Acerca de este libro

Esta es una copia digital de un libro que, durante generaciones, se ha conservado en las estanteras de una biblioteca, hasta que Google ha decidido
escanearlo como parte de un proyecto que pretende que sea posible descubrir en lnea libros de todo el mundo.
Ha sobrevivido tantos aos como para que los derechos de autor hayan expirado y el libro pase a ser de dominio pblico. El que un libro sea de
dominio pblico significa que nunca ha estado protegido por derechos de autor, o bien que el perodo legal de estos derechos ya ha expirado. Es
posible que una misma obra sea de dominio pblico en unos pases y, sin embargo, no lo sea en otros. Los libros de dominio pblico son nuestras
puertas hacia el pasado, suponen un patrimonio histrico, cultural y de conocimientos que, a menudo, resulta difcil de descubrir.
Todas las anotaciones, marcas y otras seales en los mrgenes que estn presentes en el volumen original aparecern tambin en este archivo como
testimonio del largo viaje que el libro ha recorrido desde el editor hasta la biblioteca y, finalmente, hasta usted.
Normas de uso
Google se enorgullece de poder colaborar con distintas bibliotecas para digitalizar los materiales de dominio pblico a fin de hacerlos accesibles
a todo el mundo. Los libros de dominio pblico son patrimonio de todos, nosotros somos sus humildes guardianes. No obstante, se trata de un
trabajo caro. Por este motivo, y para poder ofrecer este recurso, hemos tomado medidas para evitar que se produzca un abuso por parte de terceros
con fines comerciales, y hemos incluido restricciones tcnicas sobre las solicitudes automatizadas.
Asimismo, le pedimos que:
+ Haga un uso exclusivamente no comercial de estos archivos Hemos diseado la Bsqueda de libros de Google para el uso de particulares;
como tal, le pedimos que utilice estos archivos con fines personales, y no comerciales.
+ No enve solicitudes automatizadas Por favor, no enve solicitudes automatizadas de ningn tipo al sistema de Google. Si est llevando a
cabo una investigacin sobre traduccin automtica, reconocimiento ptico de caracteres u otros campos para los que resulte til disfrutar
de acceso a una gran cantidad de texto, por favor, envenos un mensaje. Fomentamos el uso de materiales de dominio pblico con estos
propsitos y seguro que podremos ayudarle.
+ Conserve la atribucin La filigrana de Google que ver en todos los archivos es fundamental para informar a los usuarios sobre este proyecto
y ayudarles a encontrar materiales adicionales en la Bsqueda de libros de Google. Por favor, no la elimine.
+ Mantngase siempre dentro de la legalidad Sea cual sea el uso que haga de estos materiales, recuerde que es responsable de asegurarse de
que todo lo que hace es legal. No d por sentado que, por el hecho de que una obra se considere de dominio pblico para los usuarios de
los Estados Unidos, lo ser tambin para los usuarios de otros pases. La legislacin sobre derechos de autor vara de un pas a otro, y no
podemos facilitar informacin sobre si est permitido un uso especfico de algn libro. Por favor, no suponga que la aparicin de un libro en
nuestro programa significa que se puede utilizar de igual manera en todo el mundo. La responsabilidad ante la infraccin de los derechos de
autor puede ser muy grave.
Acerca de la Bsqueda de libros de Google
El objetivo de Google consiste en organizar informacin procedente de todo el mundo y hacerla accesible y til de forma universal. El programa de
Bsqueda de libros de Google ayuda a los lectores a descubrir los libros de todo el mundo a la vez que ayuda a autores y editores a llegar a nuevas
audiencias. Podr realizar bsquedas en el texto completo de este libro en la web, en la pgina http://books.google.com

This is a reproduction of a library book that was digitized

by Google as part of an ongoing effort to preserve the
information in books and make it universally accessible.

http://books.google.com

United States Department of Commerce

National Bureau of Standards

UNITED STATES DEPARTMENT OF COMMERCE

C. R. Smith, Secretary

NATIONAL BUREAU OF STANDARDS A. V. Astin, Director

Precision Measurement and Calibration

Selected NBS Papers on

Electricity Low Frequency

F. L. Hermach and R. F. Dziuba

Editors

A compilation of previously published papers by the staff of the

National Bureau of Standards, including selected abstracts by
NBS and non-NBS authors. Issued in several volumes, see page iv.

NBS Special Publication 300 Volume 3

Issued December 1968
(Supersedes in part Handbook 77Volume II)
For sale by the Superintendent of Documents U.S. Government Printing Office
Washington, D.C. 20402 - Price $4.50

Abstract
This volume is one of an extended series which brings together
some of the previously published papers, monographs, abstracts and
bibliographies by NBS authors dealing with the precision measure
ment of specific physical quantities and the calibration of the related
metrology equipment. The contents have been selected as being useful
to the standards laboratories of the United States in tracing to NBS
standards the accuracies of measurement needed for research work,
factory production, or field evaluation.
Volume 3 contains reprints through 1967 on d-c and lowfrequency electrical measurements covering the following topics:
Electrical Units, Electrical Standards, Standard Cells, Zener Diodes,
Resistors, Resistance Apparatus, Capacitors, Inductors, Instruments,
AC-DC Transfer Standards, Transformers, Inductive Voltage Di
viders, High Voltage Measurements, Dielectric Measurements, and
Magnetic Measurements. A selected list of NBS publications on
electrical measurements is also included.
Key Words :
Electrical units; electrical standards; electrical
measurements; electrical calibrations; standard cells;
zener diodes; resistors; resistance apparatus; capac
itors; inductors; ac-dc transfer standards; current
transformers; voltage transformers; inductive volt
age dividers; high voltage measurements; dielectric
measurements; magnetic measurements.

Library of Congress Catalog Card Number: 68-60042

n

Foreword
In the 1950's the tremendous increase in industrial activity, particularly
in the missile and satellite fields, led to an unprecedented demand for precision
measurement, which, in turn, brought about the establishment of hundreds of
new standards laboratories. To aid these laboratories in transmitting the
accuracies of the national standards to the shops of industry, NBS in 1 959 gathered
together and reprinted a number of technical papers by members of its staff
describing methods of precision measurement and the design and calibration
of standards and instruments. These reprints, representing papers written
over a period of several decades, were published as NBS Handbook 77, Pre
cision Measurement and Calibration, in three volumes: Electricity and Elec
tronics; Heat and Mechanics; Optics, Metrology, and Radiation.
Some of the papers in Handbook 77 are still useful, but new theoretical
knowledge, improved materials, and increasingly complex experimental tech
niques have so advanced the art and science of measurement that a new com
pilation has become necessary. The present volume is part of a new reprint
collection, designated NBS Special Publication 300, which has been planned to
fill this need. Besides previously published papers by the NBS staff, the col
lection includes selected abstracts and references by both NBS and non-NBS
authors. It is hoped that SP 300 will serve both as a textbook and as a reference
source for the many scientists and engineers who fill responsible positions in
standards laboratories.
A. V. Astin, Director.
in

Preface
The general plan for this compilation has been reviewed by the Information Committee
of the National Conference of Standards Laboratories. The plan calls for Special Publication
300 to be published in 12 volumes having the following titles and editors:
Statistical Concepts and Procedures, H. H. Ku
Frequency and Time, A. H. Morgan
ElectricityLow Frequency, F. L. Hermach and R. F. Dziuba
ElectricityRadio Frequency, A. J. Estin
Heat, D. C. Ginnings
Temperature, J. F. Swindells
Mechanics, R. L. Bloss
Dimensional MetrologyLength and Angle, H. K. Hammond, III
Radiometry and Photometry, H. K. Hammond, III
Colorimetry and Image Optics, H. K. Hammond, III
Spectrochemical Analysis, B. F. Scribner
Ionizing Radiation, E. H. Eisenhower
This division of subject matter has been chosen to assure knowledgeable selection of
content rather than to attain uniform size. It is believed, however, that the larger volumes,
of approximately 600 pages, will still be small enough for convenient handling in the laboratory.
The compilation consists primarily of original papers by NBS authors which have been
reprinted by photoreproduction, with occasional updating of graphs or numerical data when
this has appeared desirable. In addition, some important publications by non-NBS authors,
as well as publications by NBS authors that are too long to be included, are represented by
abstracts or references; the abstracts are signed by the individuals who wrote them, unless
written by the author.
Each volume has a subject index and author index, and within each volume, contents
are grouped by subtopics to facilitate browsing. Many entries follow the recent Bureau practice
of assigning several key words or phrases to each document; these may be collated with titles
in the index. Pagination is continuous within the volume, the page numbers in the original
publications also being retained and combined with the volume page numbers, for example
100-10. The index notation 3-149 refers to volume 3, page 149 of this volume. A convenient
list of SI (Systeme International) physical units and a conversion table are to be found inside
the back cover.
The publications listed herein for which a price is indicated are available from the Super
intendent of Documents, US. Government Printing Office, Washington, D.C. 20402 (foreign
postage, one-fourth additional). Many documents in the various NBS nonperiodical series
are also available from the NBS Clearinghouse for Federal Scientific and Technical Information,
Springfield, Va. 22151. Reprints from the NBS Journal of Research or from non-NBS journals
may sometimes be obtained directly from an author.
Suggestions as to the selection of papers which should be included in future editions will
be welcome. Current developments in measurement technology at NBS are covered in annual
seminars held at either the Gaithersburg (Maryland) or the Boulder (Colorado) laboratories.
These developments are summarized, along with a running list of publications by NBS authors,
in the monthly NBS Technical News Bulletin.
H. L. Mason
Coordinator for Measurement Services
NBS Institute for Basic Standards.

Editors' Note
We have selected from the many papers published by NBS authors on electri
cal measurements those which we felt best fitted W. R. Tilley's criterion;
"References dealing with precision instruments, measurement and calibration
procedures, or treatment of measurement data, of value particularly to stand
ards and calibration laboratories or designers of laboratory instruments and
equipment." No papers or abstracts by non-NBS authors are included; how
ever, the references listed in these NBS papers can serve as a guide to some of
the excellent world-wide literature on the subject of precision measurement of
d-c and low-frequency electrical quantities.
Unfortunately, some very useful NBS Technical Notes on calibration pro
cedures could not be included because of space limitations. However, they are
listed in the abstracts sections, and are readily available from the Superintend
ent of Documents, U.S. Government Printing Office, Washington, D.C.
20402.
F. L. Hermach and
R. F. Dziuba, Editors.
v

Contents
Page
in
iv
v

Foreword
Preface
Editors' note
General Papers on Electrical Measurements
Papers
International weights and measures, C. H. Page
Electrical units, F. K. Harris
Definition of "ampere" and "magnetic constant," C. H. Page
Electrical standards and measurements, I. L. Cooter, B. L. Dunfee,
F. K. Harris, W. P. Harris, F. L. Hermach, and C. Peterson
Electrical calibration accuracies at NBS, F. L. Hermach
Suggested practices for electrical standardizing laboratories, F. B.
Silsbee
Achievement of measurement agreement among electrical standards
laboratories, F. D. Weaver
List of publications, LP 38

1
4
8
9
29
33
44
48

Abstracts
Physical entities and mathematical representation, C. H. Page
Systems of electrical units, F. B. Silsbee
Establishment and maintenance of the electrical units, F. B. Silsbee
Extension and dissemination of the electrical and magnetic units by the
National Bureau of Standards, F. B. Silsbee
Measurement of current with a Pellat-type electrodynamometer,
R. L. Driscoll
.
.
Measurement of current with the National Bureau of Standards current
balance, R. L. Driscoll and R. D. Cutkosky
Evaluation of the NBS unit of resistance based on a computable capaci
tor, R. D. Cutkosky

67
67
68
68
69
69
70

Standard Cells and Zener Diodes

Papers
Standard cells, their construction, maintenance, and characteristics,
W. J. Hamer
Oil baths for saturated standard cells, P. H. Lowrie
The operating characteristics of zener reference diodes and their measure
ments, W. G. Eicke
Making precision measurements of zener diode voltages, W. G. Eicke..

73
114
118
125

Abstracts
Designs for surveillance of the volt maintained by a small group of
saturated standard cells, W. G. Eicke and J. M. Cameron

131

Resistors and Resistance Apparatus

Papers
Stability of double-walled manganin resistors, J. L. Thomas
Errors in the series-parallel buildup of four-terminal resistors, C. H.
Page
Precision resistors and their measurement, J. L. Thomas
Calibration of potentiometers by resistance bridge methods, D. Ramaley.
VI

135
140
149
184

Resistors and Resistance ApparatusContinued

Papers
Notes on the calibration of the direct reading ratio set, P. P. B. Brooks. .
Direct ratio readings from a URS, D. Ramaley and J. F. Shafer
A method of controlling the effect of resistance in the link circuit of the
Thomson or Kelvin double bridge, D. Ramaley
Method for calibrating a standard volt box, B. L. Dunfee
Human engineering a console for the comparison of volt boxes, P. H.
Lowrie, Jr
A method for calibrating volt boxes, with analysis of volt-box selfheating characteristics, R. F. Dziuba and T. M. Souders

Page
187
198
200
204
217
222

Abstracts
Methods, apparatus, and procedures for the comparison of precision
standard resistors, F. Wenner
Notes on the design of 4-terminal resistance standards for alternating
currents, F. B. Silsbee
.
Measurement of multimegohm resistors, A. H. Scott
Calibration procedures for direct-current resistance apparatus, P. P. B.
Brooks
Practical methods for calibration of potentiometers, D. Ramaley
Some modifications in methods of calibration of universal ratio sets,
D. Ramaley
A versatile ratio instrument for the high ratio comparison of voltage or
resistance, A. E. Hess
See also, A system for accurate direct and alternating voltage measure
ments, by F. L. Hermach, J. E. Griffin, and E. S. Williams

231
231
232
232
233
233
234
290

Capacitors and Inductors

Papers
Improved ten-picofarad fused silica dielectric capacitor, R. D. Cutkosky
and L. H. Lee
.
_
.
Voltage dependence of precision air capacitors, J. Q. Shields
_.
Capacitor calibration by step-up methods, T. L. ZapL
Calibration of inductance standards in the Maxwell-Wien bridge circuit,
T. L. Zapf
.
Some techniques for measuring small mutual inductances, D. N. Homan.

237
244
254
259
265

Abstracts
New apparatus at the National Bureau of Standards for absolute capac
itance measurement, M. C. McGregor, J. F. Hersh, R. D. Cutkosky,
F. K. Harris, and F. R. Kotter
_.
..
Active and passive direct-reading ratio sets for the comparison of audio
frequency admittances, R. D. Cutkosky
Four-terminal-pair networks as precision admittance and impedance
standards, R. D. Cutkosky
.
....
Variable capacitor calibration with an inductive voltage divider bridge,
T. L. Zapf
A new type of computable inductor, C. H. Page

27 1
271
272
272
272

Instruments and AC-DC Transfer Standards

Papers
Ac-dc transfer instruments for current and voltage measurements, F. L.
Hermach
...
Thermal converters for audio-frequency voltage measurements of high
accuracy, F. L. Hermach and E. S. Williams
A system for accurate direct and alternating voltage measurements,
F. L. Hermach, J. E. Griffin, and E. S. Williams...
A comparator for thermal ac-dc transfer standards, R. S. Turgel
Calibration of peak a-c to d-c comparators D. Flach and L. A. Marzetta.
A differential thermocouple voltmeter, J. E. Griffin and F. L. Hermach.
VII

275
281
290
300
310
318

Instruments and AC-DC Transfer StandardsContinued

Abstracts
Practical aspects of the use of ac-dc transfer instruments, E. S. Williams.
Calibration of volt-ampere converters, E. S. Williams
Standard electrodynamic wattmeter and ac-dc transfer instrument, J. H.
Park and A. B. Lewis
__
Precise comparison method of testing alternating-current watthour
meters, A. W. Spinks and T. L. Zapf
Voltage ratio detector for millivolt signals, J. R. Houghton
Notes on the care and use of electrical instruments, F. D. Weaver
Transformers and Inductive Voltage Dividers
Papers
An international comparison of current-ratio standards at audio fre
quencies, B. L. Dunfee and W. J. M. Moore
Comparators for voltage transformer calibrations at NBS, W. C. Sze
An international comparison of voltage-transformer calibrations to 350
kV, F. K. Harris, W. C. Sze, N. L. Kusters, O. Petersons, and W. J. M.
Moore..
The precision measurement of transformer ratios, R. D. Cutkosky and
J. Q. Shields. _ ....
...
.
Comparison calibration of inductive voltage dividers, R. V. Lisle and
T.L. Zapf
.....
..
Comparator for calibration of inductive voltage dividers from 1 to 10
kHz, W. C. Sze
.
.
.....
An international comparison of inductive voltage divider calibrations at
400 and 1000 Hz, W. C. Sze, A. F. Dunn, and T. L. Zapf

Page
324
324
325
325
326
326

329
335

342
349
357
362
371

Abstracts
Equipment for testing current transformers, F. B. Silsbee, R. L. Smith,
N. L. Forman, and J. H. Park. . .
The design and performance of multirange current transformer standards
for audio frequencies, B. L. Dunfee
Inductive voltage dividers with calculable relative corrections, T. L.
Zapf. C. H. Chinburg, and H. K. Wolf
The calibration of inductive voltage dividers and analysis of their opera
tional characteristics, T. L. Zapf

379

High Voltage and Surge Measurements

Papers
Special shielded resistor for high-voltage d-c measurements, J. H. Park. .
Shunts and inductors for surge-current measurements, J. H. Park
Spark-gap flashover measurements for steeply rising voltage impulses, J.
H. Park and H. N. Cones

383
389
413

Abstracts
The measurement of high voltage, F. M. Defandorf

424

379

380
380

Dielectric and Magnetic Measurements

Papers
Standard tests for electrical properties, A. H. Scott
Two-terminal dielectric measurements up to 6 x 108 Hz, M. G. Broadhurst and A. J. Bur
.
Basic magnetic quantities and the measurement of the magnetic proper
ties of materials, R. S. Sanford and I. L. Cooter
The calibration of permanent magnet standards, I. L. Cooter

439
477

Abstracts
Electrical testing, A. H. Scott
An ultra low frequency bridge for dielectric measurements, D. J. Scheiber.
Low frequency dielectric behavior, W. P. Harris

485
485
486

VIII

427
431

General

Electrical

Papers

on

Measurements

Papers
International weights and measures, C. H. Page
Electrical units, F. K. Harris
Definition of "ampere" and "magnetic constant," C. H. Page
Electrical standards and measurements, I. L. Cooter, B. L. Dunfee,
F. K. Harris, W. P. Harris, F. L. Hermach, and C. Peterson
Electrical calibration accuracies at NBS, F. L. Hermach
Suggested practices for electrical standardizing laboratories, F. B.
Sflsbee.
Achievement of measurement agreement among electrical standards
laboratories, F. D. Weaver
List of publications, LP 38
_-

1
4
8
9
29
33
44
48

Abstracts
Physical entities and mathematical representation, C. H. Page
Systems of electrical units, F. B. Silsbee
Establishment and maintenance of the electrical units, F. B. Silsbee
Extension and dissemination of the electrical and magnetic units by the
National Bureau of Standards, F. B. Silsbee
Measurement of current with a Pellat-type electrodynamometer, R. L.
Driscoll
Measurement of current with the National Bureau of Standards current
balance, R. L. Driscoll and R. D. Cutkosky
Evaluation of the NBS unit of resistance based on a computable
capacitor, R. D. Cutkosky

67
67
68
68
69
69
70

D"R T? P"R FNTT /^\ instrument SOCIETY of AMERICA Preprint

X XvJCii XVLi.1 X Ig^yJP 20th Annuel ISA Conference ind Exhibit October 4-7, 1966, Lot Anfslal Number 33.3-1-65
i lo

wriu ISA. MO JW

. AUtvA Al I

International Weights and Measures

by Dr. Chester H. Page
Chief, Electricity Division
National Bureau of Standards
Washington, D. C.
ABSTRACT
A brief survey of the background, organiza
tion, and operation of the International Bureau
of Weights and Measures and related committees.

the mass of a oublo decimeter of water, for

practical reasons a solid standard was made
(of platinum-irldium alloy); this prototype was
then defined to be the kilogram.

The realization of the need for a simple

system of weights and measures, based on uni
form standards in different states and countries,
appears in the early history of our country. In
his first annual message to Congress, January
1790, Washington stated that "uniformity in the
currency, weights, and measures of the United
States is an object of great importance, and
will, I am persuaded, be duly attended to." The
matter was referred to the Secretary of State,
Thomas Jefferson; one of his proposals was "to
reduce every branch to the same decimal ratio
already established for coin, and thus bring the
calculations of the principal affairs of life
within the arithmetic of every man who can multi
ply and divide plain numbers."

For international uniformity, the various

national metric standards should be consistent;
an international conference was held in Paris
in August of 1870 to consider the advisability
of constructing new metric standards. The work
of this conference, and later ones, resulted in
the "Treaty of the Meter", or "Convention of the
Meter", signed on May 20, 1875, by 17 countries,,
including the U. S. This treaty provided for a
General Conference on Weights and Measures, meet
ing at least once every six years, an Interna
tional Committee on Weights and Measures, elected
by the General Conference and meeting every two
years, and the establishment and maintenance of
a permanent International Bureau of Weights and
Measures (BIPM). In addition to the original
primary work of verifying the new metric stan
dards the International Bureau was charged with
various duties, including:
(1) The custody of the international
prototype standards
(2) The periodic comparison of the several
national standards with the inter
national prototypes
(3) The comparison of metric standards
with other standards, such as the
British yard and pound.

In May of the same year, the National

Assembly of France requested the French Academy
of Sciences to work out a system of units suit
able for adoption by the entire world. At that
time there was a great confusion of units of
weights and measures, not only in France but
everywhere. Units in use were poorly defined
and varied from one place to the next, and from
one guild to another. In 1799, France enacted a
law making the metric system compulsory (but
later repealed it, allowing the use of both old
and new units side by sidei).

The French government donated a small his

toric manor, the Pavilion de Breteuil, located ,
in the Pare de St. Cloud, on a hillside over
looking Paris from across the Seine. This
building contains a large meeting room, offices,
and the residence of the director of the bureau.
Specially designed buildings have been erected
on the grounds, to house the metrolcgical labora
tories. The bureau has a staff of about 40 per
sons; it operated on an annual budget of about
$300,000 in 1964 but is to be escalated to about
$600,000 by 1968 with an increase of staff.

Meanwhile, in 1792, a committee of the U. S.

Senate recommended adoption of Jefferson's pro
posal of a decimal system, but no definite action
was taken.
The metric system, based on the meter as
the unit of length, the cubic decimeter as the
unit of volume, the mass of a unit volume of
water as the unit of mass, and using only deci
mal multiples and submultiples of these units,
gradually won acceptance in Europe. Eventually
the U. S. Congress (in 1866) legalized the met
ric system in this country, that is, made its
use permissive, but not compulsory, in all com
mercial and legal transactions. The metric
system has not gained favor with the man on the
street, who uses the foot, pound, gallon, and
bushel. These customary U. S. units are now
defined in terms of the meter and kilogram. Al
though the original concept of the kilogram was

The organization chart for this inter

national operation is shown in Figure 1. The
General Conference is a diplomatic operation;
our representation is via our Department of State.
The International Committee on Weights and
Measures (ICWM) is an 18-oan technical committee,
elected by the General Conference. The ICWM
nominates the director and senior staff of the
1-1

of precise measurement of high frequency

quantities.

International Bureau, and also nominates the

members of seven Advisory Committees of experts.
Host of the committee seats are held by nation
al laboratories, but there are also a few indi
viduals named as members. In these Advisory
Committees the work of the national laboratories
towards Improved standards of measurement can
be dlsoussed, and international decisions pre
pared. The International Committee has the
privilege of calling the governments partici
pating in the Convention of the Meter to meet
in a General Conference. It submits to the
General Conference proposals concerning the pro
gram and financing of the International Bureau,
and also drafts of resolutions by which the
governments agree on physical units and con
stants, or other matters of metrology, such as
research in new directions. Examples of the
recent work of this hierarchy of committees are
the redefinition of the meter in terms of the
wavelength of a particular spectral line of
krypton 86; the assignment of the frequency
9 162 631 770 hertz to the cesium oscillator;
the establishment of the thermodynamic tempera
ture scale based on the triple point of water;
and establishment of the International System
of Units, whose electrical subset is the MKSA
system incorporating the practical electrical
units instead of the confusing statvolts, abamperes, etc., found in the old physics textbooks.

Let me close with this brief sketch of

international metrology with a quotation from
a recent paper by Dr. Terrien, Director of the
International Bureau of Weights and Measures:
"Young scientists of to-day are much more aware
than formerly of the necessity and interest of
metrology, and of the Ascinating field of scien
tific research It offers; they know, or they can
learn, that the out-of-date pioture of metrology
where a patient observer spends weeks to cali
brate a divided scale, must be replaced by the
picture of a living and modem laboratory. But
what is peculiar to metrological work is perhaps
the perfection which is aimed at. No aspect of
a metrological experiment can be neglected; the
endeavour to have complete control and knowledge
of what is being done, and to get quantitative
proof with one more decimal than was possible
yesterday, that is truly science, research for
truth. Aiming at perfection is also the concern
of the artist. And that is why metrology can
be said to be not only science, but art."

The legal values of units in the united

States are those represented by or derived from,
national reference standards maintained by the
National Bureau of Standards operating under
the Treaty of the Meter. The ICWM has resolved
that "The values assigned to the national stan
dards of reference will be determined in accord
ance with the results of comparisons made with
the standards of reference of the International
Bureau." Thus we have a legal chain of connec
tion from the international actions right down
to the working standards used throughout the
U. S.
National standards are intercompared at the
International Bureau. For example, at approxi
mately three-year intervals, the major nations
submit standard cells and standard resistors
for massive lntereomparisons . Drifts in any
nation's standards are thereby spotted, and that
nation can take whatever corrective action it
sees fit. Precision three-terminal capacitors
are now being intercompared by round -robin pro
cedures. The BIPM is planning to acquire equip
ment allowing it to play a central role, such as
it does for resistance and voltage.
At a recent (May 1965) meeting of the
Advisory Committee on Electricity, arrangements
were made for international round-robin compari
sons of high-frequency measurements, such as the
measurement of power at 10 GHz. In addition,
the Advisory Committee is recommending to the
International Committee that the International
Bureau be given oompetent personnel and appro
priate equipment for getting into the business
2-2

[convention of the meter I

ESTABLISHES GCWN I BIPN
"T"
ELECTS ICWM AND MAKES INTERNATIONAL
|gcwm| DECISIONS ON STANDARDS ALLOCATES MONEY
TO BIPN

DIPLOMATIC

SETS UP ADVISORY COMNITTES AND

SUPPLIES PROPOSALS TO 6CWM,
DIRECTS OPERATION OF BIPM

TECHNICAL

TECHNICAL

ADVISORY COMMITTEES

ELECTRICITY
PHOTOMETRY
TEMPERATURE
LENCTH
TINE
IONIZING RADIATIONS
INTERNATIONAL SYSTEN

MAINTAIN & DISSEMINATE NATIONAL STANDARDS

COOPERATE WITH BIPN I ADVISORY CONNITTEES
GCWN GENERAL CONFERENCE ON WEIGHTS AND NEASURES
ICWN INTERNATIONAL CONNITTEE ON WEIGHTS AND NEASURES
BIPN INTERNATIONAL BUREAU OF WEIGHTS AND NEASURES

Fig.l

3-3

INSTRUMENT SOCIETY of AMERICA

PREPRINT^ 19th Annual ISA Conference and Exhibit October 12-15. 1964, New York
HiwxnbiUly for oil .uutmtnl, in (Ait paper rmU witk Iht aulAor.

Preprint
Number 12.1-1-61+

All rifhu rntnmL ' For pcmmioi to puMuA urru. ISA. MO Wn P.nn c, PutaW-pA. P.

ELECTRICAL UNITS
by Dr. Forest K. Harris - Non-member
Physicist
National Bureau of Standards
Washington, D. C.
ABSTRACT
The origin, development and present status of
the units of electrical measure are briefly re
viewed. The distinction between the defined
KKSA units and the "legal" units that form the
basis of the practical measuring system In the
United States Is stated, and the differences
between the two systems are examined.

Many systems of electrical units have been

used In describing electrical phenomena and the
relations between electrical quantities, and In
making electrical measurements. Some of these
units are of historical Interest only. Others
have been used by theoreticians; and their
continuing fondness for them has elements of
tradition as well as convenience. Two systems
of units are presently of concern to all of us
who must make engineering measurements --the MKSA
system of absolute units; and the "legal" units
to which any meaningful electrical measurement
in the United States must somehow be referred.
These we shall examine later, critically and in
detail. For the moment, it may be stated simply
that the MKSA units are the defined absolute
units that form the electrical portion of the
Sys teme International d ' Unite's accepted through
out the world as the common theoretical basis
for metric measuring systems; and the legal units
are the practical units maintained in the
Washington laboratories of the National Bureau
of Standards, based on physical standards that
represent the defined MKSA units as nearly as
the state of measurement art permits.
The only means we have of detecting or
measuring any electrical quantity is through a
force that it produces or an energy transfor
mation it brings about. Thus it Is natural
that one should seek to express electrical
quantities in terms of the mechanical units that
we use to evaluate force and energy (or work) ;
and It is also desirable that the system of
electrical units be compatible with the system
of mechanical units, i.e., that quantlties-such as force, torque, power, work common to
the two systems be expressed in identical units,
whatever their origin.
It Is on this basis--compatibillty between
the electrical and mechanical unlts--that the
MKSA system of electrical units was defined in
terms of the prototype mechanical units of
length, mass, and time--the Meter, Kilogram and
Second. Compatible systems of electrical and
4-1

mechanical units have been In use almost exactly

100 years. It was in 1863 that a committee of
the British Association for the Advancement of
Science, under the chairmanship of Maxwell,
recommended the adoption of an "Absolute-Practical'
system based on the centimeter -gram-second system
of electromagnetic units suggested by the earlier
work of Gauss and Weber. In this context, the
term- -absolute -practical- -meant that the units
were defined in terms of relations between
electrical and mechanical quantities, and were
of such size as to be most useful in practical
measurements. The "absolute-practical" units of
the BA committee were, in fact, decimal multiples
of the cgs electromagnetic units, and were of
such size that the product of unit current by
unit potential difference was one watt, the prac
tical unit of power in the mechanical system of
units. The BA unit of resistance the Ohm (109
cgs electromagnetic units of resistance) --was
also conveniently close to a German unit of
resistance the Siemens Elnheit* that had been
*The Siemens unit was the resistance of a mercury
column 100 cm long and 1 mm2 in section at 0C,
and was about bX smaller than the Ohm.
used to a considerable extent. Similarly, the BA
unit of potential difference the Volt (10 cgs
electromagnetic units of potential difference)
differed by little more than 201 from the emf of
a Daniell cell, which had been used to some extent,
as a unit of potential difference. The BA units
were thus of reasonable size, convenient for
measurement purposes, and soon achieved inter
national acceptance in principle. Much later
(in 1901) it was pointed out that, If the meter,
kilogram and second had been used as the defining
mechanical units rather than the centimeter, gram
and second, and if the value 10~? were assigned
as the permeability of free space rather than the
value unity that it has in the cgs electromagnetic
system, the MKS Absolute* system so defined would
*In the MKSA (meter, kilogram, second, ampere)
system presently used the ampere Is defined in
such a way that the value of the permeability of
free space is 4n x 10~7. This system has units
that are Identical with the "Rationalized MKS
System."
be identical with the Absolute-Practical units of
the British Association.

The first truly international system was the

system of "reproducible" units*, adopted by the
*In the International-Reproducible system the
ohm was represented by the resistance at 0C of
a column of mercury 106.3 cm long and about 1mm2
in section; the ampere was defined as the current
required to deposit silver from a silver nitrate
solution at a rate of 0.001118 g/sec.
International Electrical Congress of 1893, and
made the legal units of the United States in
1894 by a Congressional Act (Public Law 105).
These "International Reproducible" units* were
*not to be confused with the MKSA units of the
present Systeme International d1 Unite's.
the best approximation of the British Associa
tion's "Absolute-Practical" units possible in
the state of the art at that time. There was
med for units which were reproducible in the
sense that they would be set up independently
by a careful experimenter in his own laboratory;
but the establishment of the units soon became
the task of the various National laboratories,
established near the turn of the century. These
laboratories set up reference standards of
resistance and emf in terms of the reproducible
units; and in 1910 representatives of the British,
French, and German laboratories met with NBS
representatives in Washington to resolve certain
discrepancies in the units they maintained and
to assign their national reference standards on
a common basis. At the conclusion of this inter
national meeting at the National Bureau of
Standards, each representative returned to his
own laboratory with a group of resistance
standards and of standard cells which were in
agreement and whose values had been assigned on
the basis of the reproducible units the mercury
ohm and silver ampere. Following this, each
national laboratory maintained the electrical
units in terms of these standards.
With the development of calibration and test
services at the various national laboratories,
the need for reproducible standards diminished,
since it was easier and simpler to have a value
assigned to a standard by comparing it to a
National Reference Standard than by setting up
mercury-ohm and silver-ampere apparatus and
making the required measurements in one ' s own
laboratory.
In the decade that followed the first World
War, as Interest revived in international agree
ment of standards, it became apparent that there
were serious discrepancies between the electrical
units as maintained by some of the National
Laboratories, and further that the accuracy with
which the defined "mercury-ohm" and "silverampere" could be reproduced no longer met the
needs of science and technology. Thus a strong
movement developed to abandon the InternationalReproducible units and return to the AbsolutePractical system. This stimulated studies of

5-2

absolute measurements and the performance of

absolute ohm and ampere determinations in several
of the National laboratories.
The movement toward international agreement on
the readoption of absolute units and the reassign
ment of national reference standards on the basis
of absolute measurements was interrupted by the
second World War, but was taken up again immedi
ately afterward and culminated in the 1948
reassignment of the electrical units. The ohm.
as maintained In the United States was decreased
in value by 495 ppm (parts per million) and the
volt was decreased by 330 ppm. Historically,
this is the last change that has been made In our
electrical units. Our defined system is now the
MKSA system; and the legal units are maintained
on the basis of their 1948 assignment.
The Congressional Act that defines the elec
trical units of the United States, has the
following provision:
"It shall be the duty of the National
Bureau of Standards to establish the
values of the primary electrical units
in absolute measure, and the legal
values for these units shall be those
represented by, or derived from, national
reference standards maintained by the
National Bureau of Standards."
How then are values assigned to the "national
reference standards;" how are these standards
maintained; and how closely do the legal units of
the United States --based on these standards -agree with the defined units of the MKSA absolute
system?
If our electrical units are to be established
In such a way that the electrical and mechanical
units of energy shall be Identical, it is in
evitable that the mechanical quantities force
and distance be involved in the physical realiza
tion of the electrical units. Physical realization
of the units Involves the assignment of numerical
values to physical standards that represent and
embody the units. This Is done by means of
absolute experiments in which an electrical quan
tity is measured directly In terms of its relation
to the basic mechanical quantities length, mass
and time.
Two types of absolute measurement form the
basis of the present assignment of values to our
system of electrical units. In one experiment,
the ohm is measured as a function of length and
time; in another experiment, the mechanical force
between current-carrying conductors is used to
measure the ampere.
Absolute Ohm experiments have traditionally
evaluated the ohm in terms of a calculable
inductor and a frequency, the value of the
Inductor being computed from its measured dimen
sions. The ohm is thus established In terms of
the mechanical units of length and time --the meter
and the second . The absolute ohm experiment could
be carried out equally well In terms of a capacltive reactance In a network that takes proper
account of the phase relation between them. Here
the value of the capacitor is computed from its

measured dimension*. Such determinations have,

In fact, been made and, while their results have
not been Incorporated Into the value of the
"legal" ohm maintained by the Bureau, there Is
little doubt that any future reassignment of the
national reference standard of resistance will
rely heavily on the results of ohm experiments
that Involve a calculable capacitor. The
metrology of a calculable capacitor Is much
simpler and more direct than that of a calculable
Inductor and of equal Importance --a capacitor
can be so shielded that Its value is completely
Independent of anything outside the shield,
whereas an Inductor cannot be isolated so as to
be completely Independent of Its surroundings.
In the traditional ampere experiment, the
force between two current-carrying colls is
measured by opposing It to the gravitational
force on a known mass. The electrical force
Involves the rate of change of mutual Inductance
between the colls with linear or angular displace
ment; and the ampere Is measured In terms of the
mechanical units of length, mass and time the
miter, kilogram and second . The current, measured
In absolute amperes Is passed through a resistor
whose value Is known In absolute ohms. The
resulting voltage drop Is opposed to the electro
motive force of a standard cell, and its value
Is assigned In absolute volts . This experiment
gives us a second physical standard in terms of
which the electrical units can be preserved.
In theory the second of the two absolute
measurements required to establish our system of
electrical units could be an absolu te volt
experiment, In which the mechanical force produced
by a potential difference between the plates of
a capacitor Is measured directly and related to
the emf of a standard cell without resort to an
absolute ampere experiment. At the present time
(1964) no such absolute volt determination has
been made with sufficient accuracy to justify
Its consideration in assigning values to the
National Reference Standard, but such an experi
ment Is under study.
The National Reference S tandards consist of a
group of 10 Thomas -type 1-ohm resistors and a
group of 44 standard cells. Their values were
assigned on January 1, 1946, In terms of absolute
ohm and ampere experiments In our own and other
National Laboratories. Although additional
absolute measurements have since been reported,
there have been no reasslgnments of our National
Reference Standards. The legal units of elec
trical measure for the United States are based
on the 1946 values assigned to the National
Reference groups, and on the assumption that the
average values of these groups have remained
unchanged. The members of each group are
regularly intercompared ; and the values of the
ohm and volt as maintained at the National
Bureau of Standards are periodically compared
with the units maintained by other National
laboratories. This comparison is carried out at
the International Bureau of Weights and Measures
at Sevres (near Paris) ; and it insures that the

6-3

electrical standards used throughout the world

remain In substantial agreement. In the fifteen
years that have elapsed since the 1948 assignment,
the measured differences between the ohm main
tained by the National Bureau of Standards and
that maintained by the International Bureau have
never been greater than 1 ppm, and In only one
comparison was the measured difference greater
than 0.4 ppm. In the standard-cell comparisons,
the greatest difference between the U. S. volt
and that of the International Bureau was 3 ppm
(in 1955), and the differences in other compari
sons have been less than 2 ppm.
It must be recognized that these comparisons
do not constitute complete proof that our National
Reference Standards have kept their 1948 values.
The mast one may conclude is that, if they are
changing in value, the drift rate Is not signif
icantly different between our reference groups
and those of the International Bureau. Some
further proof of stability is needed. A means
of examining the stability of the National Refer
ence group is a repetition of an absolute
experiment.
The results of four absolute-ohm determinations
have been reported between 1948 and 1960. The
average of these determinations differs by less
than 1 ppm from the ohm as maintained by NBS, and
there is no evidence of any systematic change of
the unit with time. Preliminary results from
another ohm determination (as yet unpublished)
also Indicate that the NBS ohm differs by less
than 1 ppm from the defined absolute unit. Taking
Into account the estimated uncertainties of the
various ohm determinations, it may be said that
it is very unlikely that the difference between
the absolu te ohm and the legal unit (as maintained
by NBS) is as much as 5 ppm, and there is sub
stantial evidence that this difference does not
exceed 1 ppm.
The situation of the legal volt does not appear
to be as favorable as that of the legal ohm. The
mean of two ampere determinations made since the
1948 assignment indicate that the difference may
be about 12 ppm between the absolute ampere and
the NBS ampere the latter being determined as
the ratio of the legal volt to ohm. The estimated
uncertainty of the most recent ampere determina
tion was 6 ppm. If one accepts the evidence of
agreement between the absolute and NBS ohms, it
follows that the NBS volt may be 9-13 ppm greater
than the absolute volt.
Recently, a new means has become available
to maintain surveillance on the constancy of the
NBS ampere, and hence on the ratio of the legal
volt to ohm. If a measured current is used to
set up a magnetic field In a coll of simple
geometry and fixed dimensions (a single-layer
solenoid wound on a fused-quartz cylinder), the
precession frequency of protons in a pure water
sample immersed in this field Is a measure of the
field strength and therefore of the current in
the solenoid. A series of such measurements over
a 4-year period, using a current whose value was
always the same in terms of the NBS volt and ohm,

yielded nearly constant precession frequencies.

When the data were reduced to a fixed solenoid
te- :rature, the largest departures from a
corstant precession frequency amounted to no more
than 0.4 ppm over the 4-year Interval. Uncer
tainties In the measured dimensions of the
solenlod at the beginning and end of this period
amount to about 1 ppm In terms of the calculated
magnetic field, and this uncertainty limits any
statement concerning the constancy of the NBS
ampere. However, In the light of other evidence
concerning the constancy of the NBS ohm, It seems
likely that the "legal" volt has also remained
constant within 1 ppm over the 4-year Interval.
A number of questions may properly be asked
at this polnt--l) Apart from the challenge of
doing a better Job, what need Is there for further
refining of absolute measurement techniques?
2) Assuming that further refinements can be made
In the experimental techniques of absolute
measurement, are there other factors that limit
their accuracy? 3) Are any changes to be expected
in the National Reference Standards or In the way
they are maintained? These questions will be
considered In turn.
1) The stability of a reference standard Is
Its most Important characteristic, and an absolute
measurement is the only direct and certain test
of the stability of the National Reference
Standards. Clearly then, absolute measurements
must be repeated occasionally. Such absolute
determinations as have been reported in the
recent past had estimated uncertainties that
ranged from 2-6ppm or more. Obviously, the
smaller the uncertainty of such a determination,
the more informative it is about the stability
of the "legal" unit, maintained by the National
Reference Standard; and an immediate goal of not
more than 1 ppm uncertainty in absolute deter
minations seems appropriate. In the field of
science, the accuracy with which many of the
atomic constants can be assinged is now begin
ning to be limited by how well we know the ampere
and the volt; better assignment of these electri
cal quantities can only come from better absolute
determinations . Finally, as the measurement de
mands of space-age technology grow more severe,
It Is reasonable to suppose that more nearly
exact compatibility will be required between
mechanical and electrical units of measure.
2) Future ohm determinations will probably
be based on a calculable capacitor, and its
computed value is a function of the permittivity
of free space. In the M<SA system, it is the
permeability of space that is assigned through
the definition of the ampere (4n x 10"?). There
fore the permittivity of space must be computed
from this assigned constant and the speed of
light. Thus the value of permittivity used in
assigning a value to the calculable capacitor
can be no more certain than the accuracy with
which the speed of light is known. In an ampere
or volt determination, force is measured in
terms of the gravitational force on a known mass,
and its accuracy is limited by how well the

319-411 O - 68 - 2

7-4

acceleration of gravity is known at the place

where the experiment is performed. In a volt
determination, dependent on the measured force
between capacitor plates, the speed-of -light
uncertainty would also have to be taken Into
account. The accuracy with which the prototype
mechanical units can be reproduced and subdivided
does not impose any significant limits on absolute
electrical measurements at the present time, but
uncertainty in the speed of light and in the
acceleration of gravity are presently (1964) such
limiting factors.
3) Both the National Reference Standard of
resistance and of emf appear to be stable within
the uncertainty limits of the pertinent experimmtal information as, Indeed, they should be from
structural considerations. The 1-ohm resistors
of the National Reference Standard are thoroughly
annealed, are mounted to be free from strain, and
are sealed against contact with air and moisture.
It would appear that only the hazards of misuse
or abuse (i.e., overload or mechanical shock)
would be likely to affect their stability; and in
either instance one would expect a sudden change
in value (easily detected) followed by a period
of drift and finally stabilization about a new
value. The saturated cells of the National
Reference Standard are inherently stable, being
sealed and isolated chemical systems that are In
equilibrium when maintained at a constant temper
ature. Changes of temperature and even very
small temperature gradients across the cell limbs
must be avoided, as this would upset the equilib
rium with a consequent slow recovery period.
Another reason for avoiding temperature gradients
is the rather high thermal emf (8 \iv/C) that
would develop across the junctions between the
platinum conductors used in sealing through the
glass envelope of the cell and the copper leads
going to the measuring circuit. If proper pre
cautions are taken with temperature, it would
appear that only the hazards of misuse or abuse
remain, and most of these are more likely to
render the cell completely inoperative rather than
to exert a minor effect on its stability. One
hears occasionally the suggestion that Zener
elements with appropriate circuitry might replace
standard cells as voltage reference standards.
Certainly such a device has many attractive
features including ruggedness and ability to
withstand abuse without injury. Thus Zener refer
ence devices may eventually replace standard cells
In many laboratory operations and in most field
operations, but In their present state of develop
ment they do not compare at all favorably with
saturated standard cells either from the viewpoint
of stability or freedom from electrical noise.
Even if their development were to continue at its
present rate, it would take many years to establish
a history of stability and reliability comparable
to that of saturated standard cells. Thus there
is nothing in sight at present that suggests either
a change in the National Reference Standards of
electrical measure or in their maintenance.

Definition of "Ampere" and

"Magnetic Constant"
There appears to be some confusion
about the definition of the electrical units in
the MKSA system. Some engineers believe
that the ampere is arbitrarily defined, with
the result that the numerical value of mo is
4t10_', while others hold the view that mo
has an arbitrarily assigned value. In view of
the current 1EC (International Electrotechnical Commission) proposal to assign
the name "magnetic constant" to the quan
tity (mo) commonly called "permeability of
space," it seems desirable to clarify the
definitional basis of the electrical units.
Part of the confusion arises from two
different concepts, both called "ampere."
One meaning of ampere is the measurement
unit of current, which is a physical entity
realizable in the laboratory; the other mean
ing refers to the abstract, or symbolic, unit
in the mathematical representation of cur
rent.1 It is in this second meaning that the
ampere is chosen, for purposes of dimensional
analysis, to be independent of length, mass,
and time. This is the origin of the A in the
name MKSA system. Mathematical repre
sentation of physical relations is always
influenced by individual taste; dimensional
analysis and interpretation vary widely, and
all this plays a role in the "ampere" as an
abstract unit. The measurement unit, on the
other hand, must be the same for all lab
oratories, so I shall take the measurement
unit problem as basic, and the abstraction
as secondary.
There are several experimental situa
tions that can be used to define the nu
merical measure of a current. After one of
these is chosen, it can be used in an opera
tional procedure for realizing the unit of
current (ampere). One could use an experi
ment which determines current in terms of
a natural quantum of charge, as was done
in the old international silver coulometer
definition, or one could use magnetic effects,
such as the interaction force between two
currents. The latter is the one that has been
adopted for all "absolute" unit systems.
The electromagnetically produced force
between conductors is known to satisfy a
certain proportionality; the magnetic con
stant (r)' is the name given to the pro
portionality constant when the relation is
written in the form
4xr*
expressing the force between two current
elements in vacuo. If the ampere were de
fined independently of this relation, then the
measure {1%,} would be subject to experi
mental determination.* If, on the other
hand {r} is arbitrarily chosen, the ampere
is thereby defined because we anc deduce a

recipe (operational description) for realizing

a current of specified measure, e.g., a current
such that {/} 1. This current is then de
fined to be the ampere. Superficially, it
would appear that we could choose this
recipe as the definition of the ampere, and
thus define the ampere without assigning a
value to {r,}. The change is purely
psychological, not logical; it corresponds to
a change of word order in the definition of
{r}, since {r,} can be computed from
such a definition of the ampere and no
experiment need be performed.
The old international ampere based on
the silver nitrate coulometer was independ
ent of r. In the Systeme International
d'Unites, however, the ampere is actually
defined by assigning the number 4x10"' as
the measure of F. in the interaction-force
equation.
Choosing {r} and specifying the
equation in which r_ appears, implicitly
defines the ampere. The corresponding ex
plicit definition is the familiar "legal" form
relating the ampere to a force of 2.10-'
newton per meter between infinitely long
parallel wires in empty space.
The above viewpoint is not only logically
correct but also historically accurate. In
1938, the IEC Subcommittee EMMU (now
TC 24) followed the 1935 actions of the In
ternational Committee on Weight and
Measures, and its Advisory Committee on
Electricity, by adopting the assignment of
4-10~' as the connecting link between elec
trical and mechanical units.
Symbolic Unit
The quantity equation representation of
the facts of physics are based on a represen
tation of physical entities as mathematical
elements in a many-dimensional vector
space.1-4 Such an abstraction is much more
sophisticated than a quantitative descrip
tion of physics by measure equations alone.
It is not a consequence of experimental
physics, but a useful intellectual tool in
vented by man. Like any other tool, its de
sign is affected by individual esthetic pref
erences; in fact, many people object to the
abstraction per se.
Among practitioners of abstraction,
there is universal agreement on the choice
of mass, length, and time as orthogonal axes
in the vector space of quantity calculus.
The addition of electrical entities to this
mechanical subspace requires an additional
dimension. Various tastes prefer r,*, r /,
or Q as the fourth axis. Some physicists
even prefer to ignore this fourth axis and
interpret all electrical quantities as being
dimensionally only mechanical quantities.
In 1938, when TC 24 chose the measure
4irl0~7 as the defining link between me
chanical units and the Giorgi System of
electrical units, it made no decision on

matters of taste. In 19S0, however, this

committee decided that the ampere could
be taken as the "fourth principal unit,*
and accepted as the definition of the am
pere, the measurement unit definition ap
plicable to infinitely long parallel conduc
tors, deducible from the force relation by
assigning {r} "4rl0"'.
Thus the ampere, as a measurement unit,
is defined by the IEC both implicitly and
explicitly; as a symbolic unit, it has not
been "defined" but has been "accepted" as
a "principal unit" on which to "base" the
Giorgi system.
It is difficult to define symbolic units.
We can define symbolic or abstract quan
tities as mathematical elements that "corre
spond" (in some unspecifiable manner) to
observable physical quantities, or as mathe
matical entities that "represent" the corre
sponding physical entities. The symbolic
units then "correspond" to the measurement
units in the same unspecifiable manner. If
the quantity equations are to be formally
the same as the measure equations, there
are compatibility relations on products of
symbolic quantities and units, but there is
an independent ("basic," "principal,* "fun
damental") set of symbolic units that can
be arbitrarily chosen. This set is customarily
chosen to comprise three mechanical units
and one electrical unit. The remaining
symbolic units of electromagnetism are de
fined mathematically in terms of these four.
These basic four are in principle undefinatde,
they are postulated units in terms of which
the remainder are defined. Thus the choice
of the ampere as the "fourth principal unit*
is tantamount to stating that the ampere is
the only symbolic electromagnetic unit
which cannot be defined.
In summary, the symbol r has been
defined as the proportionality constant in a
particular equation; its measure has been
arbitrarily assigned; the consequent explicit
form of the definition of the ampere (meas
urement unit) has been adopted, and the
abstract mathematical element (symbolic
ampere) representing a physical ampere has
been chosen as the undefinable "intuitive*
element of the electrical set.
Chester H. Pace
Electricity Div.
U. S. Dept. of Commerce
National Bureau of Standards
Washington, D. C.
Manuscript received September I. 1964.
1 C. H . Page. "Physical entitles and mathematical
representation." /. Sri. Nat. Bur. St., vol. 6SB.
pp. 227-235;
October-December,
1*61.employed but is
The symbol
no is traditionally
gradually being replaced by I'm by reason of the mis
leading implications of a "permeability* symbol.
1 The notation U I indicates the numerical value
of x, without statement of its unit or dimension. See
Page.'
Wilhelm Quade. Uber die algebraische Struktur
des Grossenkalkllls der Physik, Abhandlungen der
Braunschweigischen Wissenschaftlichen Gesellschait.
Band1 The
XIII,
1961.constant, also represented by .
electric

Reprinted from the PROCEEDINGS OF THE IEEE

VOL. 53, NO. 1, JANUARY, 1065

8-100

ELECTRICAL STANDARDS AND MEASUREMENTS*

I. L. Cooter, B. L. Dunfee, F. K. Harris, W. P. Harris, F. L. Hermach, C. Peterson
National Bureau of Standards
Introduction - Many millions of measurements of
International agreement- -and was based largely on
electrical quantities made every year throughout
the results of absolute ohm and ampere determina
tions made at the National Bureau of Standards and
the United States are based on a consistent system
of reference standards established and maintained
at the National Physical Laboratory of Great
by the National Bureau of Standards in cooperation
Britain. Absolute measurements of the quality
with the National Laboratories of other countries.
needed for assignment of basic reference standards
require access to the fundamental mechanical
Our National Measurement System derives from the
International System in terms of which the legal
reference standards and, thus, can be carried out
effectively only at the larger national laborato
units of measure of 38 countries are defined.
ries of the world; and, while a number of the
Among the duties of the Bureau of Standards are
other national laboratories have programs of
included the establishment of electrical units
absolute measurements, none of their results were
that are consistent with the International System
and the assignment of values to reference standards available in final form at the time of the 1948
revision. A number of additional results have
which will preserve the legal units and serve as
since been reported from various national labora
the starting point of all the measurement chains
tories (including our own), and there is substan
whose final links are the myriads of electrical
tial evidence that the ampere determinations on
measurements made daily throughout the nation.
which the 1948 assignments were based were in error
It is only by using a consistent system that
by about 10 to 12 parts in a million. Thus, it is
scientists in different laboratories can talk
reasonable to suppose that when further evidence
in the same terms, or that one can expect proper
has been accumulated, the assignments of the basic
functioning of devices assembled from components
reference standards may be modified to bring them
made in different places at different times, or
more closely into agreement with the defined units
that competing sources of energy can be fairly
of the International System.
compared, or that complex industrial processes
can be made to yield a product of constant and
Over the past century- - s ince reference stan
acceptable quality. We shall here review briefly
dards
of resistance were first assigned by Maxwell
the means by which the Bureau has established the
and
his
colleagues of the British Association for
legal units of electrical measure and shall
the Advancement of Science--the classical ohm
discuss the experimental procedures by which the
experiment has involved the construction of a
values of the basic reference standards are made
self- or mutual- Inductor whose value could be com
available as initial links In all the measurement
puted from its measured dimensions and the assigned
chains that form the National Measurement System.
value of space permeablllty--unity In the cgs
electromagnetic system or 4tt x 10" ^ in the MKSA
System of Units. The reactance of this calculable
Fundamentals - By international agreement, the
Inductor at a known frequency was then compared to
official basis for electrical measurements
a resistance in a suitable network, and the value
throughout the world is the International System
of the resistor was thus assigned in terms of the
of Units--based on the meter, kilogram, and
units of length and time. Since the relative
second-- together with a value of the unit of
positions of all the conductor turns In the induc
electric current (the ampere) defined in such a
tor must be determined, the dimensional metrology
way that the permeability of free space has the
required for an absolute ohm experiment has been
value 4n x 10"^. One constraint on the system-very elaborate, time-consuming, and tedious.
essential to its compatibility with the correspon
ding system of mechanical units--is that the
Within the past decade, a new type of ohm
electrical unit of energy be identical with the
determination of appropriate accuracy has become
feasible, based on a calculable capacitor. A new
mechanical unit of work. A second constraint is
theorem in electrostatics, published by Thompson
that the physical laws which state the basic
relations between the units be expressed in the
and Lampard of the Australian National Standards
simplest possible terms, without the need (in
Laboratory, has led to a capacitor construction
whose configuration is such that only a single
most cases) of numerical proportionality con
stants other than unity. The first of these
length determination is critical to the calcula
constraints leads Inevitably to the experimental
tion of its capacitance. One can compare a
realization of the tie between the electrical and
capacltive reactance and resistance at a known
mechanical units in terms of units of force and
frequency and determine the resistance in units of
length--the components of work. The second con
length and time and the permittivity of space.
straint simplifies the derivation of all the
The dimensional metrology is very much simpler than
remaining electrical units, once we have fixed two
for the calculable inductor, and the experiment
of them by "absolute" experiments in terms of the
appears capable of a precision at least an order of
mechanical units. In the present context, an
magnitude better. In fact, the element of greatest
absolute experiment is one in which the value of
uncertainty in this procedure at the present time
an electrical quantity is established directly in
is the speed of light--it is from this experimental
terms of mechanical units.
constant and the assigned value of space permeabil
ity that the value of space permittivity must be
The most recent assignment of our basic refer
obtained for the calculation of capacitance. An
ence standards was made in 1948--consequent to an
*Expanded version, with corrections, of "Electrical Standards and Measurements," by I. L. Cooter, B. L.
Dunfee, F. K. Harris, W. P. Harris, F. L. Hermach, and C. Peterson, E iectro-Technology 79, S3 (Jan. 1967).
9-1

absolute ohm determination based on a calculable

capacitor has been reported which checks the
present assignment of the National Reference Stan
dard within a part In a million, and other deter
minations are under way that have not yet been
reported.
The second absolute experiment required to
assign the electrical reference standards In terms
of mechanical units has been classically the
absolute ampe re determination. In this experiment,
the value of a current In two colls Is measured In
terms of the force between them, using a balance In
which this force Is opposed to the gravitational
force on a known mass. Thus--from the balancing
mass and the local acceleration of gravlty--the
ampere Is expressed In units of length, mass, and
time. In this experiment, a resistance standard Is
connected In series with the coll system, and the
potential drop produced In It by the measured cur
rent is compared to the electromotive force of a
standard cell. By this means, the emf of the cell
Is assigned In absolute volts, and the cell pre
serves the result of the experiment. A variety of
coll shapes and arrangements have been used in
current balances at the Bureau and in several of
the other national laboratories. In any arrange
ment, dimensional measurements are required to
determine relative coil size and positions, since
the statement of force between the colls requires
that the change of mutual Inductance with movingcoil displacement be evaluated.
The units realized by these absolute measure
ments are preserved at the Bureau by a group of
10 stable 1-ohm resistance standards and a group
of 44 saturated cadmium-mercury standard cells.
To maintain the legal units, the assumption is
made that the mean of each reference group does not
change with time. To guard against drift of indi
LENGTH
T
Kr86wove - length

Reference Standards for Electrical Quantities The many electric and magnetic quantities used in
science, engineering, and industrycurrent,
voltage, inductance, capacitance, resistance,
magnetic flux, etc. are connected to each other
and to the various mechanical quantities by exact
mathematical relations. Similarly, the various
units --ampe re , volt, henry, farad, ohm, weber,
etc. --in terms of which these quantities are
measured, are linked by a parallel structure of
exact definitions. Starting with any group of
independent units, designated as "base" units,
it is possible to derive definitions of all the
others by successively applying appropriate
mathematical relations. Both the choice of "base"
units and the particular order in which the "de
rived" units are successively defined may be made

TIME INTERVAL
Space
permeability

Cs'" otomic beom

Frequency

. Mechanical
Quantities

MASS
Prototype
Kilogram no.20

" fixing the MKSA units

_ Laboratory
Reference Standards

Quartz oscillator &

frequency divider

Prototype meter no 27

l ,
i
I
I'

viduals, they are intercompared regularly; to guard

against group drift relative to the legal units of
other countries, triennial lntercomparlsona are
made at the International Bureau of Weights and
Measures at Sevres, France. Further assurance
that there is no Important group drift is obtained
by the occasional repetition of absolute ampere
and ohm determinations, and by proton precession
frequency measurements which will be discussed
below.
The relations between the electrical and
mechanical units, and the steps Involved in as
signing values to basic reference standards are
shown in the block diagram of Figure 1. It will
be noted that this diagram Indicates the units
and steps currently in use (1966) rather than
those which served In the absolute measurements
on which the 1948 reference-standard assignments
were based.

1
Local acceleration
of gravity

Speed of light

1
Space
permittivity

Measured physical
quantity

t
^ Computed from
"*
It. I

Calculable capacitor

r-j Calculable Inductor (i i

I 1
T
X
i
Absolute Ohm
Absolute Ampere
determination
determination
t
i=l
i r
Standard
Standard
Standard
Standard
Capacitors
Resistors
Inductors
Cell*
P
Figure 1. The Derivation of the Electrical Units.
It should be noted that the terms "magnetic constant" and "electric constant"
red by IEC to the terms "space permeability" and "6pace permittivity" used in
10-2

Computed from
measured dimensH
Determined
etectricol unit*
Reference
Standards

are prefer
this chart.

Figure 2.

Figure 3.
Schematic diagram of a transformerratio-arm bridge for comparing capaci
tors .

Calculable Cross Capacitor assembled

from gage rods.

arbitrarily, as taste, convenience, or practical

experience suggests. In the discussion which
follows, we shall examine In some detail the steps
by which the Bureau arrives at reference standards
for the measurement of various electrical quanti
ties. Whether these steps appear straight forward
or devious In particular cases depends solely on
the exigencies of laboratory practice.
Capacitance - For many years, the unit of capaci
tance was derived from the ohm and second by means
of a Maxwell commutator bridge. Within the past
decade, the capacitance theorem of Thompson and
Lampard has led to the construction of calculable
capacitors whose values can be assigned from length
and space permittivity with an uncertainty no more
than a part In a million. During the same Interval,
the techniques of transformer ratio-arm bridges
have been developed to an extent that permits com
parison of equal capacitors to parts In 108 and
extension of the unit upward through several
decades at the part per million level. Finally,
a stable capacitor design has been developed at
the 10 pF level whose drift rate appears to be
less than 1 x 10"' per year. Thus, the unit of
capacitance can now be established by absolute
measure as a part of the ohm determination and
values assigned directly from the calculable
capacitor to stable reference standards which can
effectively preserve the unit.
The theorem of Thompson and Lampard may be
stated as follows: "If four Infinite cylindrical
conductors of arbitrary sections are assembled
with their generators parallel, to form a com
pletely enclosed hollow cylinder in such a way
that the internal cross-capacitances per unit
length are equal, then In vacuum^ these crosscapacitances are each ln2/(ugr'V )."
In the
International System, where
has the assigned
value 4n x 10"' and V is the speed of light in
meters per second, the cross-capacitances are
expressed in farads per meter. For small in
equalities of cross-capacitance between pairs of
opposite electrodes, it can be shown that the
departure of their mean from the theoretical value

C| and
are capacitors being compared.
C3 is capacitor used to balance differ
ence between Cx and C2 , by varying tap
point on inductive divider D2 connected
to 0.1 tap of ratio transformer. R is
resistor used to balance phase differ
ence between Cx and C2 by varying tap
point of divider Dx and current divider
ratio CA/(C4 + C5) .
is approximately 0.087 (A/C)2, where A is their
difference and C their mean. Thus, if the in
equality is no more than a part in 10^, the error
In the computed value is less than 1 in lo'. A
practical realization of such a capacitor consists
of four equal closely-spaced cylindrical rods with
their axes parallel and at the corners of a square.
Arranged as a 3- terminal capacitor, the internal
cross-capacitance amounts to 1 pF for a length
about a half meter, and, if end effects are elimi
nated, this capacitance can be computed nearly as
accurately as the length can be measured. Figure
2 shows the calculable capacitor, constructed of
gage rods of known length, which is at present the
basis of the Bureau's capacitance unit. Another
calculable capacitor is now under construction
whose active length will be measured interferometrically; It is expected that its use will improve
the accuracy of the length measurement by at least
an order of magnitude. Figure 3 is a schematic
diagram of a transformer ratio-arm bridge such as
is used for comparing 3- terminal capacitors and in
extending the measurement range upward from the
calculable capacitor. The ratio-arm transformer
used in this bridge is of special design, and can
provide a bridge ratio which is known and stable to
a part in 108.
Figure 4 shows the construction of the 10 pF
capacitors used to preserve the unit. The capaci
tor Is a fused silica disc with fired-on sliver
electrodes. The measured direct capacitance is
between the plane faces of the disc, and che
cylindrical surface forms a guard electrode. The

11- 3

Figure 4.

The bridge has eight built-in fixed 3-terminal.

capacitors ranging in value from 100 pF to 10'^ pF.
Each of these capacitors can be connected to
accurate decimal fractions (0.0, 0.1, 0.2
1.0)
of the transformer secondary voltage, the adjust
ment being provided by lever switches having con
tacts connected to taps on the transformer. Thus,
eight decades of capacitance balancing adjustment
are available. Means are also provided for
balancing the loss component of capacitors being
calibrated. The maximum direct-reading range of
the bridge Is 1111.1111 pF. Small 3-termlnal
capacitors having a value of 10" pF can be
measured In this bridge with a total uncertainty
of 0.3%. While this bridge was designed for the
measurement of direct capacitance in 3-termlnal
capacitors, the use of auxiliary fixtures make it
possible to perform measurements on certain
2- terminal capacitors. The design of the trans
former Incorporated in this bridge limits its
operation-- for best measurements-- to the frequency
range from 100 Hz to 2 kHz. Other transformers
could be built to accommodate higher frequencies.
The second transformer ratio-arm bridge is a
commercial version of the first, but with only six
decades of capacitance balancing adjustment. Its
chief advantage lies in the fact that 2-terminal
measurements can be made conveniently and directly
to a maximum of 10 uF. Indeed, with appropriate
external build-up techniques, low- frequency
measurements can be extended to 10 x 10 uF, using
the bridge as a "tare" and relying upon a previ
ously calibrated 10 uF capacitor as a substitution
standard connected close to the terminals of the
large capacitor being calibrated.
The third bridge has resistance ratio arms.
It covers a wide frequency range-- from 50 Hz to
100 kHz and is used for the routine calibration of
capacitors having values up to 1.10 uF.
Each of these bridges can be calibrated by
means of working standards of capacitance evalua
ted in terms of the unit of capacitance preserved
with the group of 10 pF standards already described.
For the most accurate calibration of standards adJusted close to the decimal scale of values, no
reliance need be placed upon the bridge calibration.
In fact, the best calibration of any physical
standard is made by comparing it with a similar
standard of known value, using a simple substitu
tion technique. For capacitance measurements then,
the bridge is used at unity ratio merely to measure
the small difference between the two standards, and
no bridge errors of significance are introduced if
the upper dial settings of the bridge are not
changed during the comparison.

Cross-section of stable 10 pF capaci

tance standard with fused silica di
electric.

mounting protects the disc from mechanical shock,

and the housing contains a resistance thermometer
in addition to the capacitor so that proper allow
ance can be made for the temperature coefficient
of capacitance.
The Capacitance Scale - The Bureau uses three
types of capacitance bridges in the calibration
of capacitance standards and high-quality decade
capacitance boxes in the audio- frequency range
rs Hz to 30 kHz. Details of one of these bridges-a transformer ratio-arm bridge of NBS design and
constructionare shown schematically In Figure 5.

Dielectric Constant - (or relative permittivity)

is a property of materials that is closely related
to capacitance measurements. It is generally deter
mined as the ratio of the capacitance of a pair of
electrodes in a cell with the material occupying
all space in which there is appreciable electric
field, to the capacitance of the same electrode
configuration In vacuum. A typical cell for
dielectric-constant measurements consists of a
pair of flat parallel electrodes, a means of
maintaining and measuring their positions, and
provisions for inserting and removing a specimen:
If one electrode Is guarded (a 3-terminal capacl-

12-4

tor), the field can be more easily confined to the

volume occupied by the specimen.

a. Schematic of low-range decade capacitor

b. Basic conductance balance control circuit

MICROMHOS
OR
EQUIVALENT PICOFARADS

The practical frequency range for 3-termlnal

measurements is 10" to 10^ Hz, the lower limit be
ing determined by the time required to achieve a
bridge balance and the upper limit resulting from
difficulties In maintaining the guard and guarded
electrode at the same potential. The accuracy of
3- terminal measurements falls off from a part In
106 or 107 at 1 kHz to about a part in 10-5 at the
lower and upper frequency limits. Two-terminal
measurements can be made up to 1 GHz with an
accuracy of about 5 parts in Kp.
The dielectric constant of a fluid can be
determined to an accuracy limited only by the
mechanical stability of the cell and the accuracy
with which the cell capacitance can be measured.
Very accurate measurements can be made on fluids
by using a Thotnpson-Lampard capacitor in a con
stant temperature enclosure into which the
fluid can be Introduced and removed. In the
measurement of solid specimens, additional errors
result from practical difficulties in Insuring
that the specimen fills all the space between the
electrodes and in determining the specimen dimen
sions which enter into the calculation of the
dielectric constant. Typical solid disc specimens
are 1-5 mm thick. Discs of hard mechanically
stable materials such as ceramics can be machined
very flat and their thickness measured to better
than 0.1 micron. Uniform discs of less stable
materials, such as plastics, are difficult to
prepare and uncertainties in thickness measure
ment limit the accuracy of dielectric constant
determination to about 0.1X.
For thin films, the accuracy of dielectric
constant determination is severely limited by
uncertainty in thickness measurements. This sen
sitivity to thickness error can be reduced by
immersing the solid specimen in a fluid of similar
dielectric constant during the measurement, and
completely eliminated if the dielectric constants
are equal. A 2- fluid immersion technique Involving
four capacitance measurements makes possible the
calculation of the dielectric constant of solid
disc specimens Independent of area and thickness
measurements. This technique is valuable for very
accurate measurements on reference specimens that
will be used for checking measuring apparatus In
industrial and other laboratories.
Resistance - The unit of resistance is preserved
by a National Reference Group of ten 1-ohm resistors
hermetically sealed against air and moisture in
double-walled enclosures after being fully
annealed In an inert atmosphere at 400C. Periodic
lntercomparisons show that their Individual values
do not vary with respect to the group mean by as
much as 1 part In 10' per year, and recent abso
lute-ohm determinations Indicate that the group
mean is within 1 part In 106 of the defined ab
solute ohm.

c. Details of conductance balance control

circuit
Figure 5.

Details of NBS capacitance bridge.

The Resistance Scale - The primary instrument used

at the Bureau for the measurement of resistance Is
a precision bridge built In the Bureau shops in
1918, which incorporates many features developed
as a result of earlier experience in the field of
precise measurement. The more critical resistance
sections consist of hermetically sealed colls; the
dial-adjustment resistances are of the Waidner-

Figure 6.

Circuit for establishing 10/1 resis

tance ratio.
Wolff type, made by changing the shunting around
fixed resistors so that the effects of contact re
sistances and thermal emfs generated In the decade
switches are greatly reduced; all resistors and
contacts operate In oil In a temperature-controlled
bath; and the circuit Is shielded from stray
currents. The bridge may be used either as a
Wheatstone or a Kelvin bridge and Is commonly used
with a ratio that Is nominally either 1/1 or 10/1.
This ratio can be adjusted over a limited range
(0.57.) In the vicinity of the nominal value. The
adjustment Is made by means of four decade dials,
one step on the lowest dial corresponding to a shift
of one part In 10^ In ratio. The detector sensi
tivity Is such that parts In 10? can be Interpolated.
The step that Is fundamental to extending the
resistance scale Is the establishment of an accurate
10/1 ratio between the ratio arms of the bridge.
This Is done with a special assemblage of resistors,
shown schematically in Figure 6. Six of these
resistors are nominally 150 ohms each, and the
seventh Is 50 ohms. By successive substitution
In the precision bridge, the relative values of
the six resistors can be measured. Also by
substitution the values of the resistances of
nominally 50 ohms each, formed by connecting In
parallel each of the two sets of three 150- ohm
units, Is determined relative to the 50-ohm resis
tor with high accuracy. The three 150-ohm units
of one set are then connected In series with each
other and with the 50-ohm unit to form a 500-ohm
group. The three 150-ohm units of the other set
are connected in parallel to form a 50-ohm group.
These two groups, having a ratio close to 10/1,
are connected to form two arms of the precision
bridge, the remaining two arms being the adjustable
ratio described above. The bridge Is then balanced
and the setting
of the adjustable ratio Is noted.
The paralleling links are then shifted so that the
150-ohm colls of the first set are connected In
parallel to form a 50-ohm group, and those of the
second set are in series so that, with the 50-ohm
unit, they form a 500-ohm group. The groups are
turned end for end and again connected in the
bridge, and a second balance setting N'2 is noted.
The mean of
and N2 is adjusted by applying a
small correction derived from the observed dif
ferences between the 50-ohm coil and the mean of
the groups of three in parallel. A further cor
rection for the paralleling links is applied and,
with these corrections taken into account, the
setting of the bridge ratio dials for a 10/1 ratio
is then known to about 3 parts in 107. Other pro

cedures are available for establishing an accurate

10/1 ratio, and are sometimes used to check the
result of the regularly used procedure described
aboveagreement within a few parts in 107 Is
usually achieved.
The bridge, thus calibrated, is used to
measure in successive steps the values of standard
resistors of 10, 100, 1000, or 10,000 ohms, and
similarly to step down to 0.1, 0.01, 0.001, and
0.0001 ohm standards. Resistances of inter
mediate value are occasionally measured by using
analogous series or parallel combinations of
nominally equal resistors to evaluate other bridge
ratios that Involve small integers.
When making precise measurements with resis
tors of 1 ohm or less, it Is essential to use
resistors of the 4- terminal type to eliminate the
effects of contact resistance. The Kelvin bridge
or some equivalent circuit has proved to be
generally the most suitable for the comparison of
such resistors. Of the various procedures that
can be used to balance out the effects of the
resistance of potential leads, that which involves
applying a short-circuit across the main ratio
arms is commonly used.
At the other extreme, the measurement of high
resistance merges into measurement of the resis
tivity of insulating materials. High-grade wirewound elements are available up to 10 megohms.
Units of higher individual value are either thin
deposited films of conducting material on an in
sulating substrate or a filamentary mixture of
conducting and insulating particles. Either type
Is likely to show large changes of resistance with
temperature, voltage, and (unless very carefully
protected) ambient humidity. Also, their values
are much less stable with time than wire-wound
resistors. Thus, measurement methods of high
precision would be Inappropriate.
A series of standards, suitable for the up
ward extension of the resistance scale by sub
stitution techniques is made in the following way.
Starting with a group of ten 0.1-megohm units which
can be connected in parallel for comparison with a
10, 000- ohm standard, they can then be connected in
series and used, with the same correction in pro
portional parts, to establish a 1-megohm standard.
The error Involved in this step up is a tenth of
the sum of squares of the departures of the indi
vidual values from the group mean, so that the
step-up error becomes negligible for a group of
well-matched units. For example, If the spread of
the Individual units is 0.17., the step-up error In
a parallel-to-series reconnection is less than a
part in 10^. Similarly, groups of 1- and 10-megohm
resistors are used to extend the range in a similar
manner to 10^- ohms. Up to the latter value, com
parisons are made in a nominally 1/1 megohm bridge,
using a substitution technique, with restrictions
on the applied voltage In the upper range because
of the rather large voltage coefficients of the
film type of resistors, ranging from 1.3 x 10"6
V2 for the 107-109 group to 2.6 x 10"6 V2 for the
10^-10^ group. These approximations are valid
only for relatively low voltages--at 40 volts, a
neglected term involving
may amount to 20-30
parts in 106.
For yet higher resistances, the charge carried
by the current in the resistor under test is ac-

14-6

cumulated on an adjustable alr-dlelectrlc capaci

tor. The rate of accumulation may be measured by
changing the capacitance at a rate which maintains
a null indication on an electrometer used as a de
tector. An accuracy of 0.27. can be achieved up to
1013 ohms with this method at 1.5 V.
Voltage - The unit of voltage Is preserved by a
National Reference Group of saturated cadmium
cells, whose values were assigned In 1948 In terms
of the absolute ampere and ohm determinations avallable at that time. This reference group-whlch at present consists of 44 eel Is- -has been
selected from a much larger stock of cells that
have been made In the Bureau laboratories at
various times, using very carefully purified
materials. The cells of the reference group are
Intercompared regularly about once a month; and
every three years, a small group Is taken to the
International Bureau at Sevres, France, for com
parison with the International Reference Group
maintained there, and also with the groups of
cells carried there from the other national
laboratories .
In addition to the above comparisons-- In
which differences between the cells are measured
with a microvolt potentiometer a further Indirect
check is made on the stability of the reference
group by periodic measurements of proton preces
sion frequency in the field produced by the NBS
ampere- -defined as the ratio of the NBS volt to
ohm- - in a solenoid of fixed dimensions. In view
of other available evidence that the NBS refer
ence ohm is stable, the constancy of this measured
frequency becomes a measure of the stability of
the NBS reference volt. Over the past six years,
since this measurement technique became available,
the NBS volt--as maintained by the mean of the
National Reference Group of cellsappears to be
stable within a part In 10*.
Proton Gyromagnetlc Ratio - It has been demon
strated that if a substance whose nuclei have
non-zero spin Is placed in a steady magnetic field
of flux density B, on which is superposed a weak
radio- frequency field, the material will show a
resonance absorption at a frequency v, given by
v - yB/2TT, where v is the gyromagnetlc ratio of
the atomic nuclei of the substance used. In a
converse- -free precess lon--experiment the protons
of a water sample are first polarized in a strong
magnetic field and the sample is then moved to a
field of much lower intensity where the precession
frequency is measured, using the alternating
voltage induced in a pick-up coll by the precessing protons In the time (3-4 seconds) before
their alignments again become random. As in the
absorption experiment, the precession frequency
is a measure of the magnetic field, since the
gyromagnetlc ratio for a nucleus Is an atomic
constant that Is unaffected by the field strength.
From the free precession experiment, a value- "Y/2TT - 4.25759 x 10? hertz/tesla, uncorrected for
diamagnetism--has been determined for the proton
gyromagnetlc ratio. The absorption experiment is
appropriate for the measurement of fields greater
than about 0.02 tesla, and the free precession
experiment Is appropriate for moderately low In
tensity fields. In the free precession experiment

described above as a means for periodically moni

toring the stability of the NBS ampere, the field
of the solenoid is about 0.0012 tesla (12 gauss).
The Direct Voltage Scale - The value of a voltage
other than that of a standard cell, may be deter
mined with a ratio device, and thus referred to
the assigned emf of the reference cell. In the
general range from 1.5V down to a microvolt or
less, the ratio device used for this purpose Is a
potentiometer. Since potentiometers measure, in
effect, merely the ratio of the unknown emf to
that of the standard cell, their calibration need
not involve any direct measurement of resistance,
but merely the determination of ratios of resis
tance. Hence, for the testing of potentiometers,
It has been found convenient to use a universal
ratio set (a precision voltage divider) which is
equivalent to a slide wire having a total resis
tance of 2111.1 ohms, with a tap adjustable in
steps of 0.01 to 0.0001 ohm, depending on model
available and resolution required. In the very
low voltage range, from a few microvolts downward
to a nanovolt, extreme care must be taken In the
design and operation of the special-purpose poten
tiometers used, to avoid the inclusion of para
sitic emfs in the measuring circuit. The limited
sensitivity available In the usual types of
d'Arsonval galvanometers precludes their use at
levels much below a microvolt, and other detectors
are used, either a moving-coll galvanometer equip
ped with a photovoltaic or photoresistive network
to amplify the unbalance signal or a dc-to-ac con
version device followed by a low-level a-c ampli
fier, and finally an ac-to-dc conversion for the
signal read-out. In such detector systems, also,
extreme precautions must be observed to avoid
parasitic emfs or electrical noise that would mask
signals at a level higher than the resolution limit
Imposed by the Brownlan motion of the moving-coil
galvanometer or the Johnson noise of the measuring
circuit.
The extension of the direct voltage scale to
values higher than 1.5 volts, while still using the
potentiometer, is achieved through special resis
tive voltage dividers (commonly, but inappropri
ately, called "volt boxes") having a limited number
of discrete ratios for operation at convenient
voltage levels. In their simplest form, these
dividers comprise a group of resistors connected
in series. Taps are brought out at the ends and
from appropriate intermediate points to binding
posts to accommodate the input voltages; at the
zero end a section is brought out as the "lowside" which operates through a detector Into a
potentiometer. Twin binding posts are provided at
the ground or zero end so that the divider is used
as a 4-termlnal device (i.e., separate "input" and
"potentiometer" terminal pairs). If the input
resistance is (Ra + Rb) , where Rb is the low-side
and Ra the high-side resistance exclusive of Rj,,
the true ratio of input to output voltage is
R +
in
'pot
since no current is drawn through the output leads
at balance. It might be supposed that this ratio
could be determined conveniently from separate
measurements of Rg and Rb; but this is not desir-

15-7

0 ov^/w^AAAAAWAAAA^AA^v^^vvvywvy^A^AA GUARD

GUARD CIRCUIT

15 30 45
0

WORKING
I : vrr~ RESISTANCE
15 3.0 45
6

RATIO STANDARD
f

POTR
4
c

0
DIVIDER UNDER TEST

041
Figure 8.
Figure 7.

Circuit for self-calibration of voltageratio standard.

able. Not only would two separate resistance

measurements be required but the problems of selfheating and of leakage current would be Ignored.
The voltage ratio standard used at the Bureau
to calibrate high-grade commercial dividers in
corporates a number of special features to minimize
self-heating and leakage errors, to insure ratio
stability, and to provide capability of selfcalibration. As a result, this divider can be
operated at rated voltage (1500 V) on its top
range with a change in correction that amounts to
no more than 3 parts in 10^ from Its correction
determined at 207. of rated voltage. To reduce
leakage-current errors to negligible amounts, a
guard resistor (333 1/3 ohms /volt) operating in
parallel with the working resistor maintains,
through internal connections, proper potentials
on metal sleeves that surround the binding posts
bushings and on metal plates that support the
resistance colls. By this arrangement, voltage
across working Insulation sections is never more
than 75 volts and is zero across the bushings.
The guard resistance is also equipped with binding
posts corresponding to those on the working cir
cuit, permitting continuation of the guarding
feature to the shielding of external measuring
circuits .
The self-calibration technique for measuring
the errors of the Bureau's ratio standard to
better than 5 parts in 10^ Is made possible by
arranging the resistance sections in a particular
sequence. The first five sections have the same
nominal resistance (50 ohms) and the sixth section
has the same nominal resistance as their sum
(250 ohms). With these six sections, totaling
500 ohms, considered as a group (M]_) and also as
the first section of the next group (M) , the
sequence is repeated by making each of the next
four sections 500 ohms each, followed by a 2500ohm section. This process is repeated twice more
to form a total of four groups, with group
always considered as the first section of group
^n+1' Thus each group is characterized by five
nominally equal resistors, and a sixth nominally
equal to their sum.
In the calibration of this arrangement, a
"Direct Reading Ratio Set" with a range of 1/1
0.057. and a resolution of 1 part In 10^ forms

$
V
Test circuit for calibration of voltage
divider In terms of voltage- ratio
standard.

two arms of a bridge. One or more resistance

standards, always chosen to have the same nominal
resistance as the section (or summation of sec
tions) being measured, forms the third bridge
arm. Each successive divider section occupies the
fourth arm In turn. Corresponding guard sections
maintain proper potentials on the guard sleeves
and plates of the standard, and connection to the
bridge shielding prevents leakage currents from
entering or leaving the measuring network. Star
ting with the first section, differences between
each of the successive sections and the first
section of the group are measured, followed by a
difference measurement between the sixth section
and the sum of the first five in the group. The
correction to any ratio can then be computed by
summing the measured differences. The first
section of group >'.-. is taken as the reference, and
all measured differences are normalized to this
base. The bridge arrangement for measuring the
higher resistance sections is shown schematically
in Figure 7.
The method used in calibrating other fixedratio dividers by comparison with the Bureau's
standard is indicated in Figure 8. It also takes
advantage of the accuracy gain afforded by a
"difference" technique wherein the difference
between nearly equal quantities need be measured
with only moderate accuracy to realize high
accuracy in the measurement of the quantity itself,
when it is established in terms of a known reference
standard. The ratio standard and the divider under
test are connected in parallel for the same nominal
ratio and supplied as indicated from the same
regulated d-c source, at the required test voltage
(rated voltage or some known fraction of it) .
The junction points i andA.2 are located so that
Vl and V4 are nearly zero. The small potential
differences
to V4 that exist between correspon
ding points on the two dividers are measured in
turn by introducing an opposing voltage across
the gaps and adjusting for a detector null. A
low-range potentiometer supplies the opposing
voltage and permits measurements to 0.1 _V. These
measured voltages, together with the correction to
the standard permit computation of the ratio cor
rection. A timed sequence of measurements of
V-j permit establishing warm-up error in the divider
under test. Evidence indicates that measurement

16-8

Figure 9.

100-megohm helical resistance unit for

100 kV operation.

accuracy can be 1 part In 10^ or better.

Other voltage ratio networks and methods of
measurement are being examined by the Bureau in
an effort to improve both the standard and measure
ment accuracy. Of several experimental designs
studied, the most promising and versatile appears
to be one that will operate on a series-parallel
principle, lend itself to several methods of
calibration, Including the self-calibration
technique described earlier, and will be suitable
for low frequency as well as d-c operation. In
addition, a proposed two-stage transformer net
work looks promising as a corollary a-c ratio
standard. This could serve to further evaluate
the a-c performance of the above ac-dc model and
be useful in other experimental work involving the
measurement of voltage ratios.
For the measurement of very high direct
voltages, a new problem-- loss of current through
corona discharges at intermediate points of the
resistor s tring--becomes important, in addition to
the problem already discussed--self-heating and
leakage across the insulation of supporting
structures. Figure 9 shows the construction of a
resistor, appropriate for use to 100 kV, In which
these problems have all been adequately met. It
is made up of a large number of Individually
shielded 1-megohm wire-wound resistors connected
in series and arranged to form a vertical helix
(supported on a luclte cylinder) between a ground
plate and a high-voltage electrode. The indivi
dual shields completely enclose each 1-megohm
resistor and prevent corona formation at the re
sistor face, whatever the potential of the shield
above ground; the pitch of the helix Is such that
the gradient between turns is below what would
give rise to corona; and the configuration of the
helix together with the large high-voltage elec
trode on top serves to prevent field concentration
and consequent corona formation at the high-voltage
end of the resistor string. An experimental check

Figure 10. Maxwell-Wlen bridge for measurement of

self and mutual Inductance.
at high voltage for corona or leakage errors con
sisted of accurately comparing the current "in"
at the high-voltage end of the resistor string with
the current "out" at the ground end. Corona and
leakage errors were found to be less than 1 part
in 10* at 50 kV and less than 2 parts in 105 at
100 kV for humidities up to 50% RH. All the re
sistors had temperature coefficients less than
2 parts in 10^; some coefficients were positive and
some negative so that they could be selected in
matched parts for an overall coefficient less than
4 parts in 10? per C. The equilibrium temperature
rise of the unit is large at rated voltage, but if
it is excited less than 30 minutes, the maximum
error caused by heating was estimated to be less
than 4 parts in 10^.
Inductance - While in theory the NBS unit of in
ductance could be preserved by the calculable
inductors constructed for the Bureau's ohm deter
minations, it has been generally more expedient
In the practical work of Inductance measurement
to derive the henry from the ohm and farad through
the reference standards of resistance and capaci
tance. The calculable inductors are quite bulky
and, hence, have relatively large capacitance to
ground and considerable coupling to other circuits.
Also, they are wound with thick wire whose skin
effect would be appreciable even at moderately
high frequencies.
The Inductance Scale - The circuit regularly used
to measure inductance is that of the Maxwell-Wien
bridge shown in Figure 10. The adjustable capaci
tor is a 3-dial mica capacitor having minimum
steps of 0.001 ^F. A precision variable air capa
citor with a range of 1000 pF is used for inter
polating between the lowest steps of the mica
capacitor. The bridge is well suited for the
measurement of inductances in the range 1 uH to
10 H at frequencies in the range 100 Hz to 10 kHz.
The mica capacitor must be calibrated at each
frequency used--100, 400, 1000, and 10,000 Hz.

17-9

Over the middle range of Inductance values , and

at a frequency of 1 kHz, the total uncertainty In
the measurement Is no more than 0.017.. Wagner
arms (not shown In the diagram) must be used
generally to eliminate the effect of admittances
to ground from the bridge corners. The bridge
must also be carefully shielded and Its temperature
closely controlled If measurement uncertainties
are to be kept within the stated limit.
Mutual inductances up to 50 mH for a-c use are
regularly measured with the circuit shown, Includ
ing the dotted lines of Figure 10. With switch S
closed to the right, capacitor C is adjusted to a
balance value C^, which is a measure of Lx. Switch
S Is then closed to the left and C Is readjusted
to a new balance at C^. The mutual Inductance is
then computed from the relation M (C^ - C^)
RoifoR^/CiU + R4) , with further correction terms in
volving the residual Inductances of the resistance
arms. For inductors of larger value, or at higher
frequencies where the effects of distributed
capacitance are larger, this method becomes Inpractical. In such cases, the primary and secon
dary windings of the mutual inductor are connected
in series, first aiding and then opposing, and the
self- inductance of each combination is measured
with the Maxwell-Wien bridge. If the observed
values are La and L0 respectively, and If Lp and
L8 are the Inductances of the primary and secondary
windings measured separately,
La -L o La -L p -L s L p+L s -L o
M - "
2

2
'
Any inconsistencies among these equations give a
valuable indication of the order of magnitude of
the effects of stray capacitance between the colls.
At low values of inductance, errors from the
residual phase defects of the bridge arms may be
come appreciable. Hence, for the measurement of
the residual inductances of nominally "noninductive" 2-terminal resistors, it becomes es
sential to use a substitution method. In this,
a standard whose inductance can be computed and
whose resistance is nominally equal to that of
the impedance under test, is substituted for the
unknown in the X-arm of the bridge. Computable
standards are available for 10,000 and 1000 ohms.
Each consists of a fine wire stretched down the
axis of a brass tube about 5 cm in diameter. For
100- and 10-ohm standards, parallel wires are
used. The time constants (L/R) of these standards
can be computed to 1 x 10" second. For still
lower resistances, 4- terminal standards are used-elther parallel wire, reflex-strip, or coaxialtube types. Some of the coaxial- tube type have
their potential leads inside the inner tube, where
as others have them outside the outer tube. These
4-terrainal standards range in resistance from 0.1
to 0.0002 ohm and in current rating from 5 to
2500 amperes.
Magnetic Quantities - The unit of magnetic flux,
the weber is derived from the henry and the ampere ,
the latter being derived, in turn, from the volt
and ohm. A change in flux- linkages in a "searchcoil"--produced either by a change in the flux or
a motion of the coll is proportional to the time
integral of the voltage Induced in the coil and is
measured usually by the resulting ballistic de

flection of a galvanometer. The galvanometer may

be calibrated by connecting it to the secondary
winding of a mutual inductor and observing its
ballistic response when a known current is produced
in the primary winding of the Inductor. The change
in flux linkages, Na*, with N turns of the secon
dary winding, resulting from a change Al in the
primary current is NA* " MaI.
For convenience, the primary current is re
versed so that Al " 21. If I is in amperes and
the mutual inductance, M, is in henrys, Na* is in
webers. The ballistic response of the galvanometer
to changes In flux-linkages is strongly dependent
on the resistance of the circuit connected to its
terminals; and, normally, the resistance of this
circuit, including search coils, is adlusted to
make the galvanometer scale direct reading in
convenient decimal fractions of a weber. The
value in henries of the mutual inductor used in
the calibration is obtained by comparing it
directly with the computable mutual inductance
standard that was constructed initially for an
absolute ohm determination.
The weber, as thus established, is used to
measure the flux in test specimens in various
types of permeameter. The test specimen in the
permeameter is surrounded by a coil of a known
number of turns connected to the calibrated gal
vanometer. A measured change in magnetizing
current in the permeameter produces a change in
the flux linking the coil, and the resulting
galvanometer deflection measures the flux change.
A measurement of the cross-section, A, of the test
specimen then permits computation of the change in
magnetic induction 3 A*/A. If A Is in square
meters,
is in webers/meter^ or teslas. The
area of the specimen and the number of turns in
the coil may be lumped with other factors in the
galvanometer calibration, so that the induction
is read directly from the deflection.
Magnetizing force may be computed directly
from the current turns per unit length for a uni
form ring specimen or for a magnetic circuit in
which the magnetomotive force is distributed around the circuit in proportion to the reluctance
of the various portions so that all portions carry
the same flux. Under these conditions, the tangen
tial component of the magnetizing force is the same
inside the metal as outside at the surface of the
specimen, and the computed value applies to the
interior of the specimen if its magnetic properties
are uniform. A second method of measuring magneti
zing force is to measure the tangential component
of flux density in the air closely adjacent to the
test specimen. This measured flux density in
teslas divided by the magnetic constant fm (the
assigned permeability of free space, 4n x 10"?)
yields the value of magnetizing force in amperes/
meter.
The principal object of testing ferromagnetic
materials by d-c (ballistic) methods is to obtain
data from which normal induction curves and hystere
sis loops can be plotted. This is done either by
using suitably shaped specimens, such as rings,
which constitute the entire magnetic circuit, or
by means of permeameters . When permeameters are
used, the test specimens are bars submitted for
test by the laboratories of steel companies and
other organizations. The circulation of calibra
ted bars insures uniformity and accuracy through-

18- 10

out the Industry.

A related service Is the calibration of
search colls to determine their effective areaturns. This is done by placing the search coll
near the center of a long uniformly wound, singlelayer solenoid so that the axes of the search coll
and solenoid are aligned. Their mutual Inductance
la measured balllstlcally by comparison with a
standard; and the area turns, AN, of the search
coll are computed by the equation AN - MI/B. The
ratio of the solenoid current I to the central
flux density B is a constant that Is computed from
the measured dimensions and winding pitch of the
solenoid.
Another calibration service of Importance Is
the measurement of flux density In the air gaps of
permanent magnets that are to be used as standards.
The calibration procedure is based on the accurate
measurement of flux density in the gap of an elec
tromagnet, using the nuclear magnetic resonance
technique. This accurately known flux density Is
compared with the flux density In the gap of the
permanent magnet standard through the use of a
suitable transfer Instrument, such as a Hall ef
fect magnetometer or a rotatlng-coll magnetometer.
By using the established value of proton gyromagnetic ratio (v/2tt - 4.25759 x 107 hertz/tesla)
and a suitable resonance probe, the induction In
an unknown magnetic field can be measured by a
simple determination of the resonance frequency.
Such a resonance detector is also used to study
flux distribution in a magnetic field and, with
suitable circuitry, as a monitor to hold the in
duction constant at a given place in the field.
Direct Current. Power, and Energy Measurements The block diagram of Figure 11 shows the deriva
tion of the units of current, power, and energy
from those of resistance, voltage, and time. It
also shows the ac-dc transfer instruments on which
all a-c measurements are based.
An electric current (if one excepts the per
sistent current in a superconducting circuit and
the or or P emission from a radioactive isotope)
la in the nature of things a transitory affair,
and a "standard ampere" cannot be preserved in
the laboratory as can a standard ohm. For the
accurate measurement of direct current, the normal
procedure Is to insert a known resistance In the
circuit and to compare the voltage drop produced
In this resistance by the unknown current, with
the emf of a standard cell. This comparison is
usually made with a potentiometer. With this
technique, measurements to perhaps 10 parts In
10^ are now possible. In rare cases where extreme
accuracy is needed, special resistors are used of
such value that the IR drop at the desired current
is very closely equal to the emf of one or more
standard cells connected in series. Thus, the
circuit becomes very simple and errors from con
tacts, thermal emfs , and drift in the potentio
meter circuit are minimized or eliminated.
The extension of measurement to several hun
dred amperes depends on 4- terminal resistors
properly designed to carry the large currents
without excessive heating and without abnormali
ties of current distribution. A Kelvin double
bridge used with a substitution technique permits

STANDARD
RESISTORS

CELLS

FOR CURRENT
ME(UMTS

roTomoMETERS MB
VOLT BOXES

0-C
CURRENT
MEASMTS

0-C
MURK
MEASMTS

S-C
POWER
MEASMTS

D-C
ENERST
MEASMTS

THERMAL
CURRENT

THERMAL
VOLTACE

ELECTRODYNAMIC J

LC
STANDARD
WATTHOUR
METER

A-C
CURRENT
MEASMTS

A-C
VOLTAGE
MEASMTS

A-C
A-C
POWER
ENERST
MEASMTS
MEASMTS
Figure 11. Measurement of current, voltage, power,
and energy.
measurements to 0.01% accuracy up to a maximum
current of 1000 A. This 6-arm bridge circuit Is
of the conventional type In which auxiliary
balances minimize the effects of contact and lead
resistances which could otherwise mask the value
of the resistor. The substitution technique in
volves two bridge balances, the first being made
with a calibrated standard in the "unknown" arm
of the bridge and the second with the resistor
under test substituted for It. In this way, only
the measurement of small differences between
nearly equal quantities is required. Oil-cooled
standards of 0.0001 and 0.00001 ohm nominal re
sistance are used at the 1000 ampere level.
Recent current comparator developments at the
Canadian National Laboratory constitute a break
through in current ratio measurements, and a d-c
comparator permits the measurement of current ratios
over a wide range to an accuracy approaching a part
in 10^. Extending the direct current scale to
20 kA should soon be possible, and the application
of the direct-current comparator to comparisons of
4-termlnal resistances is expected to increase the
accuracy of that measurement by an order of magni
tude or more.
The measurement of power in a d-c circuit as,
for instance, in calibrating a calorimeter can be
made most accurately by using a potentiometer with
an accessory shunt and a voltage divider ("volt
box") to measure the current and the voltage of the
circuit separately, and by then taking their pro
duct. Electric energy is, of course, the integral
of power with respect to time, and If the power
is constant, can be measured by using the potentio
meter, voltage divider, and shunt to obtain a known
and constant power for a measured time. Accuracies
of perhaps 20 ppm in power measurements and 50 ppm
in energy measurements are possible. Of course, If
the applied power varies with time, some sort of
integrating meter must be used. The d-c watthour
meter of either the commutator or mercury-motor
type is a relatively Inaccurate device with errors
of 0.1 percent or more. More accurate measurements
can be made with digital watthour meters incorpora
ting feed-back wattmeters as explained below.

19- 11

Transfer from D-C to A-C Measurements - The stand

ard cell and the units of voltage and of current
realized from It directly or In conjunction with
a standard resistor are applicable to the measure
ment of d-c quantities only. On the other hand,
by far the more frequent and important measurements
in the power and communication fields involve
alternating currents. The procedure used to
transfer from the d-c standards to a-c measuring
apparatus, as indicated in Figure 11, therefore,
constitutes an essential link in the chain of
electrical measurements.
There are at least eight quantities that may
be determined for an unmodulated periodic a-c wave.
These are: 1) the rms or effective value; 2) and
3) positive and negative average values, i.e., the
average of all of the positive or all of the nega
tive values during one cycle; 4) the rectified
or full-wave average; 5) and 6) the positive and
negative crest (peak) values; 7) the crest-tocrest (peak- to- peak) value; and 8) the fundamental
component only. Instruments for measuring each of
these values are commercially available (with
widely varying accuracies). Ordinarily, however,
only one of these values Is really required in a
measurement. Unfortunately, it is impossible to
deduce one value accurately from the measurement
of another except for certain known wave forms.
In most cases, the rms value is the one desired,
since the rate of transformation of electrical to
other forms of energy (which is governed by the
heating produced by a current and by the electro
magnetic force produced by currents acting on
each other) is a function of the square of this
value. For these reasons, the transfer from d-c
to a-c measurements at the Bureau has traditionally
been made in terms of the rms value. However,
peak ac-to-dc measurements can also be made over
a restricted range of voltages and frequencies.
There are three general types of ac-dc
transfer instruments which are at present suitable
for rms measurements of high accuracy. They are
dependent upon these same laws of the interchange
of energy and are: 1) electrodynamic instruments,
which depend upon the force between currentcarrying conductors; 2) electrostatic instruments,
which depend upon the force between charged con
ductors; and 3) electrothermic instruments, which
depend upon some effect produced by the heating of
a current-carrying conductor. The first and third
respond essentially to current and the second to
voltage, but series or shunt resistors make all
three types suitable for both measurements, while
other circuit arrangements also make power measure
ments feasible.
The differential equation governing the angular
deflection of an electrodynamic instrument in which
two sets of coils carrying the same current, one
fixed and one rotatable about an axis, Is:
Pd" + Ao' + U6 - Bi2, where P, A, and U are the
lnertial, damping, and spring constants, respec
tively, and B is a parameter which depends upon
the rate of change of the mutual Inductance be
tween the coils with respect to the angular de
flection. If the mechanical inertia of the
movable system is sufficiently great with alter
nating current applied, the periodic fluctuations
of d are negligible compared with the average
(steady-state) value. Then by integrating each
term over one cycle and dividing by T, we have

simply
ue - |j i2dt - bi2.
*o
Thus, if U and B each have the same dc-to-ac value
on direct and alternating currents, the instrument
may be used for the dc-to-ac transfer.
Similar equations can be developed for the
electrostatic instrument, with V replacing I, or
for an electrothermic instrument with the tem
perature difference replacing the angular deflec
tion 9. (The parameter B is, of course, also
different.) In each Instrument, the basic
equation holds even if the parameters U and B are
dependent upon the response 9, This important
advantage arises because each instrument combines
In a single measuring element a function pro
portional to the square of the instantaneous
current or voltage, a restraining function, and
an lnertial function which enables it to integrate
so that the time-average value of the response is
proportional to the square of the rms current or
voltage. For this very fundamental reason, high
accuracy may be more easily attained with such
instruments than with other squaring devices or
circuits which must synthesize exact and equal
square- law responses in two quadrants (or in one
quadrant with an accurate rectifier) and provide
a separate accurate integrator.
At the Bureau, thermal converters are used
as rms ac-to-dc comparators for current and
voltage. Each of these electrothermic ac-dc
transfer standards contains a thermoelement con
sisting of a heater and thermocouple. In Its
usual form, the heater Is a short, straight wire
suspended by the two supporting lead-in wires in
an evacuated glass bulb. Its temperature rise Is
dependent on the current through it. The hot
junction of a thermocouple is fastened to the mid
point of the heater and is electrically Insulated
from it with a small bead. Thus, the thermocouple
emf (about 5 to 10 mV for a conventional thermo
element at rated current) is a measure of the
heater current.
Sets of thermal voltage converters are used
for voltage measurements. Each consists of one
or more cylindrical metal- film resistors in a
coaxial metal cylinder, connected in series with
one of two 5 mA thermoelements mounted in a sepa
rate cylinder, so that the output emf is then a
measure of the input voltage. A thermal current
converter consists of a coaxial a-c shunt with a
low-range thermoelement connected to Its potential
terminals. Alternatively, thermoelements of higher
ranges (up to 20 A) can be used directly without
shunts .
The ac-dc difference of a transfer standard
may be defined as 6 (Xa - X<j)/Xd, where Xa and
Xj are the magnitudes of a-c and d-c quantities
that are required to give the same response (such
as deflection or output emf) of the standard.
(Normally, the average for the two directions of
the d-c quantity given the best measure of the d-c
reference.) This is a useful definition, for the
a-c quantity corresponding to a given response is
then simply Xfl - Xd (1 + 6). The instrument may
be used as an ac-dc transfer standard for a-c
measurements simply by observing the response with
the a-c quantity applied and then measuring (with

20-12

Figure 12.

Circuit of determining differences In

ac-dc performance of thermoelements.

a potentiometer, d-c shunt, and voltage divider as

needed) the average for the two d-c quantities re
quired to obtain the same response, thus avoiding
many of the limitations of an ordinary instrument.
Pairs of thermoelements (or converters) are
evaluated with the emf comparator shown in Figure
12 to determine the differences In their ac-dc
performance, i.e., 6^ - 62 as defined above. With
the desired voltage applied, R^ and R2 are adjus
ted In a preliminary balance to obtain a near null
on detector Dj. Then, in regular succession, a-c,
d-c, reversed d-c, and a-c currents are applied to
the thermoelements at nearly equal time intervals
without changing
or R2. In each case, the
current is adjusted to obtain the same output emf,
E^, of TE^ as indicated by an auxiliary Lindeck
potentiometer (P with Its detector D2), and the
resulting deflections of
are observed. The rel
ative ac-dc difference is simply 6^ - 62
S (Da - Dd) , where Da and Dd are the averages of
Di with the a-c and d-c currents applied, respec
tively. The sensitivity factor (S) Is easily
determined. With this circuit, small fluctuations
In the supply current or voltage do not appreciably
change the balance point, and the sequence of
readings greatly reduces the effects of drift in
either thermoelement. In this way, a precision
approaching a few parts in 106 Is attainable.
Fourteen thermoelements, each of which the
oretically should have negligible (less than 2
parts in 106) ac-dc error at audio frequencies
form the Bureau's present ac-dc reference group.
They are of different manufacture, range, and
type. Two of them are of a new design by F. J.
Wilkins of the National Physical Laboratory in
England, having many Junctions fastened along the
length of a bifllar heater. The ac-dc differences
of this type should be very different than those
of the more conventional kind. The ac-dc differ
ences of all of them agree to a few parts in 10^
providing the necessary concordance of theory and
experiment. The extension to higher voltages and
currents Is made by step-wise intercomparisons of
adjacent ranges of the thermal voltage converters
arid the thermal current converters (or high range
thermoelements). In this way, the ac-dc differ
ences of any range can be determined in terms of
any one.
The same circuit and the same techniques are
used with the same standards to determine the acdc differences of various kinds of commercially

available thermal ac-dc transfer standards that

are sent to the Bureau for calibration. Measure
ments are normally made to 1 part in 10^, with 5
parts in 10^ projected at audio frequencies for
the near future.
An electrodynamic wattmeter is used as the
basic ac-dc transfer standard for power and
energy measurements at NBS. Its movable coils
carry a current proportional to and in phase with
the applied voltage, and Its fixed coils carry
the load current. The torque is then a function
of the applied power.
The Bueau's standard ac-dc transfer wattmeter
was carefully designed to be as free as practicable
from sources of a-c error such as eddy currents,
capacitance, and inductance. In addition, a great
many checks were made on its performance by com
paring It with a hot-wire Instrument and a quadrant
electrometer, and using It to measure the loss In
capacitors which previously had been measured In
an a-c bridge. Based on these tests It Is believed
that the errors are known to at least 0.005 percent
of the applied volt-amperes at any power factor at
power frequencies, and that the Instrument can
still be used with reasonable accuracies at fre
quencies up to perhaps 1 or 2 kHz.
This wattmeter Is used principally to deter
mine the ac-dc differences of electrodynamic
wattmeters of high accuracy, which, In turn, are
used as the reference Instruments in other labora
tories for calibrating watthour meters with alter
nating current. The sequence of ac-dc tests is
similar to that already described with the thermal
converters, and the ac-dc difference of the instru
ment to be calibrated is determined in terms of
the deflections of the standard wattmeter and a
sensitivity factor, with a correction for the
ac-dc difference of the standard. "Phanton load
ing" with separate sources to energize the voltage
and current circuits of the standard wattmeter and
the instrument under test, make it easy to obtain
any desired fictitious power and to Insure that
both wattmeters "see" the same value of power.
The standard wattmeter is used with counting
relays which are closed for periods of an Integral
number of seconds to provide a measured power for
a known time Interval, in order to calibrate a
bank of five high-grade standard watthour meters.
These meters are maintained under carefully con
trolled temperature conditions and are used only
at rated voltage and current on their 5 ampere,
120 volt ranges to serve as a standard of a-c
energy. Periodic calibrations with the wattmeter
several times a year have shown a repeatability
better than 0.03 percent for the average of the
registration of the five watthour meters. One of
the five serves as a working standard between
calibrations to determine the registration of
standard watthour meters that are sent for calibra
tion. In this way, the watthour is literally
carried from Washington to the power companies'
laboratories with an accuracy approaching 0.0S
percent, if the average of at least three meters
is used.
A number of wattmeters have been developed
with a second instrument (usually a d-c mllllammeter) connected to the same shaft as the
electrodynamic wattmeter element and arranged so
that the adjustable current through it produces

21-13

a torque opposing that of the wattmeter. This

current Is, thus, directly proportional to the
measured power when the pointer Is at zero. As
an extension of this principle, the out-of-balance
position can be detected electronically or other
wise to generate this counter-current automatically
In a feed-back circuit. The Integral of this
current with respect to time is thus proportional
to energy. This energy can be measured by passing
the current through a fixed resistor and Inte
grating the voltage drop with a vol tage- to- frequency
converter and a counter. Such "digital watthour
meters," which may be used over a wide range of
frequencies including direct current, are available
commercially and are under study at the Bureau In
the hope of providing a better reference standard
for the calibration of standard watthour meters.
The Alternating Voltage Scale - The set of thermal
voltage converters offers a convenient means for
measuring alternating voltages from about 0.2 V to
500 V with high accuracy over a wide frequency
range. Lower voltages may be generated by measur
ing the voltage applied to the input of an induc
tive voltage divider, using the thermal voltage
converter and appropriate d-c equipment (potentio
meter and resistive voltage divider). The output
of the voltage divider can be applied directly to
digital voltmeters, for example, to calibrate
their low ranges.
Inductive voltage dividers, consisting of a
group of cascaded multi-tap transformers with
special windings on high-permeability toroidal
cores, are available with six or seven decades of
voltage division. Some are capable of accurately
dividing their input voltage to about a part in
10^ at their design frequency. The accuracy of
voltage division may be determined with reference
to a divider made up of known 3- terminal capaci
tors, or may be measured against a master induc
tive voltage divider that has been so calibrated.
The Bureau now has the capability of making such
calibrations over the frequency range from 50 Hz
to 10 kHz.
For the measurement of higher alternating
voltage, at 60 Hz, Instrument voltage transformers
are used universally by the power industry. The
ratio corrections and phase angles of voltage
transformers that will be used as reference
standards In other laboratories are determined at
the Bureau either directly by means of capacltive
S

Figure 13.

voltage dividers or Indirectly by comparison with

standard transformers whose corrections have been
determined by this means. In the latter case,
the primary windings of the standard and test
transformer are connected in parallel to a suitable
high-voltage source, and the difference between
their secondary voltages is measured with a com
parator. The comparator now used at the Bureau
shown schematically in Figure 13, employs an In
ductive voltage divider for comparing the magni
tudes of the two secondary voltages and a resis
tance-capacitance network to evaluate their phase
difference. Uncertainties In the inductive com
parator have been estimated at 2 parts in 10^ In
ratio correction factor and 10 u radians. The
Bureau's group of standard transformers covers
the voltage range to 250 kV, and the capacltlvedivlder technique has been used for calibrations
to 350 kV. The direct determination of the ratio
corrections and phase angles of voltage transfor
mers, using a capacltive voltage divider, depends
essentially on the stability of the 3-terminal
capacitors and the accuracy with which their
values can be measured. In a recent international
calibration comparison that covered the range
50-350 kV, agreement was generally within 2 parts
in 105.
There is no significant degradation of ac
curacy for either method at higher frequencies,
at least up to 400 Hz. Calibration tests of vol
tage transformers have been made at this frequency
over a voltage range up to 9 kV.
The Alternating Current Scale - The transfer
principle and its embodiment in a set of thermal
current converters provide a convenient and
accurate means for measuring alternating currents
from 1 mA to 20 A at frequencies to at least 20
kHz while preserving a direct tie with their d-c
equivalents. Extension of the current scale to
several thousand amperes with high accuracy is
accomplished by use of Instrument current trans
formers and, more recently, current comparators.
In Its simple form, a current transformer
comprises two windings (primary and secondary) ,
electrically insulated but coupled magnetically
through a high-permeability core. Operating under
nearly short-circuit conditions, it reproduces in
the secondary circuit a reduced replica of the
primary current. For a perfect transformer, the
relation between these currents would be

Circuit of vol tage- trans former com

parator.

where Ns and Np are the number of turns In the

respective windings, zero phase displacement be
tween the currents being inferred. But transfor
mer operation requires that a finite flux be
established In the core, and power for the secon
dary circuit must be supplied through the windings
and core from the primary circuit, so that both
ratio and phase angle will depart from their ideal
values. Furthermore, these departures from the
ideal relations are functions of frequency, cur
rent, and secondary burden; and, for multi-ratio
units, the errors may differ for the several
ratios. Thus, a current transformer must be cali
brated under the conditions in which it will be
used.
22-14

Figure 14.

M
Circuit for direct calibration of
current transformers.

Figure 15.

The method used for many years at the Bureau

for calibrating current transformers up to
2500/5 A is shown schematically in Figure 14.
Resistor R^, in series with the transformer pri
mary, is a 4-terminal astatic network which,
based on theoretical considerations, has negli
gible skin effect and a very small known phase
angle; resistor R2, in series with the transformer
secondary, has the same features and, in addition,
permits step changes of 0.01 ohm in its 4-terminal
resistance over a range from 0.01 to 0.1 ohm, with
a continuously adjustable section having a span
of 6%, and a resolution of 10 parts in 10&; the
mutual inductance M is characterized by its broad
range with continuous adjustment and by its astatic
configuration. With the ratio R2/R1 chosen to
equal the nominal current ratio, balance is ob
tained by adjusting R2 and M for a detector null.
With zero voltage around the detector loop, the
transformer errors can be computed from the known
circuit parameters.
For primary currents greater than 2500 A and
up to 12000 A, the comparison method of Figure 15
Is used. A standard transformer and secondary
resistance R3 replace R^ of the previous method.
Comparisons are made from the secondary circuits,
but the balancing procedure is unchanged. The
error of the standard transformer must be taken
into account in the balance equation and must be
known. Its measurement rests on the principle
that the error of a current transformer, properly
designed, remains unaltered if the change- of- range
is effected by changing the numbers but not the
symmetrical distribution of the primary turns,
while the secondary turns remain the same. This
principle is realized in the Bureau's standard
transformer by dividing the primary winding into
n equal sections and arranging them so that k
sections can be connected in series to form n/k
groups connected in parallel. Deviations from
the principle arise only from the product of two
factors: 1) relative differences in current dis
tribution among the primary groups in parallel;
and 2) relative difference in the coupling
(mutual Inductance) between the primary sections
and the secondary winding. The primary winding
has 24 sections of one turn each, whose symmetri
cally disposed series and parallel combinations
provide 8 ratios for a given secondary; added
ratios are available by changing the number of
secondary turns through a series-parallel arrange
ment. When the standard is used in calibrating
315-411 0-68-3

Circuit for calibration of current

transformers in terms of a standard
transformer.

high- ratio transformers, a 1 or 2 turn arrangement

of the primary turns is used. Its errors are
measured periodically by reconnecting it for a
lower ratio and using the resistance method des
cribed above. Commercial transformers of the
highest quality can be calibrated by either
method to an accuracy of 200 ppm and when time and
effort are Justified (e.g., an international com
parison) an accuracy of 100 ppm or better is
possible.
The demand for Increased accuracy in measure
ments at power frequencies has been felt in
current-ratio measurements; and needs have devel
oped for current- ratio measurements over a fre
quency range extending to 10 kHz or more. These
requirements have led to a reconsideration of the
entire problem of current-ratio determination.
In recent years, a ratio standard, operating on a
principle quite different from the current trans
former and with potential accuracy better than 1
part In 106, has been developed by Kusters and his
co-workers in the National Research Council of
Canada (NRC) In cooperation with Professor Miljanic
of the University of Belgrade and the Nicola Tesla
Institute. The essential components of the latest
verslon--the compensated current comparator--are
Indicated in Figure 16, and its application as a
ratio standard for calibrating current transformers
is shown in Figure 17.
Ampere's circuital law states that the line
integral of magnetizing force around a closed path
equals the sum of the enclosed currents, i.e.
Hdl - i. I. In a current transformer, this in
tegral has a value not zero, and this accounts
for its magnetic or low- frequency error; In the
current comparator, however, this Integral is
zero at balance, and the ratio of currents is the
Inverse of the turn ratio. Referring to Figure 16,
a high-permeability toroidal core carries a singlelayer detection winding (d) of many turns which
senses the condition of zero flux In the core as
indicated by the detector; a compensation winding
(c) having the same number of turns as the secon
dary (S) encloses both the core and the magnetic
shield (Sh2) and provides a path for the magne
tizing current of what may be considered a "trans
former," comprised of the primary (P) and secondary
(S) windings and the shield (Shj).
A necessary and sufficient condition for pro
per functioning of the comparator is that the core

23-15

PRIMARY WINDING (P)

SECONDARY WINDING (S)
MAGNETIC SHIELD (Sh|)
COMPENSATION

WINDING (C)

MAGNETIC SHIELD (Sh2)

DETECTION WINDING
MAGNETIC
--0Figure 16.

(D)

CORE

Components of a compensated current

comparator.

flux, as sampled by the detection winding, must

truly reflect Hdl when It encompasses only those
currents pertinent to the measurement. Thus, ex
traneous magnetic fields, such as those associated
with leakage Inductance or arising from sources
external to the comparator, must not be permitted
access to the core. More Insidious, but no less
Important, Is the effect of non-uniform permea
bility around the magnetic path of the core. This
results In an abnormal distribution of the flux
produced by the primary and secondary currents
when these windings are separated in space. The
result is that the detection winding may sense
flux even though the line Integral of magnetizing
force is zero; the converse is equally possible.
The magnetic shield (Sh^), of material whose sat
uration level is high, prevents extraneous flux
from reaching the core, and essentially nullifies
the effect of the physical separation of primary
and secondary turns; the inner, high-permeability
shield (Sl^) , performs a similar function with
respect to the compensation winding. The shield
(Sh^) also enables the comparator to support Its
burden without imposing any of it on the current
transformer under test.
The operation of a compensated current com
parator used as a ratio standard to calibrate a
current transformer may be explained with the aid
of Figure 17 and an initial oversimplification.
At first, we shall assume that the current trans
former (CT) is free from error (i.e., its primary
and secondary ampere turns are exactly equal) and
that the adjustable balancing elements (G-C) are
disconnected from junction M. The primary and
secondary windings of the current comparator (CC)
together with its outer shield (Shj of Figure 16)
form, in effect, a current transformer whose ampere
turns are not equal. The compensating winding has
the same number of turns as the secondary winding.
Now from Figure 17, the sum of currents through
compensation and secondary windings must equal the
secondary current of the assumed perfect current
transformer. Thus, the total ampere turns In the
comparator must be zero, and detector D will In
dicate a null. The current in the compensation
winding is thus associated with the ampere turn
difference between the primary and secondary of
the comparator as a result of its current trans
former action with shield Shj. Injection of a
small voltage at g to compensate the impedance drop

Figure 17.

Comparls'on circuit used in calibrating

a current transformer.

in the compensation winding of the comparator can

be used to bring the point M to the same potential
as N. For this condition, no burden is imposed on
the current transformer (CT) by the comparator.
If the resistance of the compensation winding is
very low, M is sufficiently close to ground poten
tial without any voltage being injected at g.
Now the current transformer (CT) is not error
free, but it is made to appear so to the current
comparator (CC) by injecting appropriate in-phase
and quadrature currents through the G-C balancing
network. At balance, Indicated by null on the
detector, the error of the current transformer is
given by the equation r (G + jiuC) , where the
real component is the error In ratio and the
imaginary is the quadrature error. To this must
be added any error of the current comparator.
With minor modifications in the G-C loop, the re
sistance r in the transformer secondary circuit
can be eliminated, making it possible to test the
transformer with nearly zero burden.
A compensated comparator has been built at
the Bureau, providing the most common ratios en
countered from 5/5 to 600/5. Designed primarily
for 60 Hz operation, its errors remain less than
a part in 106 at 1 kHz. This unit will be used
In the calibration circuit just described and in
development work directed toward extension of the
comparator principle to higher ratios. The currentcomparator systems are Intended to replace the
existing transformer calibration methods and to
Improve calibration accuracy by an order of magni
tude. A current-comparator-current-transformer
test set, comprising the essential features of the
test circuit described above, is now commercially
available for transformer testing up to a 1200/5
ratio.
The above discussion has been concerned with
low-frequency operation, since additional param
eters become increasingly important when the fre
quency increases to 10-20 kHz. The one most pro
found effect is that of the distributed capaci
tance which shunts current around portions of a
given winding or from one winding to another.
Although the distribution of capacitance currents
depends on design, their effects are present in
both types of ratio standards, and the marked lowfrequency advantage of the current comparator over
the current transformer diminishes with increasing

24-16

Figure 18.

High Voltage Laboratory at National

Bureau of Standards.
The 2 MV, 33 kJ surge-voltage genera
tor Is at the extreme right. In the
center is the 1 MV 60 Hz cascaded
transformer set.

Figure 19.

frequency. An Important new concept regarding the

definition of ratio must be Introduced at the
higher frequencies. A primary or secondary current,
as such, can no longer be assigned to the respec
tive windings, since the current entering a given
winding may not be the same as that which leaves
it. Thus, the ratio of currents being measured
must be identified with a specific pair of ter
minals and with a given voltage distribution
along the windings. With proper attention given
to capacitance distributions, both the comparator
and the transformer can be designed for broad
frequency range of operation with small errors.
At the Bureau, two nearly Identical standard
current transformers with consecutive ratios 1/1
to 6/1 have been designed and evaluated over the
frequency range 400 Hz to 16 kHz. Their changes
In complex error with range, frequency, and secon
dary current are relatively small. The ratio error
for any combination of these variables is no more
than 25 ppm and the quadrature error no more than
50 ppm (u radians). If only the frequency decade
1-10 kHz Is considered, the ratio error is never
more than 7 ppm and the quadrature error 22 u ra
dians. NRC has pursued a corresponding develop
ment program for current comparators.
Measurement of Crest and Surge Voltages - In the
range of very high alternating voltages, engineers
are frequently concerned with crest rather than nns
values. This is because the breakdown of Insula
tion Is to a large extent dependent on the crest
value of the voltage to which it is subjected. A
technique which has been much used In this field
is to note the sparkover voltage between metal
spheres of known diameter and spacing in air under
standard test conditions. The value of crest
voltage is then determined within 2-57. by reference
to standard sphere-gap tables. The Bureau has
1-meter and 2-meter sphere pairs which can be used
for this purpose.

200 kA, 50 kJ surge-current generator.

A closely related field is the measurement of

surge voltages and currents. Transient voltage
surges are much used in the proof- testing of
electrical apparatus, since the duration and dis
tribution of electric stress can, In this way, be
made to approximate the extreme conditions to
which the equipment will be exposed In service as
a result of lightning and switching transients.
The Bureau has a 2 MV, 33 kJ surge-voltage genera
tor, shown In Figure 18, and a 200 kA, 50 kJ surgecurrent generator, Figure 19, which are used for
studying methods of surge measurement. A 3 MV,
250 kJ surge-voltage generator Is being planned
for the extension of this work.
The complications that arise from Inductance
and skin effect In measuring transient currents
are closely related to those encountered In the
precise measurement of phase angle of current
transformers. By applying similar procedures,
standard forms of shunts and inductors have been
developed for measuring surge currents and the
rates of rise of such currents.
The design of voltage dividers that will be
accurate for rapidly changing voltages, especially
during the first microsecond of the surge, Is a
difficult problem. Thin ribbon resistors have
been constructed at the Bureau, suitable for highvoltage surges and having a very low time constant,
2 x 10"9 second. They have been used In making
voltage dividers for measuring linearly rising
chopped Impulses with peak voltages to 300 kV and
time to sparkover from 30 nanoseconds to 50 micro
seconds. Divider ratio errors due to residual
inductance are computed to be less than 17.. Stray
capacitance errors are kept low by making the total
divider resistance 1000 ohms or less. By a com
bination of computation and experiments, capaci
tance errors have been deduced to be less than 17.
for time to sparkover down to 0.1 microsecond.
Using this divider, data have been derived giving
relations between rate of rise and flashover vol-

25-17

steeply rising voltage surges. This suggestion

has received international consideration and is
being studied further in a number of high-voltage
laboratories .

Figure 20. Volt-Time curve for 250 mm spheres at

60 mm spacing.
tages for various gaps. The relation is shown In
Figure 20 for 250 mm spheres spaced 60 mm apart.
It has been suggested that this curve be used as
a reference standard for lnterlaboratory compari
sons of measurement techniques in the field of

Dissemination of the Electrical Units - In addi

tion to the Bureau's statutory responsibilities
to develop, maintain, and have custody of the
national standards of measurement, it has the
further responsibility of making these standards
available through appropriate calibration services.
Thus, a great number of laboratories operated by
manufacturers, universities, public utilities,
Federal agencies, or private individualssend
their reference standards to the Bureau for com
parison with the national reference standards and
for assignment in terms of the latter. The Bur
eau's Electricity Division receives for calibra
tion standards of resistance, Inductance, and
capacitance, standard cells, ac-dc transfer
standards, ratio standards, standard watthour
meters, and other Instruments which are intended
to serve as reference standards in the laboratory
that sends them. Some must be recalibrated at
Intervals which are determined by the stability
and the type of use proposed for the standard.
The conditions of test and the fees charged for
calibration services are listed in Miscellaneous
Publications 250--"Calibration and Test Services
of the National Bureau of Standards, "--together
with listings of items which are appropriate for
calibration and other relevant Information.

Bibliography
J. L. Thomas et al, An Absolute Measurement of
Resistance by the Wenner Method, NBS J. Res.
43, p. 291 (1949).

General
31 Stat., ch. 872, p. 1449 (1901), Establishment
of the National Bureau of Standards.

R. D. Cutkosky, Evaluation of the NBS Unit of Re

sistance based on a Computable Capacitor, NBS
J. Res. 65A, p. 147-158 (1961).

Public Law 617, 81st Congress (ch. 484--2nd

Session). An act to redefine the units and
establish the standards of electrical and
photometric measurements.

Capacitance

F. B. Silsbee, Establishment and maintenance of

the electrical units, NBS Circular 475 (1949).

W. D. Volker, An Improved Capacitance Bridge for

Precision Measurements, Bell Lab Record 20,
133, 1942.

NBS Handbook 77, Vol. 1, Precision Measurement and

Calibration; selected papers on electricity
and electronics--a compilation of previously
published :echnical papers by NBS staff
members (1961) .

A. M. Thompson, D. G. Lampard, A New Theorem in

Electrostatics and Its Application to Cal
culable Standards of Capacitance, Nature 177 .
p. 888 (1956).

NBS Technical Note 262, Accuracy in Measurements

and Calibrations (1965).

A. M. Thompson, The Cylindrical Cross Capacitor

as a Calculable Standard, Proc. IEE 106B,
May (1959).

NBS Miscellaneous Publication 250, Calibration and

Test Services.

A. M. Thompson, The Precise Measurement of Small

Capacitances, Trans. IRE Instrumentation
1-7, December (1958).

Absolute Measurements
M. C. McGregor, et al, New Apparatus at NBS for
Absolute Capacitance Measurement, Trans. IRE
Instrumentation 1-7, December (1958).

R. L. Driscoll, Measurement of Current with a

Pellat-Type Electrodynamometer, NBS J. Res.
60, p. 287-296 (1958).
R. L. Driscoll, and R. D. Cutkosky, Measurement of
Current with the NBS Current Balance, NBS J.
of Res. 60, p. 297-305 (1958).

R. D. Cutkosky, Active and Passive Direct-Reading

Ratio Sets for the Comparison of AudioFrequency Admittances, NBS J. Res. 68C,
October (1964).

26-18

R. D. Cutkosky, L. H. Lee, Improved 10 pF Fused

Silica Dielectric Capacitor, NBS J. Rea. 69C,
July (1965).

R. L. Sanford, P. H. Winter, A Permeameter for

Magnetic Testing at Magnetizing Forces up to
300 oersteds, NBS J. Res. 45, p. 17 (1950).

J. Q. Shields - Voltage Dependence of Precision

R. L. Sanford, E. G. Bennett, Determination of
Air Capacitors, NBS J. Res. 69C, October (1965).
Magnetic Hysteresis with the Fahy Simplex
Permeameter, NBS J. Res. 15, p. 517 (1935).
Resistance
F. Wenner, Methods, Apparatus, and Procedures for
the Comparison of Precision Standard Resis
tors, NBS J. Rea. 25, August (1940).
J. L. Thomas, Stability of Double-Walled Manganln
Resistors, NBS J. Res. J}6, p. 107 (1946).
J. L. Thomas, Precision Resistors and Their
Measurement, NBS Circular 470 (1948).

I. L. Cooter, The Calibration of Permanent Magnet

Standards, Reprint 14, 1-3-65, Proceedings of
the 20th Annual ISA Conference, Los Angeles
(1965).
E. M. Purcell, et al, Resonance Absorption by
Nuclear Magnetic Moments in a Solid, Phys.
Rev. 69, p. 37 (1946).
P. L. Bender, R. L. Driscoll, A Free Precession
Determination of the Proton Gyromagnetlc
Ratio, Trans. IRE, 1-7, December (1958).

A. H. Scott, Measurement of MuLtlmegohm Resistors,

NBS J. Res. 50, March (1953).
AC-DC Transfer
Voltage

F. L. Hermach, E. S. Williams, Thermal Converters

for Audio-Frequency Voltage Measurements,
IEEE Trans. Instrumentation, 151, December
(1966).

W. J. Hamer, Standard Cells, Their Construction,

Maintenance, and Characteristics, NBS Mono
graph 84 (1965).
F. K. Harris, Electrical Measurements (Chapter 6,
Potentiometers), a book published by J.
Wiley (1952).
F. B. Sllsbee, F. J. Gross, Testing and Performance
of Volt Boxes, NBS J. Res. 27, p. 269 (1941).
B. L. Dunfee, Method for Calibrating a Standard
Volt Box, NBS J. Res. 67C, January (1963).

F. L. Hermach, et al, A System for Direct and

Alternating Voltage Measurements, Trans. IEEE,
1-14. December (1965).
J. H. Park, A. B. Lewis, Standard Electrodynamlc
Wattmeter and ac-dc Transfer Instrument, NBS
J. Res. 25, p. 545 (1940).
A. W. Splnks, T. L. Zapf, Precise Comparison Method
of Testing a-c Watthour Meters, NBS J. Res.
53, p. 95 (1954).

J. H. Park, Special Shielded Resistor for HighVoltage d-c Measurements, NBS J. Res. 66C,
January (1962).

R. F. Estoppey, Watthour Meter Calibration Re

ferred to d-c Volts and Amperes, Electrical
World, December 1963.

Inductance
B. Hague, Alternating Current Bridge Methods
(Pitmen, 5th Edition, Revised 1957).

Transformers and Comparators

F. K. Harris, Electrical Measurements (Chapter

15, a-c bridges), ibid.

B. Hague, Instrument Transformers, Pitman (1936).

F. B. Sllsbee, Equipment for Testing Current
Transformers, NBS J. Res. 11_, p. 93 (1933).

F. B. Silsbee, Notes on the Design of 4-Terminal

Resistance Standards for Alternating Cur
rents, NBS J. Res. 4, 73, (1930).
B. L. Dunfee, An A-C Kelvin Bridge for the Audio
Frequency Range, IEEE Trans. (Communications
and Electronics), May (1956).
Magnetic Quantities

N. L. Kusters, W. J. M. Moore, The Current Com

parator and Its Application to the Absolute
Calibration of Current Transformers, Trans.
AIEE, 80-111. April (1961).
P. N. Miljanic, et al, The Development of the
Current Comparator, a High-Accuracy a-c Ratio
Measuring Device, Trans. AIEE, 81-1. Nov. (1962).

R. L. Sanford, I. L. Cooter, Basic Magnetic

Quantities and the Measurement of the
Magnetic Properties of Materials, NBS Mono
graph 47 (1962).

N. L. Kusters, W. J. M. Moore, The Development and

Performance of Current Comparators for Audio
Frequencies, Trans. IEEE, 141, Dec. (1965).

R. L. Sanford, E. G. Bennett, An Apparatus for

Magnetic Testing at Magnetizing Forces up to
5000 oersteds, NBS J. Res. 23, p. 415 (1939).

B. L. Dunfee, The Design and Performance of MultiRange Current Transformer Standards for
Audio Frequencies, IEEE, 14-1. Dec. (1965).

27-19

B. L. Dunfee, W. J. H. Moore, An International

Comparison of Current-Ratio Standards at Audio
Frequencies, Trans. IEEE, 14-1. Dec. (1965).
W. K. Clothier, L. Medina, The Absolute Calibration
of Voltage Transformers, Proc. IEEE, 104A,
June (1957).

W. C. Sze, et al, An International Comparison of

Inductive Voltage Divider Calibrations at
400 and 1000 Hz, Trans. IEEE 14-1, p. 124
(1965).
R. D. Cutkosky, J. Q. Shields, The Precision
Measurement of Transformer Ratios, Trans. IRE,
9-1. p. 243 (1960).

F. K. Harris, et al, An International Comparison

of Voltage-Transformer Calibrations to 350 kV,
Trans. IEEE (Communications and Electronics),
January (1964).
W. C. Sze, Comparators for Voltage Transformer
Calibrations at NBS, NBS J. Res. 69C, October
(1965).
J. J. Hill, A. P. Miller, A 7-Decade Adjustable
Ratio Inductively Coupled Voltage Divider with
0.1 Part per Million Accuracy, IEEE, 109B,
March (1962).

Measurements
J. H. Park, Shunts and Inductors for Surge- Current
Measurements, NBS J. Res. 39, p. 191 (1947).
J. H. Park, H. N. Cones, Spark-Gap Flashover
Measurements for Steeply Rising Voltage
Impulses, NBS J. Res. 66C, July (1962).

28-20

Electrical

Calibration

Accuracies

Normal and

at

NBS

best capabilities
INSTRUMENTATION TECHNOLOGY
November 1967

F. L. HERMACH, National Bureau of Standards

The National Bureau of Standards concentrates
on calibration of electrical standards of high
stability and accuracy. This review presents data
on the present ranges and accuracies of basic
electrical measurements, from direct current to
50 kHz. Those interested in NBS capabilities
and services can examine charts which show
accuracies normally available and also the best
attainable.
CALIBRATION, as used in this review, is the
process of making appropriate measurements to
determine the correct value of a standard. NBS cali
brates standards of resistance, capacitance, in
ductance, voltage, and current over many decades of
both magnitude and frequency. The charts contain
the latest data on these variables.
Figure 1 shows the major calibration standards
used by NBS. Directional lines indicate the major
relationships. For clarity, some minor relationships
and the kinds of calibrations are not shown. The
diagram has prototype standards in the top row. The
meter, a unit of length, is defined as a certain number
of wavelengths of the orange-red line of Krypton-86.
The meter bar is a reference standard, rather than a
prototype (Ref. 1). A specified transistion of cesium133 defines time the atomic second, and stable
oscillators serve as working standards of frequency
(Ref. 2). The kilogram is still defined as the mass of
the prototype Pt-Ir standard.
Two experiments at NBS determine the basic elec
trical units in terms of the three prototype standards
and two measured constants, the speed of light in
vacuo c and the acceleration of gravity g. The experi
ments are simple in principle but extremely difficult
and involved in practice because of the required
accuracy.
One experimentto establish the standard ohm
consists of constructing a capacitor from gage bars
and computing the capacitance in electromagnetic
units (about 1 picofarad) from the length of the bars
and the speed of light. With suitable bridges, the
step-up is made at 1,592 Hz to two 10,000 pF
capacitors, across to two 10,000 ohm resistors of
known ac-dc difference, and down to 1-ohm resistors
(Ref. 3). The final step is taken because 1-ohm
Thomas-type resistors are still the most stable im
pedance standards known. The average resistance of a
group of such resistors serves to maintain the ohm
at NBS between absolute determinations.

The other experimentto establish the standard

voltconsists in "weighing the ampere" with a
current balance, and is based directly on the
definition of the ampere in terms of the force between
current-carrying conductors (Ref. 4). One conductor
(coil) is suspended from one arm of a balance, and
the change in force when its current is reversed is
measured in terms of the acceleration of gravity on a
known mass. The voltage drop in a 1-ohm standard
resistor carrying this measured direct current is then
used to determine the emf of a group of saturated
standard cells, which in turn maintain the volt at NBS.
Direct-reading ratio-sets were developed at NBS to
calibrate resistors by the substitution method, and to
step up and down on the resistance scale (Ref. 5). No
line leads to the ratio-sets in Figure 1 because their
accuracy depends on ratios of resistors, not on the
unit of resistance. Standard cells are calibrated by
connecting the known and unknown cells in opposi
tion and measuring the small voltage difference with
a low-range thermofree potentiometer. Potentiometers
and volt boxes are calibrated with universal and
direct-reading ratio-sets (Ref. 5 and 6). With these
the user can then extend the dc voltage scale very
accurately.
AC-DC transfer instruments are comparators for
determining the equality of ac and dc currents, ac
and dc voltages, and ac and dc powers (Ref. 7 and 8).
At NBS they serve chiefly to check other ac-dc com
parators. Since such comparators are very stable, the
user can made accurate ac measurements based on his
known dc standards.
AC bridges extend the scale of capacitance and
inductance measurements, but NBS calibrates only
the more stable capacitors and inductors, with which
bridges can be checked. AC voltage dividers or ac
potentiometric networks are used to determine the
ratios (expressed as complex numbers) of the phasor
voltages and currents of the NBS standard trans
formers and inductive voltage dividers which in turn
are used to calibrate other standards. In principle
the ratios of these networks are also independent of
units; however in practice some of the component
impedors are measured with the dc ratio-sets and ac
bridges. Build-up or "boot-strap" techniques are used
for the newer current comparators and audio-frequen
cy current transformers (Ref. 9 and 10).
The ' standard watthour meters and standards for
magnetic measurements depend on the other
standards as shown in Figure 1 (Ref. 11 and 12).

29-63

Internationally- accepted prototype standards

Krypton - 86
LENGTH

Computable
capacitor

Standard
capacitors
and bridge

Cesium -1 33
TIME

Speed
of light
in vacuo
(c)

Acceleration
of
Gravity
lg)

Oscillator
and frequency
dividers
X
Standard
isto

Current

DC ratio
sets

Standard
cells

DC
potentiometer

I
Standard
inductors
and bridge

AC dividers
and potentiometer
networks

AC- DC transfer
standards
V.I.P

Magnetic
standards

Standard
transformers

Standard DC
voltage divider

3:
Standard
watthour
meters

Figure 1. Major electrical standards used by NBS.

Accuracy at NBS
ResistanceFigure 2 shows the familiar accuracy
triangle with the peak at the value of the 1-ohm basic
Figures 2 to 6 show the uncertainly (accuracy) in the
standard. Note that only decimal multiples and subcalibrations of standards at NBS. Because of the
multiples of 1-ohm are shown. NBS does not usually
wide ranges, logarithmic scales are used. In each
offer calibrations of other values, nor does it maintain
figure the uncertainty is the "limit of error" which
an accurate resistance-measuring facility as such. How
should be exceeded only very rarely. Uncertainty
ever, stable standards of any nominal value can be
data includes the calculated imprecision (at least
calibrated
at greater cost and lesser accuracy than
three times the standard deviation of the average)
standards of neighboring values on the decimal
plus the estimated limits of the systematic errors of
scale. Thomas-type 1-ohm resistors are normally
the NBS working standards and the calibration
calibrated to 1 ppm in terms of the ohm maintained
process. Uncertainty data does not include the
at NBS.
differences between the absolute units and the legal
High-quality direct-reading and universal ratiounits maintained at NBS (Ref. 3 and 4).
sets (not shown in Figure 2) can be calibrated with
The solid lines and circles in the figures show the
an uncertainty not exceeding 1 ppm of input
accuracies normally available in the calibration of
resistance. Using relatively simple equipment and
commercially available standards of the highest grade,
established calibration procedures, many users of
by established techniques, and for the fees given in
these ratio devices could perform these calibrations
the published NBS fee schedules. These accuracies in
with comparable uncertainties. This is also true for
clude allowances for short-time variations in the
many types of range-extending ratio standards. In
standards under test under the controlled conditions
many cases the equipment and techniques are quite
at NBS, but not for possible larger variations in other
involved, so it may be very desirable to verify the
environments nor for long-time drifts or other
results by "round-robin tests" on a traveling
changes. The dotted lines and open circles show the
standard in several laboratories, or by a direct
best accuracies that might be attainable by special
calibration of a few ranges at NBS.
techniques, and with repeated tests made with great
CapacitanceThe chart for capacitance measure
care and effort in the calibration of "ideal standards"
ments, Figure 3, should be three-dimensional, with
which are perfectly stable and free from environmen
frequency as another independent variable. One would
tal influences.
30-64

o Best occuracy
Normal accuracy,
standard fee

1 'O5
m
s io4
I IO3

o
- o

-
Via Wenner bridge

0.1

Via Wheatstone bridge

o

o
o

Vio capacitance discharge-

10"

00001

1.0 Ea.
a.
io >;

Research bridge -^/"sensitivity limit at 1592 Hz

KO8
/ JI^____Expected accuracy o(

0.01

i io6-

10

100 1
1,000
(0.1%)
10,000
(1.0%)

IO"4 IO"2 1.0 IO2 IO4 IO6 IO8 IO10 IO12 IO14
DC resistance, ohms
Figure 2. NBS measures dc resistance.

Transformer rotio-arm bridge

^iransrormer
Resistance ratio-arm
bridge

E
*

KJ4
c

100 f
=>
6OI0IOOO Hz
kHz .
io2! A- for sealed capacitors (ItolO
I kHz)
E,??00 to"lOOOHzi' 10,000
(10%)
B- for unsealed capocitorsl I kHz)
UOOlolOOOHz)
-Contour line for normal calibration
IO12 IO10 IO"8 IO"6 IO"4 IO2 1.0 IO2 IO4
Capacitance, microfarads

Figure 3. NBS measures capacitance.

0.001

0*
expect, in a rough way, an "accuracy cone" with the
peak at 1 picofarad and 1,592 Hz, where the best
measurements with the computable capacitor have
been made.
The dotted line in Figure 3 shows the sensitivity
limit at 1,592 Hz of a new research bridge which is
developed in conjunction with a new calculable
capacitor. The estimated accuracy of the capacitor is
also shown. The accuracy of the research bridge for
stepping up and down from its value of 0.5 picofarad
will be somewhat poorer than the sensitivity limit
indicated. Some of the normal calibrations of the
best available commercial capacitors are shown. Note
the extremely wide ranges in magnitude of 10-9 to
102 uF and in frequency of 60 to 10,000 Hz.
Some sealed three-terminal air capacitors from 10
to 1,000 pF have shown unusual stabilities. Projected
improvements in the NBS ratio arm bridge should
make it possible to measure such capacitors to at
least 10 ppm. Even better accuracy may be required
to calibrate 10 pF sealed capacitors of fused silica
with silvered electrodes, which are apparently stable
to better than 1 ppm/year (Ref. 13).
InductanceStandard inductors are calibrated with
standard capacitors in a Maxwell Wien bridge. Some
of the contours in the accuracy cone are shown in
Figure 4. The available accuracy is generally limited
by the environmental conditions affecting the bridge
and the standards being calibrated.
VoltageThe graphs of voltage measurements, Fig
ures 5 and 6, are complicated not only because of ap
plied frequency but also because NBS does not offer a
voltage calibration service as such. Only the dc voltage
of a standard cell is available directly, and is deter
mined to 1 ppm in terms of the volt maintained at
NBS. The scale is extended in voltage and frequency
by the calibrations of ratio standards and ac-dc trans
fer standards. The user makes accurate dc and ac meaFigures 2, 3, and 4 were contributed by Chester Peterson,
Chief of the N.B.S. Resistance and Reactance Section.
31-65

Best accuracy
Normal accuracy, standard fee
7 0.1 to I kH

io"6

io"5 icr4 io"3 io"2 io"1

Inductance, henries

i.o

io

io'

Figure 4. NBS measures inductance.

0.01
0.1
slondord cells
1.0
E
Q.

Vol I bones
10
*<
Best occurocy
^^m Normol occurocy
I0"3 K)"2 hO"1 1.0 10 O2 IO3
DC voltage, volts

^
100 :
1000

IO4 I05 IO6

Figure 5. NBS measures dc voltage. Accuracy

of V/Vr is shown for ratio standards, Vr = 1 v.

OjOI

id*
T/TTT, Best accuracy
Normol accuracy

Best accuracy
Normal accuracy
Kf6-

01

NBS research transformers

ft current comparators

1.0

a.
a.

DC resistors
OCk.
(For use with potentiometer)
^^sg^I
AC-OC transfer standards
KnHT |
if ifH- 5 Hz to 50kHz

p
KX) S
n

400 H;
1000

AC-OC transfer standards

5 Hz to 50kHz
hO"3 K)2 10"' 1.0 K) I02 I03
AC voltage, volts

Current
I04

I05

K)"5 I0"4 I0"3 I0"2 10"' LO 10

Current, amperes

rO6

KJ2 I03 I04

Figure 6. NBS measures ac voltage. Accuracy

of V/Vr is for ratio standards, Vr = 100 v.

Figure 7. NBS measures current. Accuracy of

I/Ir is shown for ratio standards, Ir = 5 amp.

surements with these standards and a standard cell.

Potentiometers, fixed resistance dividers, and trans
formers are ratio standards which are generally used
with some reference voltage. The emf of a standard
cell would be used with dc ratio standards, and
approximately 100 volts for ac standards. The ratio
standards have upper and lower voltage limits. The
charts therefore show the accuracy of the ratio V/Vr,
where V is given along the abscissa, and Vr is this
reference. Note that this is the accuracy of the
indicated ratio, not of a fraction of the applied
voltage. A 1,000/1 volt box certified to 0.001 percent
of ratio has one hundredth the uncertainty of a 1ppm-of-input voltage divider at this same ratio.
Dotted lines show the "attainable" accuracies. Normal
calibrations are as follows:

References

potentiometers of the highest grade

to 3 ppm at 1 volt
volt boxes
to 10 ppm of ratio
decade transformers
to 0.5 ppm of input voltage at 1 kHz
voltage transformers
to 0.03 percent of ratio at 60 Hz
ac-dc transfer instruments for voltage
to 50 ppm from 20 Hz to 20 kHz,
and 100 ppm at 5 Hz to 50 kHz.
CurrentThe chart for current measurements, Fig
ure 7, is similar. NBS offers only calibrations of dc
shunts (for use with a potentiometer), ac ratio stan
dards, and ac-dc transfer standards. A chart for power
and energy measurements would be based on the NBS
standard transfer wattmeter, accurate to better than
50 ppm of the applied volt-amperes at 60 Hz and 0.1
percent at 2 kHz. The five NBS standard watthour
meters are calibrated periodically in terms of the dc
standards through this wattmeter, and are in turn
used to calibrate standard watthour meters to 0.05
percent at 60 Hz.

1. "Actions of 11th Conference on Weights and Mea

sures," NBS Technical News Bulletin, 44, 199, Dec. 1960.
2. "World Sets Atomic Definition of Time," NBS Tech
nical News Bulletin, 48. 209, Dec. 1964.
3. Cutkosky, R. D., "Evaluation of the NBS Unit of
Resistance Based on a Computable Capacitor," Journal
of Research NBS, 65A, 147, 1961.
4. Driscoll, R. L., and Cutkosky, R. D., "Measurement
of Current with the NBS Current Balance," Journal of
Research NBS, 60 297, 1958.
5. Thomas, J. L., "Precision Resistors and Their Mea
surements," NBS Circular, 470, 1948.
6. Dunfee, B. L., "Method for Calibrating a Standard
Volt Box," Journal of Research NBS, 67C, 1963.
7. Hermach, F. L., and Williams. E. S.. "Thermal Con
verters for Audio-Frequency Voltage Measurements of
High Accuracy," IEEE Transactions on Instrumentation
and Measurement, Vol. IM-I5, No. 4, Dec. 1966.
8. Park, J. H., and Lewis, A. B., "Standard Electrodynamic Wattmeter and AC-DC Transfer Instrument,"
Journal of Research NBS. Vol. 25, p 545, 1940.
9. Kusters, N. L.. "The Precise Measurement of Cur
rent Ratios," IEEE Transactions on Instrumentation and
Measurement, Vol. IMl3, No. 4, Dec. 1964.
10. Dunfee, B. L.. "The Design and Performance of
Multirange Current Transformer Standards for Audio
Frequencies," IEEE Transactions on Instrumentation and
Measurement, Vol. IM-14, No. 4, Dec. 1965.
11. Spinks, A. W.. and Zapf, T. L., "Precise Comparison
Method of Testing AC Watthour Meters." Journal of
Research NBS. 53, 95. 1954.
12. Sanford, R. L.. and Cooter. I. L.. "Basic Magnetic
Quantities and the Measurements of Magnetic Properties
of Materials." NBS Monograph 47, 1962.
13. Cutkosky. R. D.. and Lee, L. H.. "Improved 10-Picofarad Fused Silica Dielectric Capacitor," Journal of Re
search NBS. Sec. C. Vol. 69C. No. 3, July-Sept. 1965.
Francis L. Hermach is Chief of the Electrical Instruments
Section at the National Bureau of Standards, Washington.
D. C. Article is based on Paper M5-1-MESTIND-67
presented at the 22nd ISA Conference & Exhibit, Chi
cago, 111., Sept. 11-14. 1967.

32-66

UNITED STATES DEPARTMENT OF COMMERCE Sinclair Weeks, Secretary

NATIONAL BUREAU OF STANDARDS A. V. Aitin, Director

Suggested Practices for

Electrical Standardizing Laboratories

Francis B. Silsbee

National Bureau of Standards Circular 578

Issued August 30, 1956

33-1

Preface
Since World War II there has been a very marked increase in the number
and scope of private standardizing laboratories throughout the United States.
The function of each such laboratory is to maintain the accuracy and uniformity
of the measuring instruments and apparatus used by the organization or
organizations that the laboratory serves. A basic phase in this process is
necessarily to correlate the reference standards of the laboratory with those of
the National Bureau of Standards. This Circular has been prepared to sug
gest techniques and principles that experience has shown to be useful in such
operations. Although this Circular covers explicitly only the field of electrical
measurements, many of the principles involved are equally applicable in other
kinds of measurement.
A. V. Astin, Director.

Contents
Preface
1. Introduction
2. General principles
2.1. Competence of personnel
2.2. Position in organization
2.3. Appropriate scope and accuracy
2.4. Laboratory atmosphere
2.5. Care of standards
2.6. Timing of calibrations
3. Reference standards
3.1. Standard cells
3.2. Resistors
3.3. Inductors
.3.4. Capacitors
3.5. Volt boxes
3.6. Time standards
4. Working standards
4.1. Standard cells
4.2. Resistors
4.3. Capacitors and inductors
4.4. Volt boxes
4.5. Indicating instruments
4.6. Watthour meters
4.7. Instrument transformers
5. Comparison apparatus
5.1. Consoles
5.2. Direct-current bridges, direct-reading ratio sets, universal ratio sets, directcurrent potentiometers
5.3. Lindeck potentiometer
5.4. Voltage-calibrating transformers
5.5. Instrument-transformer test sets
6. Interlaboratory standards
6.1. Shipment
6.2. Standard cells
6.3. Resistors, capacitors, and inductors.
6.4. Instrument transformers
6.5. Indicating instruments
7. Summary
8. References

34-11

Page
"

J
*
~
*
*
*

4
*
*
j
!
jj
jj
jj
jj
jj
','

7
<
7
7
7

s
jj
9
)

Suggested Practices for Electrical Standardizing Laboratories

Francis B. Silsbee
A number of basic principles are given that experience has shown to be important in
the operation of private standardizing laboratories. Types of standard apparatus are
classified and schedules appropriate for their intercomparison and for their checking at the
National Bureau of Standards are suggested.

1. Introduction
The purpose of this Circular is to indicate
recommended procedures and schedules for the
calibration testing of electrical instruments and
measuring apparatus used in the range of fre
quency from 0 to 30 kilocycles per second.
For many years the electric-power companies
have maintained large and efficient laboratories
[1, 2] 1 suited to their particular needs, and certain
universities have similarly supplied services [3]
to their teaching and research staffs. More
recently, laboratories are being established in
many newly founded manufacturing plants, in
firms concerned with research and development,
and in Government agencies (particularly in the
military services). Although the particular prob
lems that confront the various agencies may differ
greatly, their standardizing laboratories have a
great many common functions in their work of
maintaining and disseminating the units of
measurement.
The process of measurement always involves a
succession of steps in which the units in terms of
which the measurement of any quantity is ex
pressed are transmitted by using measuring ap
paratus that has been standardized, by compari
son with standards of a higher level, to calibrate
other measuring apparatus which is to serve as a
standard at a lower level. At the source is the
National Bureau of Standards, authorized by Con
gress to establish and maintain the basic standards
of the nation [4, 5]. Next, in any organization,
may come a primary group of laboratory workers,
which maintains its basic reference standards and
in turn calibrates the working standard apparatus
used by the next lower, or secondary, group, which
in turn calibrates the instruments used in making
measurements in the plant or shop. In some cases
these latter two operations are merged in a single
In brackets Indicate the literature references at the end of this
39082536
35-1

group, while, at the other extreme, a single pri

mary group may serve many secondary groups
located in different cities.
It is the rest of the organization that uses the
shop instruments standardized by the workingstandards group, and carries on the day-to-day
work in the power systems, manufacturing opera
tions, research studies, or acceptance tests. This
use insures proper billing, adequate quality, and
valid criteria for the acceptance of manufactured
articles. In this Circular the terms "laboratory"
or "standardizing laboratory" will be used to indi
cate the segregated part of the complete organiza
tion that is charged with the duty of supplying
calibration services, at either the reference or the
working-standard level, to the measuring appara
tus used by the rest of the organization, which
will here be designated as "plant" or "shop,"
even though it may actually be more correctly
described as a research or testing laboratory.
For clarity in the rest of this Circular, the meas
uring instruments and apparatus used in an indus
trial organization will be referred to under the
following categories:
(a) Reference standards, which embody fixed
electrical quantities and serve to maintain thenrespective units during the intervals between
checks with a laboratory at a higher echelon.
(b) Working standards, calibrated in terms of
the reference standards and reserved for use in
the calibration of shop instruments and apparatus.
Working standards should never be used directly
for measurements in the plant or shop.
(c) Comparison equipment such as bridges,
potentiometers, etc., and consoles used in com
paring shop instruments with working standards.
(d) Interlaboratory standards, often similar in
construction to reference standards, used for trans
port to NBS and return in the periodic checking
of the local reference standards.
(e) Shop instruments and measuring apparatus
calibrated in the standardizing laboratory and
used in the day-to-day operation of the plant.

2. General Principles
The following general principles have been
found by experience to be applicable to many
of the problems involved in the maintenance of
accuracy in laboratory work and any program
of calibration service testing should conform to
them so far as possible.
2.1. Competence of Personnel
The person in direct charge of any standardiz
ing laboratory must be thoroughly familiar with
the fundamentals of electricity and magnetism
and with the principles of operation of all types
of electrical measuring apparatus [8, 9]. To be
competent to carry on within the laboratory the
various sequences of calibration testing outlined
below, he must have taken a college-level course
in electrical measurements or have had laboratorjr
experience equivalent thereto. It is verv desir
able that he be a graduate in physics or electrical
engineering from an accredited college. He must
be mentaSy alert to detect discrepancies and
inconsistencies in internal check tests and to rec
ognize abnormal behavior of any part of the
apparatus or circuits. Unless persons of this
type are available in a standardizing laboratory,
both its calibration service and the measurement
work that depends upon it throughout the plant
is certain to become unsatisfactory, regardless of
how often particular pieces of apparatus may bo
tested by other persons in some higher echelon
or at NBS.
2.2. Position in Organization
Experience through the years has shown that
the most satisfactory results are obtained by
maintaining a definite separation of the standard
izing laboratory from the rest of the organization.
Responsibilities for accuracy are thereby made
definite, and team spirit can be developed in the
staff of the standardizing laboratory that keeps
them interested in their work, which might other
wise seem routine and unimportant. This sepa
ration protects them from pressures that demand
improper haste and sloppy workmanship, and
gives them authority directly delegated from a
high level in management.
Closely related to the foregoing is the valuable
principle that team spirit can be developed and
maintained by recognizing that the quality of
the reference standards of a laboratory are to a
considerable extent a measure of the competence
of its staff ("a worker is known by the sharpness
of his tools"). Hence, in cases where a given
large organization maintains a plurality of stand
ardizing laboratories in its plants in different
cities, it is very desirable that each reference and
interlaboratory standard should be assigned per
manently to an individual standardizing labora
tory, and returned to it following periodic checks
at any laboratory of a higher echelon. If this

is not done, it becomes difficult to fix responsibility

for errors that may be discovered subsequently,
and the laboratory workers lose the basis for
much of their pride in workmanship.
2.3. Appropriate Scope and Accuracy
The nature and accuracy of the work done in a
standardizing laboratory may vary considerably
with the size of the plant it serves and the nature
of the output of the plant. However, there are
very few modern activities in which some phases
do not require measurements of rather high
accuracy. A standardizing laboratory must, of
course, be equipped for calibration measurement?
of a higher accuracy than any used elsewhere in
the plant, because the accuracy of measuremen:
can never rise above that of its source.
Many such local standardizing laboratories
provide service in many scientific fields, e. g
mechanics, heat, optics, and chemistry, in addi
tion to their standardizing work in electricity
and magnetism, which alone is the subject of
this Circular. The direction and extent of such
activities, as well as the degree of specialization
in particular branches of electrical science, may
greatly affect the types and ranges of the electrical
standards required.
2.4. Laboratory Atmosphere
Atmospheric conditions are of considerable
importance in an electrical standardizing labora
tory. If the relative humidity exceeds 65 per
cent, a conducting film of moisture is adsorbed
on the surface of many types of insulating mate
rial. Leakage over such surfaces, particularly to
the circuit of a sensitive galvanometer, may intro
duce serious errors.
Certain types of apparatus, for example some
wire-wound resistors of high value, show changes
of several parts per 104 with prolonged changes in
humidity. Such variations can be detected by
comparing calibrations made near the middle of
the winter heating season, during which indoor
humidities are normally very low, with those made
near the middle of the summer when the ambient
humidity is high.
Most electrical apparatus is compensated for
temperature to some extent. However, if the
ambient temperature is subject to changes, the
response of the various components may not be
equally prompt. In such circumstances the com
pensation may be temporarily ineffective. It is
therefore important that the laboratory tempera
ture be held fairly constant (2 deg F) at all
times. Such room control also makes easier the
task of the thermostats that control the baths in
which more temperature-sensitive apparatus, such
as standard cells or standard conductivity samples,
are kept. The particular temperature chosen is
of less importance. The conditions specified by
the American Society for Testing Materials for its

36-2

"Standard Laboratory Atmosphere" are a relative

humidity of 50 2 percent and a temperature of
23 1.1 deg C (73.4 2 deg F). This combi
nation is widely used in tests of dielectrics and
other materials, and, in addition, is very close to
the maximum of comfort for the workers. Hence
it is a good target to aim at in designing a new
laboratory.
Freedom from vibration and dust are also im
portant. A shop in which electrical instruments
may be opened for repair should be scrupulously
clean and dust free. Magnetic particles are
attracted to the air gaps of permanent magnet
instruments and must be constantly guarded
against. Effective air filtering or electrostatic
precipitation are very desirable.
2.5. Care of Standards
In many categories, a good standard piece of
apparatus improves with age but deteriorates
with handling and shipping. The shipping of
primary and working standards and of comparison
equipment can be minimized and the transfer of
the units be achieved best by the use of interlaboratory standards of particular types that are as
rugged as practicable.
The adjustment of a piece of measuring equip
ment to be close to its nominal value is very often
desirable in the shop instruments used in the
plant, because of the great saving in time and re
duction in possible erroneous applications of cor
rections that result. On the other hand, the exact
opposite is usually true of working and reference
standards. In general, such standards should not
be readjusted, even if their values have gradually
drifted materially away from their nominal values.
Any standardizing laboratory naturally maintains
records of the values of its reference standards [10].
These are preferably kept on an individual card
or record sheet for each standard, so that its his
tory is apparent at a glance. Each such record
provides a valuable indication of the quality of the
standard, and gives warning of deterioration and
the need for replacement toward the end of its
useful life. Frequent adjustments tend to inter
fere with such records and increase the proba
bility of improper corrections, as the proper cor
rections before and after a slight readjustment
may not be conspicuously different. Even more
important is the risk that the adjustment may
initiate a progressive drift in value of the standard
as a result of introducing mechanical strain, or
local heating.
2.6. Timing of Calibrations
The most desirable sequence of steps and the
frequency with which comparisons of measuring
apparatus should be made depend, of course, upon
a great number of circumstances, such as the volume

of equipment to be tested, the delicacy of the ap

paratus, and particularly the type of personnel
by which it is to be handled. (For measurements
made by inexperienced students or by military
recruits, the likelihood of damage is much greater
than when a laboratory is staffed by experts.)
In every laboratory a real possibility is always
present that an abrupt change may occur in some
standard as a result of an unrecognized accident
or abuse. Fairly frequent checks made locally
are therefore an indispensable supplement to any
program of calibrations to a higher echelon.
Cooperative arrangements with a similar lab
oratory in the same city may allow frequent checks
to be made against an independently maintained
group of standards. This may permit a much less
frequent checking schedule between either labora
tory and one of higher echelon.
In any laboratory, the accuracy needed in meas
urement may differ widely on different projects.
Hence the frequency with which the working in
struments should be checked may vary consider
ably from project to project, even though instru
ments of the same accuracy class are used [2, 6, 7].
The time intervals suggested elsewhere in this
Circular for check tests represent in most cases
the considered opinion of persons who have had
long experience with standard apparatus of the
types now commonly used. Similar test intervals
are listed in the American Standards Association
Code for Electricity Meters [2]. When a new
type comes into use, it must be appraised by being
checked at frequent intervals during its first few
years of service. After its good stability has thus
been demonstrated, the interval between tests
may safely be made longer. Several types of
standard capacitors and inductors are currently
in this probational stage.
A procedure that is often of value, particularly
for the checking of consoles and complex equip
ment such as a-c bridges, is to use an interlaboratory standard that is measured in one laboratory
and shipped to a laboratory of lower echelon for a
repeat measurement, the lower laboratory not
initially being aware of the value obtained at the
higher. This procedure can detect significant de
fects in the wiring or insulation of the test console,
improper procedure and careless operations at the
lower-level laboratory, as well as changes or de
terioration in the reference or working standards
of the lower laboratory.
When systematic local intercom parisons indicate;
changes or abnormal performance in primary or
working standards, checks with a higher echelon
are in order regardless of whether any formal
scheduled time for such a check has been reached.
If the local periodic checks uncover unsteadiness
of operation, large sudden changes in value, or
other indication of a constructional defect in the
reference or working standard, this standard
should be sent to its manufacturer for repair
before it is submitted to NBS for calibration.

37-3

3. Reference Standards
The function of the reference standards of a
laboratory is to maintain locally a continuity of
value in the units of measurement that they
embody. Successive comparisons with the higherechelon standards of NBS by means of interlaboratory standards shipped back and forth will give
an indication of any slow drift of the reference
standard. When this cumulative drift exceeds
the confidence interval of the values derived by
the comparisons with the interlaboratory stand
ards, the value assigned to the reference standard
is corrected accordingly. The reference standards
are used primarily to calibrate periodically the
working standards of a lower echelon, such as
indicating instruments. Also, on occasion they
may be used as working standards themselves.
It is evident that both reference standards and
working standards should be as permanent and
reliable in construction as possible. Definiteness
and repeatability are of major importance, whereas
sensitivity, low losses, and freedom from extrane
ous influences are relatively much less important.
Closeness of adjustment to nominal value is of
decidedly minor importance. An undetected
change in a reference standard may easily initiate
a chain of error that will propagate throughout
the plant and cause losses in time and material
exceeding manyfold the original cost of the stand
ard. Hence, these reference standards should be
fiurchased only on very strict specifications, and
rom manufacturers of high reputation and of long
experience in producing shop instruments and
apparatus of demonstrated permanence.
3.1. Standard Cells
A large and important laboratory, which is
expected to need a standard of voltage with an
accuracy of 0.002 percent and to certify the emf of
unsaturated standard cells used in its plant or by
its subcontractors, normally maintains as a refer
ence standard a group of 5 or 6 cadmium standard
cells of the saturated type. These cells are rela
tively permanent but have a temperature coeffi
cient of about 0.005 percent per degree Celsius
(centigrade) at room temperature. They must
therefore be kept in a bath thermostatted to
0.01 deg C [11, 12]. A highly refined, waterwhite, acid-free mineral oil, having a viscosity of
about 0.25 poise at 25 C and a flash point of 170
C, has been found suitable for baths for standard
cells and for resistors. Such a group of saturated
cells should preferably be checked initially at NBS
and its value reassigned annually for the first 3
years and biennially thereafter, by using a second
group of 2 or 3 saturated cells as an interlaboratory
standard. The individual cells of the reference
group can be intercompared monthly. The refer
ence group is used to check the laboratory's work
ing-standard cells of the unsaturated type and,
on occasion, shop-standard cells also.

For laboratories requiring an accuracy of not

over 0.01 percent in their standard of voltage, the
reference standard often consists of a group of
three cadmium standard cells of the unsaturated
type. The individual cells are intercompared
weekly by connecting them by pairs in series oppo
sition and measuring the differential emf. These
cells have a very small temperature coefficient and
can be shipped safely by parcel post (if carefully
packed). However, they are less constant th&n
saturated cells and their emf usually decreases aU
rate ranging from 40 to 120 microvolts per year
Therefore, their values should be reappraised at
least annually on the basis of a periodic check from
a higher echelon. A second group of 2 or 3
unsaturated cells can be used as an interlaboratory
standard for such checks. When the emf of an
unsaturated standard cell has fallen below 1.0183
volts, it is approaching the end of its useful life
and can no longer be used as a reference or an
interlaboratory standard. Following a moderate
change in temperature, many cells tend to show s
very considerable change in emf, which may per
sist for several days. This thermal "hysteresis''
must be guarded against by allowing cells to stand
for some time after shipment before taking read
ings. Temperature troubles with unsaturated
standard cells can be minimized by keeping themiii
a thermally lagged copper-lined box. This re
duces temperature fluctuations and differences in
temperature between the two electrodes. It is
important that the leads be brought out through
very high grade insulation to an external termini
board. [13,14].
3.2. Resistors
If the laboratory possesses at least two standard
resistors of each decimal value covering the range
over which it expects to make accurate measure
ments, one of these can be submitted to NBS
every 2 years as an interlaboratory standard, while
the other resistor of each pair remains undisturbed
as a reference standard in its laboratory.
3.3. Inductors
Many laboratories find it useful to possess two
fixed standard inductors of each decimal value
over the range it expects to cover with accurate
measurements. In recent years the quality of
standard inductors has been greatly improved,
and this Bureau does not yet have sufficient data
on the stability of the newer types to make &
definite estimate of their expected stability. It
is suggested therefore that one inductance standard
of each denomination be submitted annually te
the NBS, until a sufficient history is obtained to
predict its performance.
3.4. Capacitors
Reference standards of capacitance include smali
fixed air-dielectric units, precision-variable air
capacitors, and solid-dielectric capacitors using
mica or an equivalent dielectric. Fixed standards
1,000 picofarads (micromicrofarads) or less in

38-4

value must be of three-terminal construction in

order to avoid uncertainties due to stray capaci
tances. Precision-variable air capacitors are pre
ferably hand carried, although shipment of the
capacitor in its wooden container within a padded
carton is usually satisfactory. Variable capacitors
with noticeable backlash must be adjusted or
repaired before submission to NBS for calibration.
Capacitors with worm reduction gear are cali
brated at the cardinal points corresponding to each
whole revolution of the worm wheel. If the worm
is eccentric, capacitance increments between
cardinal points will depart from linearity, but it is
usually not necessary to calibrate the vernier dial
for every turn.
Each laboratory should possess two capacitance
standards of each decimal value and type needed
to cover the range of concern. After its initial
check at NBS, one standard of each value can be
kept as a reference standard while the other is
submitted to NBS at regular intervals as an
interlaboratory standard. This interval should
initially be 1 year for any one pair of standards.
After the first 3 or 4 calibrations, an examination
of the record will indicate the appropriate future
frequency of calibration, taking into account the
actual accuracy demands made on particular
standards. On variable air capacitors the eccen
tricity correction ordinarily will be determined
only during the initial calibration; thereafter,
calibration at a few cardinal points is usually
sufficient.
3.5. Volt Boxes
Each laboratory should have one volt box with a
plurality of ratios, which can serve as a reference
standard. This can be checked initially at NBS.
Unless the laboratory is rather large, this same volt
box may be rechecked at NBS at intervals of 2
years, and thus serve as an interlaboratory
standard also.
3.6. Time Standards
Although time is not an electrical quantity,
many electrical laboratories require reference
standards of time or of frequency. A high-grade
seconds pendulum clock with photoelectric pickup
or a standard crystal-controlled oscillator may
form the reference standard. Either can be
calibrated by reference to the standard-frequency
radio signals emitted by NBS stations WWV or
WWVH.2 By the use of a multivibrator in com
bination with the reference-standard oscillator, the
frequency of working-standard oscillators can be
calibrated over a wide range. Such an oscillator
can be used to control a standard frequency
circuit to which, in turn, synchronous timers can
be connected to serve as working standards for
measuring time intervals. The use of the fre
quency of local electric-power circuits as a time
standard, while very convenient, may be subject
> For information on this service, consult Radio Standards Division,
National Bureau of Standards, Boulder Laboratories, Boulder, Colo.
315-411 O - 68 - 4

to errors approaching 1 percent for short periods

of time, even though the average frequency over
a longer time as shown by a clock is very high.
These short-time fluctuations in frequency are
materially less if the supply is tied in synchronism
with a large power system.
4. Working Standards
The working standards constitute the principal
tools of the standardizing laboratory. They are
calibrated at intervals by comparison with the
reference standards, and used in the daily work of
checking shop instruments. The number needed
of any one kind and range will depend upon the
volume of testing service demanded in that range.
If the volume is very small, and also in special
cases where extreme accuracy is needed, a refer
ence standard can be used as a working standard
also.
4.1. Standard Cells
Most laboratories will need a number of un
saturated cadmium standard cells to serve as
working standards and relieve their reference cells
of excessive use and of the hazard of accidental
abuse. Such working-standard cells should be
checked against a reference standard at intervals
of 1 or 2 weeks.
4.2. Resistors
Depending upon the nature of the work in the
laboratory, there will probably be required, in
addition to sets of fixed standard resistors, a num
ber of dial-type resistance boxes, perhaps including
resistors in the megohm and multimegohm ranges,
and also resistors or shunts capable of carrying
larger currents than are appropriate to the refer
ence standards. These working standards may
initially be tested at NBS to determine the effects,
if significant, of current and temperature on their
resistance. Thereafter these working standards
need to be checked, using a moderate current, at
intervals of about 6 months by comparison with
the reference standards of resistance, using a
direct reading ratio set, Wheatstone bridge, or
double ratio set [15], or the potentiometer method.
If it is believed that a resistor has been overloaded,
it should be checked against the appropriate refer
ence standard without delay, ana if a significant
change in value has occurred since the last regular
check, the overloaded resistor should be checked
at frequent intervals until its resistance again
becomes steady.
Most standard resistors are made of manganin.
This alloy has the valuable property of showing
low thermal electromotive force to copper and of
changing relatively little in resistance with change
in temperature. The resistance, R at a tempera
ture / C is related to that, RK, at 25 C by the
formula
R,=R2s{l + a(t-25)+Kt-25)1}.
Here the coefficient a is usually less than 10 X 10-4
and 0 usually lies between 3 X 1 0"7 and 6 X 1 0~7.

39-5

4.3. Capacitors and Inductors

4.6. Watthour Meters

Each laboratory should compare all of its work

ing standards with its reference standards once a
year, and also immediately after an interlaboratory
group has been calibrated at NBS. Working
standards that are used frequently, or upon
which great dependence is placed, may be com
pared with the laboratory reference standards
whenever an important series of plant calibrations
is undertaken.

Laboratories that have occasion to test large

numbers of watthour meters, as do those of power
companies, are best guided by the Electrical
Metermen's Handbook [1]. This book was pre
pared by a committee of experienced meter
engineers for the instruction of meter-laboratory
personnel.
If the laboratory has occasion to test watthour
meters only rarely, it is sufficient to have available
an electrodynamic wattmeter of suitable range
and a standard of time accurate to the degree
needed in the energy measurement. The watt
meter can be calibrated by using direct current
and a potentiometer, and then used to hold a
known constant power while the revolutions of
the watthour meter are timed.
In intermediate cases it may be desirable to
install a group of three working-standard watthour
meters. These can be calibrated as indicated in
the preceding paragraph, and each can be used to
check a number of other meters. To minimize
friction they should be used without register
mechanisms but with photoelectric pickups. The
complete but time-consuming calibration again?'
the wattmeter need be made only at intervals of a
month or so, provided quick intercomparisons
among the three working-standard meters shows
no relative change in their rates. Provision can
be made by means of suitable precision current
transformers so that the standard meter can be
operated always at about its full-load speed, even
when the meter under test is at light load.

4.4. Volt Boxes

A small laboratory may find it sufficient to use
its reference volt box as a working standard also.
This can be checked initially and at intervals of
2 years at NBS. It is well to measure and record
the resistance of each section of each volt box
initially, and monthly thereafter, as a means for
detecting possible changes such as might be caused
by inadvertently overloading one of the lowervoltage ranges. More often other working-stand
ard volt boxes will be used as auxiliaries to poten
tiometers for the calibration of both workingstandard indicating instruments and shop instru
ments. These can De compared monthly with the
reference-standard volt box by a null method [16].
4.5. Indicating Instruments
Direct-current and alternating-current amme
ters, voltmeters, and wattmeters of either the
0.1-percent or the 0.25-percent class [17] will be
needed for the range of current, voltage, and
power over which the shop instruments used in
the plant are to be checked. The a-c instruments
must be of the electrodynamic, electrostatic, or
electrothermal (thermocouple) types, which can
be used on direct as well as on alternating current
[6, 8, 18]. The usual a-c instrument of the
moving-iron type is not suitable as a transfer
standard. The working-standard indicating in
struments can be submitted to NBS for an initial
test, and the a-c instruments for the additional
determination, on appropriate ranges, of the
ac-dc difference by comparison with a transfer
instrument. This ac-dc difference test should
cover the full range of frequency over which the
instrument is likely to De used. After this
initial test, they can be retained in the laboratory
and checked in terms of a standard cell and
standard resistor, using a potentiomenter. Pref
erably, the frequency of these checks ranges from
2 weeks to 2 months, depending upon the fre
quency of use of the working standards and their
reliability as indicated by earlier check tests.
Direct-current working-standard indicating in
struments, when built into a console, usually have
terminals so arranged that the instrument and its
range-extending resistors can be checked by using
a potentiometer and a bridge. Alternatingcurrent instruments may have to be checked by
comparison with interlaboratory standard instru
ments of a multirange type.

4.7. Instrument Transformers

Many laboratories will need a set of multirange
current and voltage transformers covering the
range of current and voltage over which other
shop transformers are to be calibrated. An initial
calibration at NBS, using the burden of the stand
ard circuit of the transformer testing set (plus
ammeter or voltmeter) , can be made at 60 cycles
per second. The errors of current transformers in
general are smaller at higher frequencies, but the
initial tests should include tests at 400 or 800
cycles per second if the transformer is to be used
at such frequencies. Subsequent tests at NBS
need be made only at intervals of 5, or even 10,
years.
5. Comparison Apparatus
The term "comparison apparatus" includes
equipment by means of which tne calibration of a
shop instrument or standard is checked by com
paring it with an appropriate working standard
of the laboratory. In many cases the working
standard is substituted for the device under test
in the same circuit of the comparison equipment,
and the change in its indication is taken as the
measure of their difference. Such substitution
methods are in general capable of very high ac
curacy, and should be used wherever practicable.

40-6

r-

In other cases a working standard is in effect built

in as part of the comparison equipment. Ex
amples of the latter are (1) the working standard
indicating instrument built into an instrument
testing console and (2) the rheostat arm of a
Wheatstone bridge when used directly (as con
trasted with its use by substitution).

the bridge or ratio set to measure or compare

interlaboratory-standard resistors. A potentiom
eter can be given a rough check at least annually
at one point by using it to measure a standard
cell of known emf or, alternatively, b}' using a
single cell first as the standard by which to adjust
the potentiometer current and then as the
unknown to be measured.

5.1. Consoles
5.3. Lindeck Potentiometer
These" devices, containing appropriate sources,
adjusting transformers and rheostats, and panel
instruments for approximate adjustment, will pro
vide the circuits for the comparison of shop indi
cating and recording instruments with workingstandard instruments [19, 20]. In some types of
console the working-standard instruments and
their auxiliary range-extending apparatus also are
permanently built into the equipment [20]. If
suitable special terminals are available, these
working standards can be calibrated like portable
standards. If not, some secondary procedure must
be set up by which the console in effect tests other
working standards that have previously been
checked over the full range. The console can be
shipped direct from the manufacturer to the lab
oratory. The person in charge of calibration
work at the laboratory can then satisfy himself,
by appropriate measurements of insulation re
sistance and of circuit resistance, that the connec
tions are correct and that the leakage and lead re
sistance are not such as to introduce errors. In
fgeneral, these particular hazards are less if the
eads can be run directly in the open between the
working standard and the instrument under test.
The working-standard indicating instruments
are sometimes mounted at an angle of 45 for
greater convenience in reading. This arrange
ment, however, tends to introduce additional
pivot friction. Care must also be taken to insure
that no ferromagnetic material is located near
enough to either the instrument under test or
the standard instrument to affect its calibration.
Even nonmagnetic metal supports can cause
trouble by providing eddy-current circuits which
can affect unshielded a-c instruments. Stray
magnetic fields from supply transformers and
rheostats (particularly those wound on enameled
steel tubes) must be eliminated. The familiarity
that the supervisor will obtain by carrying on
this acceptance test forms an essential part of his
training for the job. His success in it can be
verified adequately by the use of a group of indi
cating instruments as interlaboratory standards
in annual tests. The local checks of insulation
and of lead resistance should be repeated every
6 months.
5.2. Direct-Current Bridges, Direct-Reading Ratio Sets,
Universal Ratio Sets, Direct-Current Potentiometers
These can be
most cases need
vals of 3 years.
can be made at

tested initially at NBS and in

only to be resubmitted at inter
During the interim, local checks
intervals of 6 months by using

Any laboratory will find very useful a combi

nation of a standard resistor and milliammeter
(of the 0.25% class) to use as a low-range Lindeck
potentiometer for the purpose of intercomparing
the various standard cells in the laboratory by
measuring the differences between them in pairs.
The resistor and milliammeter can then be
checked on the same schedule as the workingstandard resistors and the working-standard in
struments, respectively [12, 21, 22]. Care must
be taken to minimize thermal emf in such a circuit.
5.4. Voltage-Calibrating Transformers
The ratios of the voltages of the various tapped
sections of the secondary winding to the voltage
of the tertiary winding to which the standard
voltmeter is connected can be determined initially
at each operating frequency. This should be
done with no load and with rated load on the
secondary while the standard voltmeter, or an
impedance duplicating it, is connected to the ter
tiary. Subsequent tests need be made only at
intervals of 5 or even 10 years.
5.5. Instrument-Transformer Test Sets
The instrument-transformer test sets can easily
be checked annually at the 100-percent point and
also at one other point.3 These checks can be
supplemented by using the test set and workingstandard instrument transformer to measure once
a year the ratio and phase angle of an inter
laboratory-standard instrument transformer of
the same range, which is checked at NBS every
5 years. These test sets are currently available
for use at 60 and 25 cycles per second only.
Radical modifications in procedure or in compo
nent values are required if they are to be used
at other frequencies.
6. Interlaboratory Standards
Interlaboratory standards in general are similar
in nature and inherent accuracy to reference stand
ards and working standards. The smaller labor
atories may well use some of their reference stand
ards to serve as interlaboratory standards also.
These should, so far as possible, be of rugged
construction to minimize damage and change of
value in shipment. On this account they usually
are standards of fixed value rather than contin
uously adjustable devices.
1 Simple methods for such one-point checks on instrument-transformer
test sets are now being developed at the Nutional Bureau of Standards.

41-7

The function of an interlaboratory standard is

to transmit some one of the electrical units of
measurement from a laboratory of higher echelon,
such as the NBS, to the local laboratory. It is
sent systematically to NBS for a calibration and
is compared before and after this operation with
the appropriate reference standard of the local
laboratory. In those cases in which a large organ
ization maintains a plurality of separate stand
ardizing laboratories under its authority, a single
set of interlaboratory standards may profitably
be circulated to give a "round-robin test" by
being sent to 3 or 4 of the company's laboratories
in succession between trips to NBS. Such a pro
gram provides a check both on the comparison
apparatus and on the operating competence of
the personnel at the various laboratories. It also
performs the function of maintaining the assigned
values of the reference standards of each laboratory
in concordance with NBS standards.
6.1. Shipment
Electrical measuring instruments such as am
meters, voltmeters, wattmeters, and watthour
meters contain extremely delicate jewels and
pivots, upon which the operation of each instru
ment depends. These delicate parts must be
carefully protected from mechanical shocks and
jars during shipment. Sensitive instruments will
not arrive in satisfactory operating condition un
less great care is taken in packing. Every effort
is made to handle and to repack these instruments
carefully at the Bureau, and whenever possible the
return shipment is made in the original container.
Before each instrument is packed, all binding
posts should be tightened, and any externally
operated clamping device for the moving system
should be switched to the "clamp" or "transit"
position. Plugs and other small accessories should
be enclosed in a small separate container tied to
the instrument. Glass windows of instruments
lacking protective cases should be protected by
pieces of thin wood or heavy cardboard before
wrapping. Each instrument should then be
wrapped in heavy manila paper or similar cover
ing and sealed with gummed tape to exclude dust
and excelsior.
Boxes in which instruments are packed should be
strong, preferably of wood, with screwed-on tops
to avoid damage to pivots or jewels, which may
be caused by a hammer or nail puller.
Clean, fresh excelsior or its equivalent in special
packaging material should be used as the shockabsorbing material. A layer of excelsior at least 3
to 4 inches deep, pressed down firmly, should
surround each wrapped instrument. Instruments
having pivoted components should be packed
upside down.
High-grade pivoted instruments of the labora
tory-standard type, which have comparatively
heavy moving systems without clamping devices,
should be packed with special care and should
always be individually shipped in wooden boxes
with 4 to 6 inches of excelsior around the wrapped

instrument. Portable standard watthour meters

(rotating standards) should also be individually
packed.
Certain heavy accessories used with instru
ments, such as ammeter shunts, current transformers, and voltage (potential) transformers,
should be packed in separate boxes to avoid
possible damage to the instruments. Heavy
pieces should always be shipped in wooden boxes
and held in place, if necessary, by checks or cleats
Large transformers, especially those having oilfilled iron cases, should ba crated singly, and
arranged whenever possible so that the terminals
can be made accessible for tests without removing
the entire crate.
The tops of boxes and crates must be marked
"This Side Up." Boxes containing delicate instru
ments should be marked "Fragile, Handle With
Care." Those containing any glass parts should
be marked "Glass." Failure to use such markings
precludes recourse in the event of loss or damagi
in shipping.
6.2. Standard Cells
A laboratory having reference-standard cells of
the saturated type would logically provide itself
with a group of about three saturated standard
cells which can be sent to the NBS annuaUv,
while the reference cells are new, but biennially
thereafter. Cells of this type must be kept
upright at all times and protected as far as possible
from shock and temperature changes. This
necessitates hand carrying and arrangements at
each end for installing them in a thermostat ted
bath [11, 12]. Smaller laboratories having un
saturated reference standards may use 2 or 3
unsaturated cadmium standard cells as inter
laboratory standards. These can be shipped by
parcel post. Shipments in extremely cold or hot
weather should be avoided. If each interlabora
tory cell is compared with the cells of the labora
tory reference group before and after their trans
port to NBS, a very desirable check is obtained
on any changes that may have occurred during
transport. If any one cell shows the same value
relative to the local group, before and after its
travels, it is highly probable that its emf did* not
change and recover by an equal amount. If all
cells are unchanged, the probability of the com
parison being valid is greatly increased.
6.3. Resistors, Capacitors, and Inductors
Fixed standard resistors, capacitors, and in
ductors, whether of the fixed-unit type or groups
of these combined in dial-type boxes, are satis
factory as interlaboratory standards and can be
used at intervals of 1 to 2 years for comparison
with NBS, depending on the stability of the
reference standards that they serve. Intercomparison between the laboratories of a single
organization at intervals of 1 year may prove
useful in cases where the volume of testing at the
individual laboratories is large and the working
standards at these laboratories are therefore un

42-8

usually liable to deterioration or accidental

damage. The values of the interlaboratory stand
ards of this group should be such as to cover the
range of measurements with which the laboratory
is concerned.
6.4. Instrument Transformers
Standard multirange current and voltage trans
formers can be obtained which are of quite rugged
construction and give reliable performance for
long periods of time. For a small laboratory, a
single set of such transformers covering the com
plete range may be tested initially at NBS and
resubmitted at intervals of 5 or even 10 years for
verification. A larger laboratory in which the
program of transformer testing cannot be inter
rupted will need a duplicate set of standardinstrument transformers, one set being used as
interlaboratory standards at intervals of 5 years,
while it and the other set both serve as working
standards the rest of the time.
6.5. Indicating Instruments
The use of indicating instruments as inter
laboratory standards is often of great value as an
over-all check on the comparison equipment and
on personnel and procedures. On the other hand,
in general, the actual transport of the units of
measurement from a higher to a lower echelon is
done more accurately by standard cells and re
sistors. The transfer of the electrical units from
d-c to a-c standards is based on the initial tests
at NBS [23] of suitable 0.1-percent or 0.25-percent
wattmeters, ammeters, and voltmeters. It may
be found desirable, as a guard against accidental
changes, to verify the performance of the transfer
standards by comparing them with similar inter
laboratory standards at 5-year intervals. A group
of multirange a-c instruments may be used as
interlaboratory standards to check the over-all
accuracy of the a-c working standards built per
manently into some types of consoles.
7. Summary
In the foregoing sections, some of the basic
principles on which the operations of an electrical
standardizing laboratory should be based have
been listed; the types of standard equipment
needed have been classified; and the intervals at
which these pieces of equipment should be intercompared locally and checked by comparison with
a laboratory of higher echelon.have been suggested.
The most important considerations in such an
enterprise are:
(1) The leader must have a high degree of
technical knowledge and competence in the spe
cialized field of electrical measurements ;
(2) The measuring apparatus must be adequate
and chosen specifically to fit the kinds of measure
ment and level of accuracy demanded ;
(3) The checking procedures must be definite
and followed carefully, but should be flexible
enough to meet emergencies;

(4) The laboratory must accept responsibility

for the internal consistency of its measurements,
and should look to a higher echelon (such as
NBS) only for its initial calibration and for
periodic checks to detect drifts in the values of its
reference standards.
8. References
[I] The meter laboratory, Electrical metermen's handbook,
6th ed., chap. 12 (Edison Electric Inst., New York,
N. Y., 1960).
[2] American standard code for electricity meters, Am.
Standards Assoc. C12-1941 (Edison Electric Inst.,
New York, N. Y.)
[3] H. N. Hayward, The calibration, checking and testing
of electrical instruments in an education laboratory,
Eighth Natl. Conf. Instr. Soc. Amer., paper 53-2-2
(1953).
[4] F. B. Silsbee, Establishment and maintenance of the
electrical units, NBS Circ. 475 (1949).
[5] F. B. Silsbee, Extension and dissemination of the
electrical and magnetic units, NBS Circ. 531 (1952).
[6] F. L. Hermach, The testing of electrical instruments,
Eighth Natl. Conf. Instr. Soc. Amer., paper 53-2-3
(1953).
[7] F. D. Weaver, Notes on the care and use of electrical
instruments, Instruments 23, 1236-1239 (1950).
[8] F. K. Harris, Electrical measurements (John Wiley &
Sons, Inc., New York, N. Y., 1952).
[9] F. A. Laws, Electrical measurements (McGraw-Hill
Book Co., Inc., New York, N. Y., 1938).
[10] J. B. Dowden, Organizing an electrical instrument
standardizing laboratory, Weston Engineering
Notes 2, 6 (June 1947); 4, 3 (Feb. 1949).
[II] E. F. Mueller and H. F. Stimson, A temperaturecontrol box for saturated standard cells, J. Research
NBS 13, 699 (1934) RP739.
[12] A. W. Spinks and F. L. Hermach, Portable poten
tiometer and thermostatted container for standard
cells. Rev. Sci. Instr. 26, 770 (1955).
[13] J. H. Park, Effect of service temperature conditions
on the electromotive force of unsaturated portable
standard cells, BS J. Research 10, 89 (1933)
RP518.
[14] G. W. Vinal, Primary batteries, chap. 6, Standard
cells (John Wiley & Sons, Inc., New York, N. Y.,
1950).
[15] J. L. Thomas, Precision resistors and their measure
ment, NBS Circ. 470 (1948).
[16] F. B. Silsbee and F. J. Gross, Testing and performance
of volt boxes, J. Research NBS 27, 269 (1941)
RP1419.
[17] American standard, Electrical indicating instruments,
ASA C39. 1-1955 (American Standards Association,
New York, N. Y.).
[18] F. L. Hermach and E. S. Williams, Multirange audio
frequency thermocouple instruments of high accu
racy, J. Research NBS 52, 227 (1954) RP2494.
[19] F. D. Weaver, An easily assembled console for rapid
testing of electrical indicating instruments, Instru
ments 22, 396-399 (1949).
[20] E. A. Gilbert, Equipment for instrument calibration,
Elec. Eng. 68, 1065 (1949).
[21] H. B. Brooks, The standard cell comparator, a spe
cialized potentiometer, BS J. Research 11, 211
(1933) RP586.
[22] J. H. Miller, Simplified standard cell comparator,
Weston Engineering Notes 9, 1 (Dec. 1954);
Trans. Am. Inst. Elec. Engrs. 73, (Pt. 1), 413 (1954).
[23] J. H. Park and A. B. Lewis, Standard electrodynamic
wattmeter and ac-dc transfer instrument, J. Re
search NBS 25, 545 (1940) RP1344.
[24] S. C. Richardson and F. A. Ludewig, A solution to
the problem of electrical instrument standardization
service, First Intern. Conf. Instr. Soc. Amer., paper
54-21-2 (1954).
Washington, May 10, 1956.

43-9

A STANDARD is a suitable device (physical

embodiment) for accurately maintaining a unit
of measurement and for transferring a unit
of resistance, capacitance, inductance, current or emf,
or a measurement of ratio, from one laboratory to
another.
A number of standards and measuring equipments,
such as resistors, saturated cells, capacitors, induc
tors, and bridges, must be obtained when setting
up facilities for a low-frequency calibration program.
An important consideration in the selection of such
equipment is the specification of characteristics that
one would like embodied in a good standard suitable
for transferring a unit from one laboratory to an
other.
Some of the characteristics are:
1. Ruggedness
2. Simplicity in construction and use
3. Stability (short term and long term repeatabil
ity)
4. Accuracy ( close to nominal value )
5. Portability
6. Insensitiveness to changes in environmental con
ditions, temperature, pressure, humidity, sun
light, vibration, position, capacitance, magnetic
material, etc.
7. Readability
8. Low voltage or current coefficient
9. Reasonable price.
There are likely no standards that fulfill all these
requirements. The Thomas-type resistor probably
comes most nearly to approaching the ideal. On the
other hand, the emf of the saturated cell is affected
by nearly all environmental conditions and, therefore,
the cells must be kept under carefully-controlled
ambient conditions.
A number of things can be done to improve the
quality of standards. Recently, manufacturers have
taken some steps in this direction. For example, one
manufacturer builds a capacitor enclosed in a tem
perature-controlled air bath. Another places the
resistors in his VVheatstone bridge in a temperaturecontrolled oil bath, and in so doing raises his quoted
accuracy by a factor of two. Other manufacturers
build resistors of improved resistance material hav
ing low temperature coefficients, long-term stability,
and actual values within a few ppm of nominal.
The accuracy to which a standard may be cali
brated depends on the quality of the standard. There
is no point in attempting to calibrate a resistor hav
ing an accuracy rating of 0.12 to an accuracy of
0.00012 because a measurement 10 minutes later
may show an entirely different value. In order to
improve accuracy, laboratories must be provided with
better standardsfor example, resistors having an ac
curacy rating of 0.01% rather than an accuracy rat
ing of 0.05?. In addition, they must have more
accurate bridges for comparison and must be pro
vided with better-trained measurement personnel.
There is no object in hiring an inexperienced and
untrained person for a laboratory and placing a
Thomas-type resistor and a Wenner Bridge at his
disposal. Such ;i person would not know what to
do with either piece ol equipment. Yet, with the
same type of resistor and bridge in the hands of

FRANK D. WEAVER
Low Frequency Calibration Services
National Bureau of Standards
Boulder Laboratories

ACHIEVEMENT

OF

MEASUREMENT

AGREEMENT

AMONG

ELECTRICAL

STANDARDS

LABORATORIES

Trained measurement personnel with a de

termination and a planned program, good
equipment, and good laboratories can achieve
near measurement agreement with the National
Bureau of Standards over a period of years. De
tails are given whereby the NBS Electronic Cali
bration Center, Boulder, Colorado, established
and maintains the low-frequency units and
achieves measurement agreement with NBS,
Washington. Other laboratories can follow this
precedent.
44-1

trained personnel in a good laboratory, comparison

measurements to one part in ten million can be ob
tained.
Traceability Concept
Efforts to improve the accuracy of standards and
measurements have lead to the concept of traceability. This concept implies knowledge of the value
of a standard through either direct or indirect meas
urements, with reference to the unit maintained as
a National Standard by the Bureau. It should also
imply the capability of making accurate measure
ments. As a result of the misuse of the idea of
traceability, NBS is called upon to calibrate and
issue Certificates on an unnecessarily large number
of standards. In many instances it is believed that
the corrections reported on the Certificate are not
applied by the user to the nominal value of the
standard. The Certificate too often is hidden in a
file to be made available upon request to prove
traceability. Certificates issued on this basis result
in a waste of NBS time and effort, and a waste
of the user's money.
Methods of Achieving Equality
The procedure whereby the Electronic Calibra
tion Center (ECC) of the Badio Standards Labo
ratory of NBS, Boulder, achieved equality with NBS,
Washington, has an important lesson for all stand
ards labs.
The Electronic Calibration Center began opera
tion in late 1957. The establishment of this facility
grew out of the needs of the Department of Defense
and of industry for additional calibration service
covering the frequency spectrum from dc through
microwaves, at a convenient location. In the LowFrequency (dc-30kc) Unit of the Center, services are
provided for the calibration of electrical standards
and instruments such as standard cells, standard
resistors and resistance apparatus, standard inductors
and capacitors, volt boxes and other d-c ratio de
vices, standard resistors for current measurements,
electrical instruments, inductive voltage dividers, and
instrument transformers.
The Electronic Calibration Center started issuing
Certificates on low-frequency standards in 1958. Last
year about 3000 low-frequency standards were cali
brated and certificates issued. The measurements
on which Certificates are issued are determined from
intercomparison with high-quality working standards.
Periodically the working groups at both Boulder and
Washington are calibrated by comparison with the
National Beference Group. The main purpose of
the reference group of resistors and cells is for main
taining the units' of resistance and voltage.
The ohm is maintained at NBS Washington by
a group of Thomas-Type resistors. These are referred
to as national reference standards of resistance. The
volt is maintained by a number of saturated stand
ard cells. These are referred to as national reference
standards of voltage. The Center has, at the present
time, working standards of resistance and voltage
Abstracts from papt-r presented ill ISA Conference, I .(is
Angeles, September, 1961.
45- 2

NATIONAL REFERENCE STANDARDS

THOMAS TYPE I OHM RESISTORS
AND
SATURATED CELLS
(used to maintain the ohm and the volt)
NBS -WASHINGTON
I
WORKING STANDARDS

WORKING STANDARDS

Good quality
standard resistors
and
saturated cells

Good quality
standard resistors
and
saturated cells

ECC- BOULDER

NBS- WASHINGTON

FIG. I. RELATION of the calibration facilities at

NBS-Washington and ECC-Boulder.
that have been assigned values by comparison with
the national reference standards. Fig. 1 shows the
relation of the calibration facilities at the two labora
tories, for resistance and voltage, to the national
reference standards.
Laboratory standards sent to either location for
calibration are compared with working standards,
and are not ordinarily directly compared with the
national reference standards. The working standards
are of equal quality and reliability whether they are
across the hall from the reference standards in Wash
ington, D. C, or seventeen hundred miles away in
Boulder, Colorado.
In 1957 it was believed that within a few years
it would be possible to establish the ohm and the
volt at the Electronic Calibration Center with the
highest precision and that the units would, if a care
fully planned inter-comparison program was followed,
approach closely the value of the volt and the ohm
maintained at Washington. There were no plans for
establishing the units at the Center by the long
ard tedious process involved in the absolute determi
nation of the volt and the ohm. Instead, by using the
best of standards and techniques with many intercomparisons over a period, possibly five years, the
same goal could be achieved. Evidence indicates
that this objective has been approached closely, and
that practically identical calibration accuracies have
been achieved at the two locations.
The laboratory wishing to obtain the best pos
sible calibrations on its electrical standards, resistors,
capacitors, inductors, bridges, etc., may have these
standards calibrated in either Washington or Boulder
on the basis of convenience.
Resistance Standards
Several methods are used to be sure that the
measurement accuracies in the two laboratories,
Boulder and Washington, are equivalent. The present
working group for resistance measurements includes
five Thomas-typ<' resistors. Each year three of these,
after being measured in Boulder, are hand-carried
to NBS Washington, for comparison with the nation

al reference standards. To date there lias been little

change in the value of the Boulder Thomas-type
resistors. As indicated in Table 1, the maximum
change over four years in any one resistor is 1.2
ppm, and the minimum about 0.4 ppm. In the most
recent intercomparison ( April 1961 ) two of the three
resistors showed no change from a year ago in parts
in ten million; the other showed a change of 1 part
in ten million.
TABLE IECC THOMAS-TYPE RESISTORS
Resistor
No.
Corrections (ppm)
Fobr
June
April
April
April
I9S7
1958
1959
I 960
1961
149
+3.1
+2.8*
+2.4*
+1.9*
+2.3*
148
+8.1
+7.7
+7.3
+7.3
+7.3
147
+7.1
+7.0
+6.6
+6.7
+6.8
146
+5.1
+5.2*
+4.8*
+4.7*
+5.0*
145
+2.1
+2.4*
+2.1
+2.1
+2.1
"Indicates Boulder calibrations

are made with other resistors in an effort to re

solve the disagreement. Sometimes values for past
measurements are useful in resolving such differences.
This procedure continues up through the 100,000ohm value and down to the 0.0001-ohm resistors.
The records show that the expected differences
between the Washington values and the ECC values
at the 1-ohm level should be less than 1 ppm. At
nominal values other than one ohm, the uncertainty
may amount to N ppm; where N is the number
of decades removed from the one-ohm level. At a
value of 100,000 ohms we would be concerned if
the difference exceeded about 4 ppm.
The Center frequently receives resistors from other
laboratories that were previously calibrated in Wash
ington. The Washington data on resistors have been
made available to ECC personnel. This has provided
an opportunity to make a large number of intercomparisons with good quality resistors over the past
three years. Table 3 indicates close agreement. Large
differences probably indicate a change in the re-

TABLE 2COMPARISON OF WASHINGTON

TABLE 3CORRECTIONS FOR SOME RESISTORS
AND BOULDER CALIBRATIONS
CALIBRATED IN WASHINGTON AND
(Step-Up and Step-Down from the Thomas l-Ohm Resistors).
CALIBRATED IN BOULDER APPROXIMATELY
Corrections (ppm)
Nominel
A YEAR LATER.
Washington
Boulder
Resistance
Nominal
April 1961
June 1961
ohms
Date of Calibration
Corrections (ppm)
Resistance
Washington
Washington
Boulder
Boulder
(ohms)
- 10
- 7
100000
1 00000
+ I*
+ 14
+ 27
Mar 61
+ 33
Jan 60
100000
- 21
- 22
10000
-24
Jan 60
Mar 61
-25
10000
- 24
10000
- 25
-29
-28
Jan 58
Sep 59
10000
- 25
- 26
1000
Jul 57
Jul 58
+ 52
+ 54
+ 40
10000
1000
+ 40
- 6
100
- 6
+ 22
+ 23
1000
Jun 59
Jul 60
- 29
- 29
100
- 2
+ 1
Sep 59
Jan 58
1000
- 7
- 7
10
Jul 58
+ 12
+ 12
Jul 57
1000
- 3
- 3
10
15
Jun 60
10
Nov 59
- 9
- 10
100
0.1
- 24
0.1
- 26
+ 9
Nov 59
Nov 60
+ 22
100
+
0.01
+ 5
+ 6
Jan 58
Sep 59
100
+ *
+ M
0.01
+ 10
Jul 58
15
17
Jul 57
100
0.001
+ II
+ II
-89
Sep 60
-91
Sep 59
10
0.001
0
0
0.0001
+ 145
+ 145
+ 4
+ 2
Jan 58
Sep 59
10
0.0001
+
+ 1
Jul 58
74
77
Jul 57
10
-58
-59
1
Apr 58
Jul 59
In addition to the Thomas-type resistors, the Center
1
64
67
Apr 59
Feb 58
also sends yearly to Washington two of each other
-80
!
Sep 60
80
Sep 59
nominal values from 0.0001 ohm to 100,000 ohms.
Nov 59
+ 24.9
+ 26
Oct 58
I*
This enables ECC personnel to check all working
+ 22
+ 23
Oct 58
Jan 60
1*
standards by two methods with the Wenner bridge,
3
3.4
Aug 60
I*
Jun
59
by (1) the direct comparison method and by (2)
- 5.3
- 5
I*
Jun 59
Sep 60
step-up or step-down procedures from the one ohm.
Nov. 59
+ 13.0
+ 12.1
I*
Mar 58
Table 2 shows close agreement in this step-up and
+ 367
+ 359
Jul 57
Aug 58
0.1
step-down procedure.
- 33
Apr 58
- 18
0.1
Nov 57
When the resistors are returned to the Center
- 12
Mar 59
6
Mar 58
0.1
- 18
24
Jul 59
Jul 58
0.01
after intercomparison in NBS Washington, the new
2
- 2
Feb 58
Mar 59
0.01
values of resistance for the three Thomas-type re
+ 550
+ 549
Jul 58
041
Jul 57
sistors are used to calibrate the other one-ohm re
+ 507
0.001
Jul 58
+ 508
May 57
sistors by direct comparison. Two of the Thomas-149
Jul
60
-137
0.0001
Dec
59
type resistors then are used to step-up to the two
"Denotes Thomas-type 1-ohm resistors.
10-ohm resistors for which recent Washington values
sistor itself rather than imprecision in the measure
are available. If the values (Washington value and
ECC step-up value) for these 10-ohm resistors agree
ments. The table indicates, in general, that the re
sistance of a good resistor changes very little (a
within 1 ppm, then they are used to calibrate the
other 10-ohm resistors and also to step-up to the two
few ppm) from one year to another. With all this
recently calibrated 100-ohm resistors.
information, considerable confidence has been built
up in the ECC ohm to the point where it is be
If differences between the Washington value and
lieved to be within a lew parts per ten million oi
tin- Boulder step-up value exceed about 1 to 2 ppm,
depending on nominal value, additional comparisons
the ohm maintained in Washington.
46 3

Voltage Standards
To establish the volt in Boulder the low-frequency
unit started with eighteen saturated cells, three
groups of six each, with each group in an air bath.
It was evident in 1958 that the Center would be
called upon to calibrate saturated cells. This meant
increased accuracy and the need for placing ECC
cells in an oil bath for better temperature control.
Two oil baths were designed and constructed; one
to regulate at 28 C and the other to regulate at
35 C. Of the original eighteen cells only nine turned
out to be stable enough to use as a working group.
Later, additional cells were obtained. The Center
has a group of six that now appear stable enough to
add to the working group, making a total of fifteen.
This group is never moved. Six cells are used as
a traveling group and hand-carried to Washington
and returned for intercomparison purposes. It now
appears that the volt, as maintained at the ECC,
is within about 0.6 microvolt of the volt maintained
at NBS Washington, as shown in Table 4. This
gap will be closed to possibly 0.2 microvolt with a
little more history on ECC cells. An additional period
of a year or so may be required to achieve this
goal.
TABLE 4COMPARISON OF CALIBRATIONS
ON SIX SATURATED CELLS
Difference
Washington
Boulder
Position
Number
Value
Value
Mar 1961
Jan 1961
microvol1
(volts)
(volts)
+ 0.9
1.0182259
1.0182250
I
1.0182229
+0.7
1.0182222
2
1.0182494
1.0182492
+0.2
]
1.0 182501
1.0182507
+ 0.6
4
1.0182462
1.0182467
+ 0.5
S
1.0182485
+ 0.7
1.0182478

1.0 182401
1.0182407
Mean
+0.6
In helping to establish a volt, some laboratories
have been thoughtful enough to have their cells
delivered to the Center after calibration in Washing
ton and prior to their return to their own laboratory.
This enables additional intercomparisons to be made.
Volt boxes have been used as an intercomparison
device by the Electronic Calibration Center and
Washington to assure measurement agreement. Table
5 shows the results of a number of intercomparisons
on a volt box.
TABLE 5RATIO COMPARISONS
(CORRECTIONS, PPM) FOR A VOLT BOX.
Washington
Voltage
Boulder
Washington
Boulder
Range
Aug 58
Aug 58
Jun 61
Jul 60
*
*
3
0
+
+
+ *
+ 1?
+ 25
+ 21
7.5
+ 26
15
+ 34
+ 38
+ 32
+ 38
+ 20
+ 24
30
+ 23
+ 32
- 85
75
79
78
74
-120
112
150
1 14
118
175
-152
300
-158
161
- 96
750
- 86
- 82
- 86
The data, in ppm, show close measurement agree
ment in d-v ratio as determined at NBS Washington
and Boulder.
47-

Capacitance and Inductance

In the area of capacitance and inductance the
Center has three standards of each nominal value
that manufacturers ordinarily make. One standard
of each of the nominal values is sent to Washington
each year for intercomparison purposes.
Other interlaboratory standards such as inductive
voltage dividers, current and potential transformers,
standard resistors for current measurement, volt
meters and a-c transfer standards are periodically
shipped between the two laboratories and calibrated
for intercomparison purposes. In addition to giving
information about the stability of standards that
have been transported bv the usual me ins of public
transportation, repeatability of measured values on
these items gives us additional assurance of measure
ment agreement.
Conclusion
With the svstem of intercomparison described above
it appears that close measurement agreement has
been achieved between the two laboratories. The
achievement of the Electronic Calibration Center in
this area can be closely approached by another labo
ratory. Some laboratories already have long his
tory of comparisons on their standards by NBS. It
woidd seem that another laboratory, in a period of
a few years, could establish a volt and ohm that
closely approaches the value of the volt and ohm
maintained at NBS.
Another standards laboratory with well-trained
personnel having a determination to make the best
measurements and working with high quality stand
ards and facilities can achieve near measurement
equality with the National Bureau of Standards. To
do this their Thomas-type resistors should have annual
comparisons with the national working standards, and
their saturated cells should be intercompared by a
traveling group annually with the national working
standard cells. The time can be foreseen when a
number of laboratories will reach this objective.
What Is the Best Calibration Interval?
Some items may require NBS certification once
per year; and others at less-frequent intervals. The
calibration by NBS each year of every standard in a
laboratory generally is not necessary. The calibra
tion interval for a given type of standard depends
upon a number of factors, but experience indicates
the recommended period for some items to be as
much as 5 years. Furthermore, NBS should not be
called upon to calibrate standards that the labora
tory can do with sufficient accuracy. With only a
few selected calibrated standards as a nucleus, it is
sometimes possible to calibrate a wide range of other
standards by step-up and intercomparison techniques.
It may be advisable to consider the use of the
service of a commercial organization. A number of
commercial testing laboratories have adequate facili
ties for calibrating many of the standards associated
with the usual standards laboratory.
Minimum requirements should be established by
some national organization to guide laboratories seek
ing to establish traceability.
( Pre-Printrd from Instruments & Control Systems, 1%2I

\ugust
1966

U. S. DEPARTMENT OF COMMERCE
NATIONAL BUREAU OF STANDARDS

List of
Publications
LP 38

Washington, D. C.

ELECTRICAL UNITS , INSTRUMENTS , AND MEASUREMENTS

Publications by the Members of the Staff of
The National Bureau of Standards

CONTENTS
PAGE
1.

Introduction

2.

Library of Congress Photoduplication Service

3.

General Information:

3.1 Periodicals of the National Bureau of Standards

3.2 Catalogs of National Bureau of Standards Publications
3.3 Field Offices of the U. S. Department of Commerce

1
2
3, 4

....

4.

Units and Absolute Electrical Measurements

5.

Standard Cells

6.

Resistors

7.

Inductors

8.

Capacitors

9.

Potentiometers, Volt Boxes, and Bridges

10.

Instruments, Meters, and AC-DC Transfer Standards

11.

Transformers and Inductive Voltage Dividers

11

12.

Galvanometers

12

13.

High Voltage and Surge Measurements

12

14.

Dielectric Measurements

13

15.

Magnetic Measurements

X}

16.

Miscellaneous

14

17.

Reference to Books and Standards on Electrical Instruments and

Measurements

14

17.1
17.2

15
16

18.

Textbooks
Standards and Handbooks

Addendum

17

48-1

1.

Introduction

This list of publications is a selec

tion which describes the methods and equip
ment used in the establishment and mainte
nance of the electrical units and the stan
dards that have been developed at the Nation
al Bureau of Standards for the calibration
of measuring apparatus. It includes refer
ences to older publications of the Bureau
which describe basic principles and methods
and are therefore still useful. References
to articles on electrical measurements at
radio and higher frequencies are not in
cluded.

tions for research use. National Bureau of

Standards publications are on file at the
Library, so that a copy of any Bureau docu
ment that is out of print can usually be
obtained. Full information concerning this
service may be secured by writing to the
Library of Congress at the address noted
above. In making such inquiry, it is im
portant to give an accurate and complete
identification (author, title, place of
publication, name of the series and number,
if known) of the document desired.
3.

Many inquiries sent to this Bureau can

best be answered by reference to a recogni
zed national standard, or by reference to a
particular textbook or handbook. According
ly a few such standards and books are list
ed that include the information on instru
ments and measurements most frequently re
quested.

3.1

Periodicals of the National

of Standards

Bureau

Journal of Research

This Bureau makes no tests on motors,

generators, or transformers used for power
or lighting service, and has no current
publications on their design or performance.
It cannot undertake to answer questions con
cerning the design, construction, repair, or
rewinding of such apparatus. A number of
pertinent references are listed in a folder,
"Electric Motors and Generators, Basic In
formation Sources," available on request from
the Office of Publications, Department of
Commerce, Washington, D. C. 20230.
Some of the publications by Bureau
authors have appeared in the regular series
of publications of this Bureau, and others
in various scientific and technical journals.
Reprints of some of the more recent papers
may be available upon request from the au
thors. Where the price is stated Govern
ment publications can be purchased from the
Superintendent of Documents, U. S. Govern
ment Printing Office, Washington, D. C.,
20402. Remittances should be made either
by coupons (obtainable from the Superinten
dent of Documents in sets of 20 for $1.00
and good until used), or by check or money
order payable to the "Superintendent of
Documents, U. S. Government Printing Office"
and sent with the order. Publications may
be consulted in leading libraries.
2.

General Information

Library of Congress Photoduplication

Service

The Photoduplication Service, Library

of Congress, Washington, D. C. 20540, makes
photoduplicates of material in its collec
49-1

The Journal of Research reports Nation

al Bureau of Standards research and develop
ment in physics, mathematics, chemistry, and
engineering. Comprehensive scientific papers
give complete details of the work, including
laboratory data, experimental procedures,
and theoretical and mathematical analyses.
Illustrated with photographs, drawings,
and charts.
All NBS nonperiodical publications and
articles by the Bureau staff in professional
journals will be abstracted in the appropri
ate section of the Journal. In addition,
each section will carry a complete listing
of all Bureau publications that are not
abstracted in that section.
Three sections
A.

PHYSICS AND CHEMISTRY

Papers of interest primarily to scien

tists working in these fields. This section
will cover a broad range of physical and
chemical research, with major emphasis on
standards of physical measurement, funda
mental constants, and properties of matter.
Issued six times a year. Annual subscrip
tion: Domestic, $5.00; Foreign, $6.00
B.

MATHEMATICAL

SCIENCES

Studies and compilations designed main

ly for the mathematician and theoretical
physicist. Topics in mathematical statis
tics, theory of experiment design, numeri
cal analysis, theoretical physics and chem
istry, logical design and programming of
computers and computer systems. Short nu
merical tables. Issued quarterly. Annual

subscription:
$3.50.

Domestic, $2.75; Foreign,

cations issued after January 1, 194-2 are

included. Copies of these catalogs may be
obtained from the Superintendent of Docu
ments, U. S. Government Printing Office,
Washington, D. C. 20402, as follows:

C. ENGINEERING AND INSTRUMENTATION

Reporting results of interest chiefly
to the engineer and the applied scientist.
This section will include many of the new
developments in instrumentation resulting
from the Bureau's work in physical measure
ment, data processing, and development of
test methods. It will also cover some of
the work in acoustics, applied mechanics,
building research, and cryogenic engineer
ing. Issued quarterly. Annual subscrip
tion: Domestic, $2.75; Foreign, $3.50.

Circular 46O, publications issued

1901 through June 30, 1947,
(375 pages), ($1.25).
Supplement I to Circular 460, publi
cations issued July 1, 1947 through
June 30, 1957 (373 pages), ($1.50).
Miscellaneous Publication 240, publi
cations issued July 1, 1957 through
June 30, I960, (391 pages), ($2.25).

Technical News Bulletin

Summaries of current research at the
National Bureau of Standards are published
each month in the Technical News Bulletin.
The articles are brief, with emphasis on
the results of research, chosen on the basis
of their scientific or technologic impor
tance. Lists of all Bureau publications
during the preceding month are given, in
cluding Research Papers, Handbooks, Applied
Mathematics Series, Building Materials and
Structures Reports, Miscellaneous Publi
cations, and Circulars. Each issue con
tains 12 or more two-column pages; illus
trated. Annual subscription: Domestic,
$1.50; Foreign, $2.25.
All NBS periodicals issued by the
Bureau may be obtained by purchase on an
annual subscription basis from the Super
intendent of Documents, U. S. Government
Printing Office, Washington, D. C. 20402.
3.2

Catalogs of National Bureau of

Standards Publications

Supplement to Miscellaneous Publi

cation 240, publications issued
July 1, 1960 through February 26,
1965.
3.3

Field Offices of the U. S. Department

of Commerce

Department of Commerce Field Offices

are maintained in the cities listed be
low. Their purpose is to provide ready
access, at the local level, to the Depart
ment's reports, publications, statistical
statements, surveys, as well as to the
specialized and experienced staff in charge.
Each Field Office serves as an official
sales agent of the Superintendent of
Documents, U. S. Government Printing Office,
making available for purchase locally a
wide range of Government publications. The
reference library maintained by each Field
Office contains many Government and private
publications, periodicals, directories,
reports, and other reference material.

For additional information concern

ing available publications consult the
Field Office in your area.

National Bureau of Standards Circular

460 and its supplement list all Bureau
publications issued from 1901 through June
30, 1957. Brief abstracts for the publi

50-2

LIST OF FIELD OFFICES

Cincinnati, Ohio 45202
8028 Federal Office Building
550 Main Street
Robert M. Luckey, Director
Area Code 513 Tel. 684-2944

Albuquerque, New Mexico 87101

U. S. Courthouse
William E. Dwyer, Director
Area Code 505 Tel. 247-0311
Anchorage, Alaska 99501
306 Loussac-Sogn Building
Clyde S. Courtnage, Director
Area Code 907 Tel. 272-6331

Cleveland, Ohio 44101

4th Floor, Federal Reserve Bank Building
East 6th Street and Superior Avenue
Charles B. Stebbins, Director
Area Code 216 Tel. 241-7900

Atlanta, Georgia 30303

4-th Floor, Home Savings Building
75 Forsyth Street, N. W.
Daniel M. Paul, Director
Area Code 404 Tel. 526-6000

Dallas, Texas 75202

Room 1200, 1114 Commerce Street
Harry C. Meyers, Director
Area Code 214 Tel. Riverside 9-3287

Baltimore, Maryland 21202

305 U. S. Customhouse
Gay and Lombard Streets
Carroll F. Hopkins, Director
Area Code 301 Tel. Plaza 2-8460

Denver, Colorado 80202

16407 Federal Building
20th and Stout Streets
Charles E. Brokaw, Director
Area Code 303 Tel. 297-3246

Birmingham, Alabama 35203

505 Title Building, 2030 Third Avenue, North
Gayle C. Shelton, Jr., Director
Area Code 205 Tel. 325-3131
Boston, Massachusetts 02110
Room 230, 80 Federal Street
Paul G. Carney, Director
Area Code 617 Tel. CA 3-2312

Des Moines, Iowa 50309

1216 Paramount Building
509 Grand Avenue
Raymond E. Eveland, Director
Area Code 515 Tel. 284-4222
Detroit, Michigan 48226
445 Federal Building
Frank A. Alter, Director
Area Code 313 Tel. 226-6088

Buffalo, New York 14203

504 Federal Building
117 Ellicott Street
Robert F. Magee, Director
Area Code 716 Tel. 842-3208

Greensboro, North Carolina 27402

412 U. S. Post Office Building
Joel B. New, Director
Area Code 919 Tel. 275-9111

Charleston, South Carolina 29403

Federal Building - Suite 631
334 Meeting Street
Paul Quattlebaum, Jr., Director
Area Code 803 Tel. 747-4171

Hartford, Connecticut 06103

18 Asylum Street
James E. Kelley, Director
Area Code 203 Tel. 244-3530

Charleston, West Virginia 25301

3002 New Federal Office Building
500 Quarrier Street
J. Raymond DePaulo, Director
Area Code 304 Tel. 343-6196

Honolulu, Hawaii 96813

202 International Savings Building
1022 Bethel Street
H. Tucker Gratz, Director
Tel. 588977

Cheyenne, Wyoming 82001

6022 Federal Building
2120 Capitol Avenue
Albert B. Kahn, Director
Area Code 307 Tel. 634-5920

Houston, Texas 77002

5102 Federal Building
515 Rusk Avenue
Edward T. Fecteau, Jr. , Director
Area Code 713 Tel. 228-0611

Chicago, Illinois 60604

i486 New Federal Building
219 South Dearborn Street
Anthony J. Buchar, Director
Area Code 312 Tel. 828-4400

Jacksonville, Florida 32202

512 Greenleaf Building
208 Laura Street
William B. Curry, Director
Area Code 305 Tel. 354-7111
51-3

Kansas City, Missouri 64IO6

Room 2011, 911 Walnut Street
Nathan L. Stein, Director
Area Code 816 Tel. BA 1-7000

LIST OF FIELD OFFICES - Cont.

Pittsburgh, Pennsylvania 15222
2201 Federal Building
1000 Liberty Avenue
John A. Donely, Director
Area Code 412 Tel. 644-2850

Los Angeles, California 90015

Room 450, Western Pacific Building
1031 South Broadway
Stanley K. Crook, Director
Area Code 213 Tel. 688-2833

Portland, Oregon 97204

217 Old U. S. Courthouse
520 S. W. Morrison Street
James W. Goodsell, Director
Area Code 503 Tel. 226-3361

Memphis, Tennessee 38103

345 Federal Office Building
167 North Main Street
John M. Fowler, Director
Area Code 901 Tel. 534-3214

Reno, Nevada 89502

2028 Federal Building
300 Booth Street
Jack M. Howell, Director
Area Code 702 Tel. 784-5203

Miami, Florida 33130

928 Federal Office Building
51 S. W. First Avenue
Marion A. Leonard, Director
Area Code 305 Tel. 350-5267

Richmond, Virginia 23240

2105 Federal Building
400 North 8th Street
William S. Parker, Director
Area Code 703 Tel. 649-3611

Milwaukee, Wisconsin 53203

Straus Building
238 West Wisconsin Avenue
David F. Howe, Director
Area Code 4H Tel. BR 2-8600

St. Louis, Missouri 63103

2511 Federal Building
1520 Market Street
Alfred L. Rascher, Jr. , Director
Area Code 314 Tel. MA 2-4243

Minneapolis, Minnesota 55401

306 Federal Building
110 South Fourth Street
Ernest G. Booth, Director
Area Code 612 Tel. 334-2133

Salt Lake City, Utah 84IH

3235 Federal Building
125 South State Street
Stephen P. Smoot, Director
Area Code 801 Tel. 524-5116

New Orleans, Louisiana 70130

909 Federal Office Building, South
610 South Street
Edwin A. Leland, Jr., Director
Area Code 504 Tel. 527-6546

San Francisco, California 94102

Federal Building, Box 36013
450 Golden Gate Avenue
Philip M. Creighton, Director
Area Code 415 Tel. 556-5864

New York, New York 10001

6lst Floor, Empire State Building
350 Fifth Avenue
Arthur C. Rutzen, Director
Area Code 212 Tel. Longacre 3-3377

Santurce, Puerto Rico 0097

Room 628, 605 Condado Avenue
John H. Shoaf, Director
Phone: 723-4640

Philadelphia, Pennsylvania 19107

Jefferson Building
1015 Chestnut Street
David Jamieson, Director
Area Code 215 Tel. 597-2850

Savannah, Georgia 31402

235 U. S. Courthouse and
Post Office Building
125-29 Bull Street
Joseph G. Stovall, Director
Area Code 912 Tel. 232-4321

Phoenix, Arizona 85025

5413 New Federal Building
230 North First Avenue
Donald W. Fry, Director
Area Code 602 Tel. 261-3285

Seattle, Washington 98104

809 Federal Office Building
909 First Avenue
William H. Flood, Director
Area Code 206 Tel. 583-5615
52-4

4.

Units and Absolute Electrical Measurements

1.

A Determination of the Absolute Ohm, Using an Improved Self Inductor, H. L. Curtis,

C. Moon, and C. M. Sparks, J. Research NBS 21, 375, (1938), RP 1137.

2.

An Absolute Determination of the Ampere, Using Helical and Spiral Coils, R. W. Curtis,
R. L. Driscoll, and C. L. Critchfield, J. Research NBS 28, 133, (1942), RP 1449.

3.

Review of Recent Absolute Determinations of the Ohm and the Ampere, H. L. Curtis,
J. Research NBS 22, 235, (1944), RP 1606.

4.

An Absolute Measurement of Resistance by the Wenner Method, J. L. Thomas, C. Peterson,

I. L. Cooter, and F. R. Kotter, J. Research NBS 42, 291, (1949), RP 2029.

5.

Establishment and Maintenance of the Electrical Units, F. B. Silsbee, NBS Circular

475, 38 p., (1949), 35 cents.

6.

Measurement of the Proton Moment in Absolute Units, H. A. Thomas, R. L. Driscoll,

and J. A. Hippie, J. Research NBS 4A, 569, (1950), RP 2104.

7.

A New Method for Determining the Value of the Faraday, D. N. Craig and J. I. Hoffman,
Phys. Rev. 80, No. 3, 487, (1950).

8.

Standards for Electrical Measurement, F. B. Silsbee, Physics Today 4, 19, (1951).

9.

Measure for Measure: Some Problems and Paradoxes of Precision, F. B. Silsbee,

J. Washington Academy of Sciences 41_, No. 7, 213, (1951).

10.

Fundamental Units and Standards, F. B. Silsbee, Instruments 26, 1520 (1953).

11.

Extension and Dissemination of the Electrical and Magnetic Units by the National
Bureau of Standards, F. B. Silsbee, NBS Circular 531, 33 p., (1952).

12.

Electrochemical Constant, NBS Circular 524 (1953), $2.00

This circular includes:
a) Standard Cells and the Unit of Electromotive Force, W. J. Hamer and
L. H. Brickwedde, P. R. Robb, pp. 103 - 118.
b) Determination of the Faraday Constant by the Electrolytic Oxidation of Oxalate
Ions, D. N. Craig, and J. I. Hoffman, pp. 13 - 20.
c) The Faraday and the Omegatron, H. Sommer and J. A. Hiple, pp. 21 - 26.

13.

Nuclear Magnetic Resonance and the Measurement of Magnetic Fields, R. L. Driscoll,

Natl. Phys. Lab. Symp. of Precision Electrical Measurements, (1955).

14.

Measurement of Current with a Pellat-Type Electrodynamometer, R. L. Driscoll,

J. Research NBS 60, 287, (1958), RP 2845.

15.

Measurement of Current with the National Bureau of Standards Current Balance,

R. L. Driscoll and R. D. Cutkosky, J. Research NBS 60 297, (1958) RP 2846.

16.

New Apparatus at the National Bureau of Standards for Absolute Capacitance Measurement,
M. C. McGregor, J. F. Hersh, R. D. Cutkosky, F. K. Harris, and F. R. Kotter, IRE Trans.
Instrumentation, 1-7, No. 3-4, 253, (Dec. 1958).

17.

A Free Precession Determination of the Proton Gyromagnetic Ratio, P. L. Bender and

R. L. Driscoll, IRE Trans. Instrumentation, 1-7, No. 3-4, 176 (Dec. 1958).

18.

The Ampere, F. B. Silsbee, Proc. IRE 42, No. 5, 643, (May, 1959).

19.

Simplification of Systems of Units, F. B. Silsbee, AAAS Symposium, 1959.

53-5

20.

Some Results on the Cross Capacitances -Per Unit Length of Cylindrical Three -Terminal
Capacitors With Thin Dielectric Films on Their Electrodes, D. G. Lampard and R. D.
Cutkosky, Inst. Elec. Engs. (London) Mono. No. 351M, January I960.

21.

Determination of the Value of the Faraday With a Silver -Perchloric Acid Coulometer,
D. Norman Craig, C. A. Law, J. I. Hoffman, and W. J. Hamer, J. Research NBS 64A5.
381, (1960).

22.

Refining Measurements by Capacitance Techniques, F. K. Harris, R. D. Cutkosky,

ISA Journal, (1961).

23.

An Evaluation of the NBS Unit of Resistance Based on a Computable Capacitor, R. D.

Cutkosky, J. Research NBS 65A, No. 3, 97, (1961).

24.

The Systems of Electrical Units, F. B. Silsbee, J. Research NBS 66C. No. 2, 137, (1962).

25.

Four Terminal Pair Networks as Precision Admittance and Impedance Standards, R. D.

Cutkosky, IEEE Trans. Paper 63-928, (1963).

26.

A New Type of Computable Inductor, Chester H. Page, J. Research NBS 67B, No. 1, (1963).

27.

Definition of "Ampere" and 'Magnetic Constant", Chester H. Page, Proc. of the IEEE,
Vol. 53, No. 1, 1965.

28.

The Value of the Faraday, W. J. Hamer and D. N. Craig, J. Electrochem. Soc., III.
No. 12, (Dec. 1964).
5. Standard Cells
(see Section 9 also)

1.

Effect of Service Temperature Conditions on the Electromotive Force of Unsaturated

Portable Standard Cells, J. H. Park, J. Research NBS 10 89, (1933), RP 518.

2.

Standard Cells and the Unit of Electromotive Force, W. J. Hamer, L. H. Brickwedde,

and P. R. Robb, NBS Circular 524, p. 103 - 118, (1953).

3.

A Temperature -Control Box for Saturated Standard Cells, E. F. Mueller and

H. F. Stimson, J. Research NBS 13_ 699, (1934), RP 739.

4.

Standards of Electromotive Force, G. W. Vinal, D. N. Craig, and L. H. Brickwedde,

Trans. Electrochem. Soc. 68, 139, (1935).

5.

Metastability of Cadmium Sulfate and its Effect on Electromotive Force of Saturated

Standard Cells, G. W. Vinal and L. H. Brickwedde, J. Research NBS 26, 455 (1941),
RP 1389.

6.

Standard Cells and the Change From International to Absolute Electrical Units,
G. W. Vinal, J. Electrochem. Soc. 22, 95, (1948).

7.

New Quartz Container for Standard Cells at the National Bureau of Standards, G. W.
Vinal, L. H. Brickwedde, and W. J. Hamer, Compt. Rendu., Sept. 5 - 10, 92, (1949).

8.

Controlled Temperature Oil Baths for Saturated Standard Cells, Patrick H. Lowrie, Jr.,
NBS Tech. Note 141, (1962).

9.

Effect of Vibration and Shock on Unsaturated Standard Cells, R. J. Brodd, W. G. Eicke,

J. Research NBS 66C, No. 2, 85, (1962).

10.

Comments on Zener Diodes as Voltage Standards, W. G. Eicke, Jr., Comite Consultatif,

D'Electricite Aupres, Du "Comite Internationale Des Poids et Mesures," 10th
Session 'les, (1963).
54-6

11.

Making Precision Voltage Measurements on Zener Diodes, W. G. Eicke, Jr., IEEE Trans.
Paper CP 63-416, (1963).

12.

The Operating Characteristics of Zener Reference Diodes and Their Measurements,

W. G. Eicke, Jr., ISA Trans., 3_, No. 2 (April 1964).

13.

Standards of Electromotive Force, W. J. Hamer, J. Washington Academy of Sciences,

54, (Oct. 1964).

14.

Standard Cells, Their Construction, Maintenance, and Characteristics, W. J. Hamer, NBS

Mono. 84, (Jan. 1965), 35 cents.

15.

On the Long-Term Stability of Zener Reference Diodes, W. G. Eicke, Jr. and H. H. Ellis,
Proc. of the 11th Session, Comite Consultatif D'Electricite, (May 1965).

16.

The "Volt Standard" Moves to Gaithersburg, Maryland, W. J. Hamer, J. Washington

Academy of Sciences, 5.6, pp. 101 - 108, (1966).
6.

Resistors

1.

The Four-Terminal Conductor and the Thomson Bridge, F. Wenner, Bui. BS 8, 559,
(1912), S181.

2.

Adjustments of the Thomson Bridge in the Measurement of Very Low Resistances, F.

Wenner and E. Weibel, Bui. BS 11, 65, (1915) S22.

3.

A Study of the Inductance of Four-Terminal Resistance Standards, F. B. Silsbee,

Bui. BS 13_ 375, (1916-17), S281.

4.

Notes on the Design of Four-Terminal Resistance Standards for Alternating Currents,

F. B. Silsbee, BS J. Research
73, (1930), RP 133.

5.

A Method of Adjusting the Temperature Coefficient and Resistance of Low-Valued

Resistance Standards, F. Wenner and J. L. Thomas, J. Research 12, 147, (1934), RP 639.

6.

Methods, Apparatus, and Procedures for the Comparison of Precision Standard

Resistors, F. Wenner, J. Research NBS 2%, 229, (1940), RP 1323.

7.

Stability of Double -Walled Manganin Resistors, J. L. Thomas, J. Research NBS 26,

107, (1946), RP 1692.

8.

Precision Resistors and Their Measurement, J. L. Thomas, NBS Circular 470, 32 p.

(1948), 30 cents.

9.

Best Arrangement of Resistors in a Series Group, H. B. Brooks, Rev. Sci. Instr. 21,
No. 5, 491, (1950).

10.

Measurement of Multimegohm Resistors, A. H. Scott, J. Research NBS 50, 147,

(1953), RP 2402.

11.

Alloys For Precision Resistors. C. Peterson, Natl. Phys. Lab. Symp. of Precision
Electrical Measurements, (1955).

12.

Rack for Standard Resistors, P. H. Lowrie, Jr., Rev. Sci. Instr. 3_0_, 291 - 292,
(April 1959).

13.

A Method of Controlling the Effects of Resistance in the Link Circuit of the Thomson
or Kelvin Double Bridge, D. Ramaley, J. Research NBS 64 C4. 267, (I960).

14.

Calibration Procedures for D-C Resistance Apparatus, P. B. Brooks, NBS Mono. 39,
(1962), 40 cents.

315-411 O - 68 - 5

55-7

15.

Errors in the Series-Parallel Buildup of Four-Terminal Resistors, Chester H. Page,

J. of Research, Vol. 69C. No. 3, July-Sept. 1965.
7. Inductors
(See Section 4 also)

1.

Design of Standards of Inductance, and the Proposed Use of Models in the Design of
Air-Core and Iron-Core Reactors, H. B. Brooks, J. Research NBS 7, 289, (1931), RP 34-2.

2.

Improved Continuously Variable Self and Mutual Inductor, H. B. Brooks and A. B. Lewis,
J. Research NBS 19_, 493, (1937), RP 1040.

3.

Formulas for Computing Capacitance and Inductance, C. Snow, NBS Circular 544, 64 p.
(1954).

4.

A Study of Absolute Standards of Mutual Inductance and in Particular the Three-Section

National Bureau of Standards Type, F. W. Grover, J. Research NBS
-297, (1954),
RP 2548.

5.

Methods for Measuring the Q of Large Reactors, C. Peterson, B. L. Dunfee, and F. L.

Hermach, .Am. Inst. Elec. Engrs. Trans. Paper 56-87, (1956).

6.

Inductive Efficiency of Reactive Coils, H. B. Brooks, Elec. Engr. , (1956).

7.

Calibration of Inductance Standards in the Maxwell -Wien Bridge Circuit, T. L. Zapf,

J. Research NBS 6JC No. 3, 183, (1961).
8. Capacitors
(See Section 4 also)

1.

Measurement of Relative and True Power Factors of Air Capacitors, A. V. Astin, J.

Research NBS 21, 425, (1938), RP 1138.

2.

Nature of Energy Losses in Air Capacitors at Low Frequencies, A. V. Astin, J.

Research NBS 22, 673, (1939), RP 1212.

3.

Standards for Low Values of Direct Capacitance, C. Moon and C. M. Sparks, J. Research
NBS 41, 497, (1948), RP 1935.

4.

A Standard of Small Capacitance, C. Snow, J. Research NBS 2, 287, (1949), RP 1970.

5.

Formulas for Computing Capacitance and Inductance, C. Snow, NBS Circular 544,
69 p., (1954).

6.

Variable Capacitor Calibration with an Inductive Voltage Divider Bridge, T. L. Zapf,

NBS Tech. Note 57 (PB 161558) $1.60. (Available only by purchase from the Office of
Technical Services, Department of Commerce, Washington, D. C. 20230).

7.

Capacitor Calibration by Step-Up Methods, T. L. Zapf, J. Research NBS 64C. 75, (1960).

8.

New Apparatus at the National Bureau of Standards for Absolute Capacitance Measurement,
M. C. McGregor, J. F. Hersh, R. D. Cutkosky, F. K. Harris, and F. R. Kotter, IRE Trans.
Instrumentation, 1-7, No. 3-4, 253, (Dec. 1958).

9.

The Construction and Behavior of a Transportable Ten Picofarad Capacitor, R. D.

Cutkosky, L. H. Lee, Comite Consultatif, D'Electricity Aupres, Du "Comite Inter
nationale Des Poids et Mesures," 10th Session 'les (1963).
9.

1.

Potentiometers. Volt Boxes, and Bridges

Deflection Potentiometers for Current and Voltage Measurements, H. B. Brooks, Bui.

BS 8, 395, (1912), S172.
56-8

2.

A Multi -Range Potentiometer and its Application to the Measurement of Small Temperature
Differences, H. B. Brooks and A. W. Spinks, BS J. Research % 781, (1932), RP 506.

3.

The Standard-Cell Comparator, A Specialized Potentiometer, H. B. Brooks, J. Research

11, 211, (1933), RP 586.

4.

The Waidner-Wolff and Other Adjustable Electrical-Resistance Elements, E. F. Mueller

and F. Wenner, J. Research NBS 15, A77 , (1935) RP 842.

5.

Les Potentiometres, H. B. Brooks, Cong. Ind. d'Elec. (Paris), 3_, 275, (1932).

6.

Testing and Performance of Volt Boxes, F. B. Silsbee and F. J. Gross, J. Research

NBS 27, 269, (1941), RP 1419.

7.

Portable Potentiometer and Thermostatted Container for Standard Cells, A. W. Spinks

and F. L. Hermach, Rev. Sci. Instr. 26, No. 8, 770, (1955).

8.

Lindeck Potentiometer, D. W. Oliver, Rev. Sci. Instr. 26_, No. 11, 1078, (1955).

9.

An A-C Kelvin Bridge for the Audio Frequency Range, B. L. Dunfee, Am. Inst. Elec.
Engrs. Trans. Paper 56-25, (1956).

10.

The Use of an A-C Bridge to Measure Core Loss at High Inductions, I. L. Cooter and
W. P. Harris, Am. Inst. Elec. Engrs. Trans. Paper 56-26, (1956).

11.

A Low-Cost Microvolt Potentiometer, W. H. Wood, Rev. Sci. Instr. 28, No. 3, 202, (1957).

12.

Method for Calibrating a Standard Volt Box, Bernadine L. Dunfee, J. Research 67C. No. 1,
(1963).

13.

Practical Methods for Calibration of Potentiometers, D. Ramaley, NBS Tech. Note 172,
(1963), 30 cents.

14.

A Method for Comparing Two Nearly Equal Potentials Directly in Parts Per Million,
C. J. Saunders, Rev. Sci. Instr. 2A, (12), M52, (1963).

15.

A Method for Calibrating Volt Boxes, With an Analysis of Volt-Box Self-Heating

Characteristics, R. F. Dziuba and T. M. Souders, IEEE Conv. Record, March 1966.
Calibration Procedures
(Copies available at NBS. Resistance and Reactance Section)

1.

Notes on the Calibration of the Direct Reading Ratio Set.

2.

Calibration of the Wenner Low Range Potentiometer.

3.

Calibration of the Mueller Thermometer Bridge.

4.
5.

Calibration of the Universal Ratio Set.

The Six -Dial Thermofree Potentiometer.
10.

1.
2.
3.
4.

Instruments. Meters, and AC-DC Transfer Standards

Accuracy of Commercial Electrical Measurements, H. B. Brooks, Trans. Am. Inst.

Elec. Engrs. 22, 495, (1920).
A Suppressed -Zero Electrodynamic Voltmeter, F. K. Harris, J. Research NBS 3_, 445,
(1929), RP 105.
Composite-Coil Electrodynamic Instruments, F. B. Silsbee, J. Research NBS 8, 217,
(1932), RP 411.
Temperature Compensation of Mi Hivoltmeters, H. B. Brooks, J. Research NBS 17,
(1936), RP 926.
57-9

5.

Standard Electrodynamic Wattmeter and AC-DC Transfer Instrument, J. H. Park and A. B.

Lewis, J. Research 25_, 545, (1940), RP 1344.

6.

Performance of Portable Electrical Instruments in Magnetic Fields, A. E. Peterson,

Trans. Am. Inst. Elec. Engrs. 67, 1228, (1948).

7.

An Easily Assembled Console for Rapid Testing of Electrical Indicating Instruments,

F. D. Weaver, Instruments 22, 396, (1949).

8.

Notes on the Care and Use of Electrical Instruments, F. D. Weaver, Instruments 23_,
No. 12, 1236, (1950).

9.

Thermal Converters as AC-DC Transfer Standards for Current and Voltage Measurements
at Audio Frequencies, F. L. Hermach, J. Research NBS 48_, 121, (1952), RP 2296.

10.

A-C Measurements to 10,000 cps, F. D. Weaver, Instruments 25, No. 6, 757, (1952).

11.

The Testing of Electrical Instruments, F. L. Hermach, Proc. Instr. Soc. Am. 8,

18, (1953).

12.

Testing Electrical Instruments, F. D. Weaver, Instruments 26, No. 9, 1362, (1953).

13.

Lead Resistance Errors in Watthour Meter Tests, F. L. Hermach and T. L. Zapf, Elec.
World
No. 16, 113 (1954).

14.

Multirange, Audiofrequency Thermocouple Instruments of High Accuracy, F. L. Hermach

and E. S. Williams, J. Research NBS 52, 227, (1954), RP 2494.

15.

Precise Comparison Method of Testing Alternating -Current Watthour Meters, A. W. Spinks,

and T. L. Zapf, J. Research NBS 5X 95, (1954), RP 2521.

16.

Scale and Reading Errors of Electrical Indicators, F. D. Weaver, Instr. and

Automation 27, No. 11, (Nov. 1954).

17.

Power Supplies for 60-Cycle Tests of Electrical Instruments and Meters, F. L.

Hermach, Proc. Instr. Soc. Am. 11, Paper No. 56-21-3, (1956).

18.

AC-DC Transfer Instruments for Current and Voltage Measurements, F. L. Hermach,

IRE Trans. Instrumentation, 1-8, 235, (1958).

19.

The Definition and Measurement of the Time Constant and Response Time of Thermal
Converters, F. L. Hermach, Trans. Am. Inst. Elec. Engrs. 77, 277, (1958).

20.

A Wide Range Volt-Ampere Converter for Current and Voltage Measurements, F. L.

Hermach and E. S. Williams, Communications and Electronics, Trans. Am. Inst. Elec.
Engrs. Paper 59-161, (1959).

21.

Thermal Voltage Converters for Accurate Voltage Measurements to 30 Megacycles, F. L.

Hermach and E. S. Williams, Trans. Am. Inst. Elec. Engrs. Paper 60-135, (I960).

22.

An Analysis of Errors in the Calibration of Electrical Instruments, F. L. Hermach,

AIEE Trans. Paper 61 -63, (I960).

23.

A Differential Thermocouple Voltmeter, F. L. Hermach and J. E. Griffin, AIEE Trans.

Paper 62-819, (1962).

24.

The Calibration of Volt-Ampere Converters, E. S. Williams, NBS Tech. Note 188,

(1963), 20 cents.

25.

Practical Aspects of the Use of AC-DC Transfer Instruments, E. S. Williams, NBS

Tech. Note 257, (March 9, 1965), 15 cents.
58-10

26.

A System for Accurate Direct and Alternating Voltage Measurements, F. L. Hermach,

J. E. Griffin, E. S. Williams, IEEE Transactions on Instrumentation and Measure
ments, Vol. IM-14, No. 4, Dec. 1965.

27.

Calibration of Peak A-C to D-C Comparators, Donald Flach and L. A. Marzetta, ISA
Preprint No. 14.2, March 1965.

23.

Thermal Converters as AC-DC Transfer Standards for the Measurement of Alternating

Current and Voltage, F. L. Hermach, Proc. of the 11th Session of the International
Committee of Weights and Measures - Advisory Committee on Electricity, May 1965.

29.

Thermal Converters for Audio Frequency Voltage Measurements, F. L. Hermach and

E. S. Williams, IEEE Trans, on Instrumentation Measurements, 1966.

30.

A Comparator for Thermal AC-DC Transfer Standards, R. S. Turgel, ISA Convention

Record, Fall 1966.
11.

Transformers and Inductive Voltage Dividers

1.

A Method for Testing Current Transformers, F. B. Silsbee, Bui. BS 1^, 317,

(1918-19), S309.

2.

Lead Resistance for Current Transformers, F. B. Silsbee, Elec. World, 81, 1082, (1923).

3.

Methods for Testing Current Transformers, F. B. Silsbee, Trans. Am. Inst. Elec.
Engrs.,
282, (1924).

4.

A Shielded Resistor for Voltage Transformer Testing, F. B. Silsbee, BS Sci. Paper 20,
489, (1924-26), S516.

5.

Equipment for Testing Current Transformers, F. B. Silsbee, R. L. Smith, N. L. Forman,

and J. H. Park, BS J. Research 11, 93, (1933), RP 580.

6.

Accuracy of High-Range Current Transformers, J. H. Park, NBS J. Research 14_, 367,

(1935), RP 775.

7.

Information for the Amateur Designer of Transformer for 25- to 60-Cycle Circuits,
H. B. Brooks, NBS Circular 408, 25 pages, (1935).

8.

Effect of Wave Form Upon the Performance of Current Transformers, J. H. Park,

NBS J. Research 19^ 517, (1937), RP 1041.

9.

Measurement of Voltage Ratio at Audio Frequencies, W. C. Sze, Trans. Am. Inst.

Elec. Engrs., Paper 57-648, (1957).

10.

A Standard Transformer and Calibration Method - A Basis for Establishing Ratios of

Currents at Audio Frequencies, B. L. Dunfee, IRE Trans. Instrumentation 1-9, (I960).

11.

The Precision Measurement of Transformer Ratios, R. D. Cutkosky and J. Q. Shields,

IRE Trans. Instrumentation 1-9, (i960).

12.

Voltage Ratio Measurements With A Transformer Capacitance Bridge, T. L. Zapf,

J. Research NBS 66C, No. 1, 25, (1962).

13.

The Calibration of Inductive Voltage Dividers and Analysis of their Operational

Characteristics, T. L. Zapf, ISA Trans. 2, No. 3, (1963).

14.

An International Comparison of Inductive Voltage Divider Calibrations at 400 and

1000 Hz, W. C. Sze, A. F. Dunn and T. L. Zapf, IEEE Trans, on Instr. and Meas.,
Vol. IM-14, No. 3, Sept. 1965.

15.

Comparators for Voltage Transformer Calibrations at NBS, W. C. Sze, J. Research NBS,

Vol. 69C, No. 4, Oct. - Dec. 1965.
59-11

16.

An International Comparison of Current Ratio Standards at Audio Frequencies, B. L.

Dunfee and W. J. M. Moore, IEEE Trans, on Inst, and Meas., Vol. IM-14, No. 4, Dec. 1965.

17.

The Design and Performance of Multirange Current Transformer Standards for Audio
Frequencies, B. L. Dunfee, IEEE Trans, on Inst, and Meas., Vol. IM-14, No. 4, Dec. 1965.
12.

Galvanometers

1.

General Design of Critically Damped Galvanometers, F. Wenner, Bui. BS 12, 211,

(1916-17), S273.

2.

A New Form of Vibration Galvanometer, P. G. Agnew, BS Sci. Paper 16, 37, (1920), S370.

3.

Sensitivity of a Galvanometer as a Function of its Resistance, H. B. Brooks,

J. Research NBS 4, 297, (1930), RP 150.

4.

Galvanometer Efficiency as a Design Parameter, F. K. Harris, Trans. Am. Inst.

Elec. Engrs. 56-60, (1956).
13.

High-Voltage and Surge Measurements

1.

An Experimental Study of the Corona Voltmeter, H. B. Brooks and F. M. Defandorf,

J. Research NBS 1, 589, (1928), RP 21.

2.

Calculations of Electrical Surge-Generator Circuits, A. B. Lewis, J. Research NBS 17,

585, (1936), RP 929.

3.

An Absolute Electrometer for the Measurement of High Alternating Voltages, H. B.

Brooks, F. M. Defandorf, and F. B. Silsbee, J. Research NBS 20, 253, (1938), RP 1078.

4.

A Transformer Method for Measuring High Alternating Voltages and its Comparison
with an Absolute Electrometer, F. B. Silsbee and F. M. Defandorf, J. Research NBS 20,
317, (1938), RP 1079.

5.

Shunts and Inductors for Surge-Current Measurements, J. H. Park, J. Research NBS 29_,
191, (1947), RP 1823.

6.

The Measurement of High Voltage, F. M. Defandorf, J. Wash. Acad. Sci. 28, No. 2,
33, (1948).

7.

A Fifty-Fold Momentary Beam Intensification for a High Voltage Cold -Cathode

Oscillograph, J. H. Park, J. Research NBS 4Z, 87, (1951), RP 2231.

8.

Puncture Tests on Porcelain Distribution Insulators Using Steep-Front Voltage

Surges, J. H. Park and H. N. Cones, AIEE Paper 53-257, June 1953.

9.

Surge Voltage Breakdown of Air in a Non-Uniform Field, J. H. Park and H. N. Cones,

J. Research NBS $6, 201, (1956), RP 2669.

10.

High Voltage Pulse Generator and Tests on an Improved Deflecting System of a ColdCathode Oscillograph, H. N. Cones, J. Research NBS 5J7, 143, (1956), RP 2704.

11.

Surge Measurement Errors Introduced by Coaxial Cables, J. H. Park, AIEE Paper 58-110,
February 1958.

12.

Special Shielded Resistor for High-Voltage D-C Measurements, J. H. Park, J. Research

NBS Paper 66C1-83; page 19, (1962).

13.

Spark -Gap Flashover Measurements for Steeply Rising Voltage Impulses, J. H. Park,
H. N. Cones, J. Research NBS Paper 66C3-96, page 197, (1962).

14.

An Experimental 350 kv, 1 Picofarad Air Capacitor, Alvin Peterson, IEEE Trans.
Paper No. 63-926, (1963).
60-12

15.

Some Impulse Sparkover Voltages for a 200 cm Sphere Gap, H. N. Cones, IEEE Conf .
Paper No. 63-929, (1963).

16.

An International Comparison of Voltage Transformer Calibration to 350 kv, Wilbur

Sze, Forest K. Harris, N. L. Kuster, 0. I. Petersons, W. J. M. Moore, IEEE Trans.
Paper No. 63-992, (1963).
14-.

Dielectric Measurements
(see also Section 8)

1.

Edge Correction in the Determination of Dielectric Constant, A. H. Scott and

H. L. Curtis, J. Research NBS 22, 747, (1939), RP 1217.

2.

Dielectric Constant, Power Factor and Conductivity of the System Rubber-Calcium

Carbonate, A. H. Scott and A. T. McPherson, J. Research NBS 28, 279, (194-2),
RP 1457. (Describes measurements by a three -terminal bridge.)

3.

Measurements of Dielectric Properties at Temperature up to 500C, A. H. Scott, P.

Ehrlich, and J. F. Richardson, Symp. on Temperature Stability of Electrical
Insulating Materials, Am. Soc. Testing Materials Spec. Tech. Publ. No. 161, (1954).

4.

Precise Measurements of Dielectric Constant over a Wide Range of Frequencies and

Temperatures, A. H. Scott, Proc. Instr. Soc. Am. 11, Paper No. 56-8-2, (1956).

5.

An Ultra Low Frequency Bridge for Dielectric Measurements, D. J. Scheiber,

J. Research NBS 6j>C, 23, (1961).

6.

Residual Losses in a Guard -Ring Micrometer-Electrode Holder for Solid-Disk

Dielectric Specimens, A. H. Scott and W. P. Harris, J. Research NBS 65C, 101, (1961).

7.

Precise Measurement of Dielectric Constant by the Two-Fluid Technique, W. P. Harris

and A. H. Scott, CEI Report NRC-NAS, 51, (1962).

8.

Standards for Plastics - Standard Tests for Electrical Properties, A. H. Scott,

S. P. E. Journal, 1375, (1962).

9.

Dielectric Properties of Semicrystalline Polychlorotrifluorethylene, A. H. Scott,

D. J. Scheiber, A. J. Curtis, J. I. Lauritzen, Jr., J. D. Hoffman, J. Research
66A, No. 4, 269, (1962).

10.

Insulation Resistance Measurements, A. H. Scott, Fourth E. I. Conf., ALEE Paper

No. P-137-52, (1962).
15.

Magnetic Measurements

1.

An Apparatus for Magnetic Testing at Magnetizing Forces Up to 5000 Oersteds, R. L.

Sanford and E. G. Bennett, J. Research NBS 23, 415, (1939), RP 1242.

2.

Permanent Magnets, R. L. Sanford, NBS Circular 448, 39 p., (1944).

3.

A Permeameter for Magnetic Testing at Magnetizing Forces up to 300 Oersteds, R. L.

Sanford and P. H. Winter, J. Research NBS 4Jb 17, (1950), RP 2109.

4.

The Use of an AC Bridge to Measure Core Loss at High Inductance, I. L. Cooter

and W. P. Harris, Trans. Am. Inst. Elec. Engrs., (1956).

5.

Investigation of an Alternating Current Bridge for the Measurement of Core Losses

in Ferromagnetic Materials at High Flux Densities, I. L. Cooter and V/. P. Harris,
J. Research NBS 7,103, (1956), RP 2699.

6.

Improved Bridge Method for the Measurement of Core Losses on Ferromagnetic Materials
at High Flux Density, W. P. Harris and I. L. Cooter, IRE Natl. Conv. Record, Pt. 5,
217, (1958).
61-13

7.

A Feedback Amplifier with Negative Output Resistance for Magnetic Measurements, I. L.

Cooter and W. P. Harris, IRE Natl. Conv. Record, Pt. 5, 217, (1958).

3.

Basic Magnetic Quantities and the Measurement of the Magnetic Properties of Materials,
R. L. Sanford, I. L. Cooter, NBS Mono. 47, (1962), 30 cents.

9.

Absolute Magnetic Susceptibilities by the Gouy and the Thorpe -Senftie Methods, George
A. Candela, IRE Trans, on Instrumentation, I -11, Nos. 3 and 4, (1962).

10.

Absolute Magnetic Susceptibilities by the Thorpe -Senftie Method, G. A. Candela and

R. E. Mundy, Rev. of Sci. Inst., Vol 22, No. 9, pp. 1056-1057, Sept. 1961.

11.

A New Absolute Null Method for the Measurement of Magnetic Susceptibilities in Weak
Low-Frequency Fields, Charles T. Zahn, Rev. Sci. Inst. 2A, No. 3, 285 - 291 (1963).

12.

The Calibration of Permanent Magnet Standards, Irvin L. Cooter, Reprint No.

14.1-3-65, Proc. 20th Annual ISA Conf., Oct. 1965.
16.

Miscellaneous

1.

Accuracy Tests for Meggers, H. B. Brooks, Elec. World 85^ 973, (1925).

2.

Precautions Against Stray Magnetic Fields in Measurements with Large Alternating

Currents, F. B. Silsbee, Trans. Am. Inst. Elec. Engrs. 48, 1301, (1929).

3.

Suggested Practices for Electrical Standardizing Laboratories, F. B. Silsbee, NBS

Circular 578, 9 p., (1956), 15 cents.

4.

Copper Wire Tables, NBS Handbook 100, February 1966, 50 cents.

5.

Measurement of the Resistance-Strain Relation and Poisson's Ratio for Copper Wires,
T. E. Wells, Proc. Instr. Soc. Am. 11, (1956).

6.

Electrical Measurement in the Core Curriculum, F. R. Kotter, Proc. Am. Inst. Elec.
Engrs., Analog & Digital Instrument Conf., (April 1959).

7.

Phase Angle Master Standard for 400 cycles per second, J. H. Park and H. N. Cones,
J. Research NBS 64C3, 229, (i960).

8.

Shielded Coaxial Leads for Low Temperature Electrical Measurements, N. L. Brown and
R. N. Barfield, Rev. Sci. Inst. 21, (5), 517, (I960).

9.

Achievement of Measurement Agreement Among Electrical Standards Laboratories,

F. D. Weaver, Instruments and Control Systems, 26, No. 7, 128, (1963).

10.

Method for Comparing Two Nearly Equal Potentials Directly in Parts Per Million, C. J.
Saunders, Review of Scientific Instruments, Vol. 34, No. 12, 1452 - 1453, Dec. 1963.

11.

Low Frequency Electrical Calibrations at NBS, F. L. Hermach, National Conf. of

Standards Laboratories, NBS Misc. Publ. 248, (1962), $1.75.
17.

References to Books and Standards on Electrical

Instruments and Measurements

The National Bureau of Standards receives frequent inquiries for information on

electrical instruments and measurements which is not specifically covered in its publi
cations. To meet the needs of such inquiries a list of books and standards is given
below.

62-14

17.1

Textbooks

(Arranged alphabetically by author's name)

1.

Electrical Measurements and Instrumentation, D. Bartholemew, (Allyn & Bacon,

Boston, 1963).

2.

Principles of Electrical Measurements, H. Buckingham and E. M. Price,

(Philosophical Library, New York, New York, 1957).

3.

Electrical Measurements, H. L. Curtis, (McGraw-Hill Book Company, New York,

New York, 1937) .

4.

Electrical Measuring Instruments, Drysadle and Jolley - revised by Tagg (John

Wiley and Sons, New York, New York, 1952).

5.

Industrial Instrumentation, D. P. Eckman, (John Wiley and Sons, New York,

Mew York, 1950).

6.

Electrical Measurements and Their Application, E. Frank, (McGraw-Hill Book

Company, New York, New York, 1959).

7.

Electrical Measurements and Measuring Instruments, E. W. Golding, (Pitman

Publishing Company, New York, New York, 1955).

8.

Inductance Calculations, F. V/. Grover, (D. Van Nostrand & Co., Princeton, New
Jersey, 1946).

9.

Alternating -Current Bridge Methods, B. Hague, (Pitman Publishing Co., New York,
New York, 1957).

10.

Instrument Transformers, B. Hague, (Pitman Publishing Co., New York, N. Y. , 1936).

11.

Structure of Electrolytic Solutions, Edited by W. J. Hamer, Published by John

Wiley and Sons, New York, New York, (1959).

12.

Electrical Measurements, F. K. Harris, (John Wiley & Sons, New York, N. Y. , 1952).

13.

Applied Electrical Measurements, I. F. Kinnard, (John Wiley & Sons, New York,
New York, 1956) .

14.

Electrical Measurements, F. A. Laws, (McGraw-Hill Book Company, New York, N. Y. , 1938).

15.

Electrical Measurements and Their Applications, W. C. Michels, (D. Van Nostrand

and Company, Princeton, New Jersey, 1957).

16.

Elektrische Messgerate und Einrichtungen, A. Palm (Springer, Berlin, Germany, 1948).

17.

Principles of Electronic Instruments, G. R. Partridge, (Prentice Hall, Englewood

Cliffs, New Jersey, 1958).

18.

Electronic Measurements, F. E. Terman & J. M. Pettit, (McGraw-Hill Book Company,

New York, New York, 1952).

19.

Elektrische Messgerate und Messverfahren, P. M. Pflier (Springer, Berlin, Germany, 1957 ) .

20.

Basic Electrical Measurements, M. B. Stout, (Prentice Hall, Englewood Cliffs, New

Jersey, I960).

21.

Storage Batteries, 4th Edition, George W. Vinal (John Wiley & Sons, New York,
New York, 1955).

63-15

17.2

Standards and Handbooks

Standards and Handbooks are frequently revised, and the current edition should
always be consulted.
1.

American Standard for Electrical Indicating Instruments, (Am. Standards Association,

10 East 40th Street, New York, New York, C39.1) (Supersedes AIEE Standard No. 33).

2.

American Standard for Instrument Transformers, (American Standards Association,

10 East 40th Street, New York, New York, 056.13).

3.

Code for Electricity Meters, (Edison Electric Institute, New York, New York).

4.

Standard Handbook for Electrical Engineers, (McGraw-Hill Book Company, New

York, New York).

5.

Electrical Metermen's Handbook, (Edison Electric Institute, New York, New York).

6.

Master Test Code for Resistance Measurements, (Institute of Electrical & Electronic
Engineers, IEEE Publication 118, Supersedes AIEE Publication No. 550).

7.

Automatic Digital Voltmeters and Ratio Meters, Part 1 - Direct Current Instruments,
(American Standards Association, C39.6).

64-16

ADDENDUM
September 1967

Units and Absolute Electrical Measurements

1.

Physical entitles and mathematical representation, C. H. Page, J. Research NBS 65B,

No. 4, 227 (1961).

2. Klectrical units, F. K. Harris, ISA 19th Ann. Conf., Preprint 12.1.1.64 (1964).
3.

International weights and measures, C. H. Page, ISA 20th Ann. Conf., Preprint 33.3-1-65
(1965).

4. Electrical standards and measurements, I. L. Cooter, B. L. Dunfee, F. K. Harris, W. P.

Harris, F. L. Hermach, and C. Peterson, Electro-Technology 79, No. 1, 53 (1967).
Standard Cells
L. Oil baths for saturated standard cells, P. H. Lowrie, Jr., ISA J. 9, No. 12, 47 (1962).
2. The establishment and maintenance of the unit of voltage at the NBS Boulder Laboratories,
B. A. Wickoff, ISA 18th Ann. Conf., Preprint 28.2.63 (1963).
3. Designs for surveillance of the volt maintained by a small group of saturated standard
cells, W. G. Eicke and J. M. Cameron, NBS TN 430 (1967).

Inductors
1. Some techniques for measuring small mutual inductances, D. N. Homan, J. Research NBS
70C, No. 4, 221 (1966).
Capacitors
1. Active and passive direct- reading ratio sets for the comparison of audio- frequency
admittances, R. D. Cutkosky, J. Research NBS 68C, No. 4, 227 (1964).
2.

Improved ten-picofarad fused silica dielectric capacitor, R. D. Cutkosky and L. H. Lee,

J. Research NBS 69C, No. 3, 173 (1965).

3. Voltage dependence of precision air capacitors, J. Q. Shields, J. Research NBS 69C,

No. 4, 265 (1965).
Potentiometers, Volt Boxes, and Bridges
1.

Calibration of potentiometers by resistance bridge methods, D. Rama ley, Instr. Control

Syst. 37, 106 (1964).

2. Some modifications in methods of calibration of universal ratio sets, D. Rama ley, NBS
TN 220 (1964).
3. Human engineering in the design of a console for the comparison of volt boxes, P. H.
Lowrie, Jr., ISA J. 12, No. 7, 67 (1965).
4.

Console for the rapid and precise comparison of volt boxes, P. H. Lowrie, Jr., J. Re
search NBS 70C, No. 3, 173 (1966).

5.

A versatile ratio instrument for the high ratio comparison of voltage or resistance,
A. E. Hess, J. Research NBS 70C, No. 3, 169 (1966).

6.

Direct ratio readings from a URS , D. Rama ley and J. F. Shafer, Instr. Control Syst. ^9
No. 1, 73 (1966).
Instruments, Meters, and AC-DC Transfer Standards

1.

DC differential current meter, E. Nlesen, Rev. Sci. Instr. 32, No. 12, 1407 (1961).

2.

Voltage ratio detector for millivolt signals, J. R. Houghton, NBS TN 266 (1965).

Transformers and Inductive Voltage Dividers

1.

Inductive voltage dividers with calculable relative corrections, T. L. Zapf, C. H.

Chinburg, and H. K. Wolf, IEEE Trans. IM-12, No. 2, 80 (1963).

2.

Comparison calibration of Inductive voltage dividers, R. V. Lisle and T. L. Zapf, ISA

Trans. 3, No. 3, 238 (1964).

3.

The accurate measurement of voltage ratios of inductive voltage dividers, T. L. Zapf,

ACTA IMEKO 3, 317 (1964).

4.

Comparator for calibration of Inductive voltage dividers from 1 to 10 kHz, W. C. Sze,

ISA 21st Ann. Conf., Preprint 12.3.2.66 (1966).

High Voltage and Surge Measurements

1.

The design and operation of a high voltage calibration facility, W. W. Scott, Jr.,
NBS TN 349 (1966).
Dielectric Measurements

1.

Two-terminal dielectric measurements up to 6 x 10

J. Research NBS 69C, No. 3, 165 (1965).

Hz, M. G. Broadhurst and A. J. Bur,

2.

Electrical testing, A. H. Scott, Ency. Chem. Tech. ]_> 716 (1965).

3.

Low frequency dielectric behavior, W. P. Harris, Proc. Electrical Insulation Conf.,

IEEE-NEMA (1967).
Miscellaneous

1.

A standard for accurate phase-angle measurements at audio frequencies, W. W. Scott, Jr.,

NBS TN 347 (1966).

2.

Low- frequency electrical calibrations at the National Bureau of Standards, F. L.

Hermach, ISA 22nd Ann. Conf. (1967).

66-18

JOURNAL OF RESEARCH of the National Bureau of StandardsB. Mathematics and Mathematical Physics
Vol. 65B, No. 4, October-December 1961

Physical Entities and Mathematical Representation

Chester H. Page
(August 22, 1961)
Certain basic postulates about physical observables yield the structure of their mathe
matical representation. Measure equations are contrasted with quantity equations, and
measurement units with abstract units. The abstract vector spaces in which observables
are represented comprise the core of dimensional analysis.
Systems of equations, units, and dimensions are discussed, along with comments on
rationalization. The problem of assigning a dimension to angle is discussed, and a new
proposal offered.

NBS Monograph 56
September 1962

Systems of Electrical Units

Francis B. Silsbee
(January 19, 1962)
The various systems of measurement, with their respective sets of units, used in the
literature on electricity and magnetism are described in detail. Their historical develop
ment is summarized. The manner in which each is derived from either of the two alternative
points of view of the experimentalist and the theoretician is compared and contrasted.
The desirability of recognizing both points of view in international standardization, partic
ularly when discussing rationalization, is pointed out. The present status of the absolute
measurements on which all electrical units are based is reported, and tables are included for
the conversion of equations and numerical values from one system to another.

67-1

NBS Circular 475

June 1949

Establishment and Maintenance of the Electrical

Units
By F. B. Silsbee
Abstract
A history is given of the establishment of the "international" system of electrical units,
its operation during the interval 1911 to 1947, inclusive, and of the developments that
caused it to be superseded. It includes a record of the international comparisons, which
indicated that the units of some countries have at times drifted nearly 0.01 percent from the
mean, an account of the maintenance procedures used at the National Bureau of Standards,
and brief descriptions of the methods currently available for the absolute measurement of
resistance and of current, which will be used in the future as a check on the maintenance
of the units.

NBS Circular 531

July 1952

Extension

and

Dissemination

of

the

Electrical

and

Magnetic Units by the National Bureau of Standards

By Francis B. Silsbee
Starting with the ohm and the volt as maintained by groups of standard resistors and
cells, this paper describes the experimental processes by which the other electric and magnetic
units, e. g., farad, henry, ampere, watt, joule, gauss, and oersted are derived. It also de
scribes the series of steps by which the scales of measurement of resistance, direct and alter
nating current, and voltage are derived experimentally. Brief mention is made of the pro
cedures for the dissemination of these standards of measurement throughout the world by
the calibration of standard electrical measuring apparatus. An extensive bibliography lists
papers describing the measurement procedures in greater detail and serves as a historical
report of the work of the National Bureau of Standards in the field of electrical measurements
during its first 50 years.

68-2

Jourrid! of Research of the National Bureau of Standards

Vol. 60, No. 4, April 1958

Research Paper 2845

Measurement of Current with a Pellat-Type Electrodynamometer

R. L. Driscoll
The value of an electric current has been determined in absolute measure by means of
an electrodynamometer, and simultaneously by standard cells and standard resistors as
currently maintained. The electrodynamometer used was of the Pellat type, and featured
a fused silica balance beam and single layer helical coils.
The relation of the NBS ampere to the absolute ampere, from this determination, may
be expressed as
1 NBS ampere = 1.000013 absolute amperes. *
The uncertainty in this result from all known sources is estimated to be eight parts per
million.

Journal of Research of the National Bureau of Standards

Vol. 60, No. 4, April 1 958

Research Paper 2846

Measurement of Current with the National Bureau of

Standards Current Balance
R. L. Driscoll and R. D. Cutkosky
Prior to the adjustment of the electrical units in 1948, the value of a current had been
determined in absolute units by means of a current balance and simultaneously measured
in XBS amperes by comparison with standard resistors and standard cells. This work
was reported in RP1449. Similar measurements made recently with an electrodynamom
eter indicate a possible change in the values of the standards. The present paper reports
a repetition of the work described in RP1449. The purpose, of this remeasurement was
to determine whether or not the standards had changed. Only minor changes were made
in the equipment in order that factors which might have introduced small systematic errors
in the results would remain unchanged.
According to the work described in this paper, 1 NBS ampere = 1.000008 absolute
amperes. Recent work with the Pellat electrodynamometer gave the result 1 NBS ampere
= 1.000013 absolute amperes. The weighted mean of these two values is
1 NBS ampere-1.000010 0.000005 absolute amperes *
The results given above for the current balance differ by 6 ppm from those obtained in
1942. This indicates, in view of the uncertainties of measurement, that any change in
the ampere as maintained by standard resistors and standard cells does not exceed a few
parts in a million.

*The above values for the NBS ampere are based on the Dryden
reduction value of the acceleration of gravity, g. In light of new
determinations of g, the NBS ampere is
1 NBS ampere - 1.000012 0.000005 absolute amperes.
This value is the weighted mean of the results obtained with the
NBS current balance and the Pellat electrodynamometer.

R. L. Driscoll
November 1967
69-3

JOURNAL OF RESEARCH of the National Bureau of StandardsA. Physics and Chemistry

Vol. 65A, No. 3, May-June 1961

Evaluation of the NBS Unit of Resistance Based on

Computable Capacitor
Robert D. Cutkosky
(January 16, 1961)
An evaluation of the unit of resistance maintained at the National Bureau of Standards,
based on the prototype standards of length and time, is described. The evaluation is based
on a nominally one-picofarad capa:itor whose value may be calculated from its mechanical
dimensions to high accuracy. This capacitor is used to calibrate an 0.01-microfarad ca
pacitor. A frequency-dependent bridge involving this capacitor establishes the value of a
lO'-ohm resistor. Comparison of that resistor with the bank of one-ohm resistors main
taining the NBS unit of resistance establishes that this unit is
tlBV= 1.000002 3 ohms 2.1 ppm*
The indicated uncertainty is an estimated 50 percent error of the reported value based on
the statistical uncertainty of the measurements and allowing for known sources of possible
systematic errors other than in the speed of light, assuming that the speed of light
c = 2.9971)25 X 1 010cm/sec.

*Since the publication of this paper, it was discovered that

the small correction resulting from the ac-dc differences of the
10^-ohm resistors was misapplied. The corrected value for the
NBS unit of resistance is
1.000000

ohms 2.1 ppm.

R. D. Cutkosky
November 1967

70-4

Standard

Zener

Cells

and

Diodes

Papers
Standard cells, their construction, maintenance, and characteristics,
W. J. Hamer
Oil baths for saturated standard cells, P. H. Lowrie
The operating characteristics of zener reference diodes and their measure
ments, W. G. Eicke
.
Making precision measurements of zener diode voltages, W. G. Eicke. .

73
114
118
125

Abstracts
Designs for surveillance of the volt maintained by a small group of
saturated standard cells, W. G. Eicke and J. M. Cameron
315-411 O - 68 - 6

71

131

UNITED STATES DEPARTMENT OF COMMERCE

Luther H. Hodges, Secretary

NATIONAL BUREAU OF STANDARDS A. V. Astin, Director

Standard Cells

Their Construction, Maintenance,

and Characteristics

Walter J. Hamer

National Bureau of Standards Monograph 84

Issued January 15, 1965

73-1

Foreword
Accurate measurement of electromotive force is important in many areas of science and
technology. Physical standards for such measurement are provided by standard cells, which
are long-lived electrochemical systems of highly stable electromotive force.
This publication gives the origin and derivation of the unit of electromotive force and outlines
the procedures by which the National Bureau of Standards maintains and disseminates this unit
by means of standard cells. Information is also given on the construction, maintenance, and
characteristics of standard cells as well as a history of their development. Emphasis is placed
on the precision and accuracy of electromotive force measurements; the stability of standard cells,
especially those of the National Reference Group; and efforts made to construct standard cells of
high quality.
A. V. Astin, Director

Library of Congress Catalog Card Number: 64-60065

74-11

Contents
Foreword
1. Introduction
2. The unit of electromotive force
2.1. Realization
2.2. History
2.3. Maintenance
2.4. Dissemination
2.5. International comparisons
3. Early standard cells
4. The Clark cell
5. The Weston (or cadmium sulfate) cell
5.1. General
5.2. Preparation and properties of materials
5.3. Containers for standard cells
5.4. Assembly and mounting of standard cells
5.5. Electromotive forces of newly made cells
6. Effect of variations in components on the electromotive force of stand
ard cells
6.1. Concentration of solution
6.2. Acidity of solution
6.3. Composition of amalgam
6.4. Crystal phases of cadmium sulfate
7. Characteristics of standard cells
7.1. Emf-temperature coefficient
7.2. Emf-temperature hysteresis
7.3. Temperature range
7.4. Emf-pressure coefficient
7.5. Internal resistance
7.6. Effect of current
7.7. Effect of light
7.8. Effect of shock
7.9. Effect of vibration
8. Life of standard cells
9. References

10. Appendix 1
11. Appendix 2
12. Appendix 3
13. Appendix 4
14. Appendix 5
15. Appendix 6
16. Appendix 7

75-IV

IV'
11
1
1
1
4
5
8
10
11
13
15
15
16
19
20
21
22
22
22
23
23
23
23
26
27
27
28
28
30
30
30
30
31
33
34
35
35
36
37
37

Standard Cells
Their Construction, Maintenance, and Characteristics
Walter J. Hamer
This Monograph contains information on the construction, maintenance, and characteristics of
standard cells. The effects of temperature, pressure, electric current, light, shock, and vibration on
standard cells are discussed. A history of the realization and maintenance of the unit of electromotive
force is also included. A record of international comparisons of the unit of electromotive force is pre
sented as well as information on the constancy of the National Reference Croup of Standard Cells.
1. Introduction
Standard cells are physical representations of
the unit of electromotive force (emf), serve in the
maintenance of the unit, and are used as standards
with which the emf of other cells and systems and
IR drops are compared. Together with standards
of resistance (R) they are also used in the measure
ment of current, /. When measurements of electric
power, P, are made in terms of standards for emf
() and resistance, the expression for power,
P = E2/R, shows the necessity of knowing E ac
curately, since a small error in the standard for E
would produce a percentage error twice as- great
in the value for the power, P.
Standard cells are electrochemical systems com
posed of two dissimilar electrodes immersed in an
electrolytic solution. They are not intended to
supply electric current and, therefore, are of dif

ferent design from those electrochemical systems

which are intended for such purpose. Owing to
their special use, standard cells are required to
meet certain performance criteria and, for precise
measurements, to have certain inherent charac
teristics. They must be reasonably reproducible,
exhibit good permanency, possess low emf-temperature coefficients, have a low or moderately
low internal resistance, be relatively insensitive
to current drains of low magnitude, and, if pos
sible, have an emf of convenient magnitude. Since
a standard cell is a physical representation of a unit
it is obvious why permanency is of prime importance
in a standard cell. The precision with which the
emf of standard cells is measured, accordingly,
exceeds that normally required for other types of
galvanic cells.

2. The Unit of Electromotive Force

2.1. Realization
The practical unit of emf, the volt, is not an arbi
trary one but like the other electrical units is de
rived from the basic mechanical units of length,
mass, and time using the principles of electromagnetism with the value of the magnetic constant
(the so-called permeability of free space) taken as
47r/107 in the rationalized mksa (meter-kilogramsecond-ampere) system of units. 1 It has been
customary, following the first use of the term by
Gauss [9],2 to refer to electrical units based on the
basic units of length, mass, and time as absolute
electrical units.3 The transition from arbitrary to
absolute units began with the work of Gauss [9]
in 1833 and of Weber [2] in 1851, who showed that
1 This is the Giorgi system [1], which is a pirt of ihe Systeme International d'Unitfs
(SI), adopted in a resolution, 11th Genera] Conference on Weights and Measures,
Paris, October 1960. Other systems for the basic units of length, mass, and time
have been used. These include the millimeter milligram-second system of Weber [2),
the meter gram-second system recommended by the first committee of the British
Association for the Advancement of Science appointed to consider electrical units (31,
the quadrant-(elevenlh-gram>-second or Q.E.S. system of Maxwell [4). the foo|.grainsecond system used in England for a time [5], and the cgs (centimeter-gram-second
system [6] (see Appendix 1). Heaviside also proposed a rationalized system wherein
the cgs unit of emf was increased by a factor V4ir (unit of resistance increased by a
factor 4rr and the unit of current decreased by the factor V4ir j7J. In ail of these,
Ihe Giorgi system excepted, the magnetic constant (the permeability of free space) is
taken as unity. Regardless of which system is chosen the practicaJ system remains
unaffected. Additional information on electrical units is given in reference {8].
1 Figures in brackets indicate the literature references on page 31.
' It is unfortunate that the name "absolute" has persisted. It is sometimes wrongly
interpreted to imply that there are no errors involved in the measurements or that per
fection has been attained.

it was possible to measure electrical quantities

in terms of mechanical units. Weber pointed out
the desirability of making the electrical units con
sistent with those used in other branches of science
and engineering.
The electrical units determined in the cgs electro
magnetic (em) system are of inconvenient size for
practical use. For example, Sir William Thomson
(Lord Kelvin) [10] in 1851 showed that the emf of
a Daniell cell was about 1 X 108 cgs em units. Even
so, a Committee of the British Association for the
Advancement of Science in 1873 [6] recommended
the cgs em system for use both in basic science and
practical engineering. However, the practitioners,
although they agreed with the desirability of re
lating electrical units to mechanical ones, objected
to the use of the cgs em system of units in practice
both because of the magnitudes involved and be
cause they had been using such terms as ohms and
volts for their units. Their views prevailed and in
1881 the International Congress of Electricians
[11] meeting in Paris adopted the cgs em system
of units as the fundamental system and the voltohm-ampere system for practical use, with the
practical units being made larger or smaller than
the corresponding cgs em units by an appropriate
power of 10.

76-1

The factors chosen for emf, resistance, and

current were:
1 volt (practical unit)

= 10s cgs electromagnetic

units of emf

1 ohm (practical unit)

= 109 cgs electromagnetic

units of resistance

1 ampere (practical unit)= 10"' cgs electromagnetic

unit of current.
The factor 10s was chosen for emf since then the
emf of a Daniel! cell, widely used at that time as a
rough standard of emf, became approximately 1
volt. The factor 10* was chosen for resistance, for
in this way the value of the Siemens mercury col
umn, already used as a resistance standard,
especially on the European continent, became
approximately 1 ohm (actually about 0.94 ohm).
The factor for the ampere was then fixed as 10-1
by the requirement of Ohm's law.
The unit of emf, although a most important unit,
is obtained from the ohm and the ampere. To
date, no direct absolute measurement of emf in
the em system of units has been found feasible.
Instead its value is established experimentally
in em units through Ohm's law and the measure
ment of the fail of potential produced in a resistance
by a current, each of these being capable of deter
mination in absolute measure.4
The ohm in absolute measure is usually obtained
in terms of length and time by means of inductance
and frequency. In the Wenner [15] method, a
mutual inductor of known dimensions, and thus of
calculable inductance, is placed in a suitable cir
cuit containing the resistor the value of which is to
be determined, a battery, a galvanometer, and two
rotary reversing switches. The primary of the
mutual inductor is placed in series with the battery
and resistor, and with one rotary reversing switch
so arranged as to reverse the connections to the
primary. The secondary of the mutual inductor
is connected to the potential terminals of the re
sistor through a galvanometer and a rotary revers
ing switch for reversing the connections to the
secondary terminals. The galvanometer detects
the balance between the induced emf in the sec
ondary and the IrR drop across the resistor. When
balance is obtained, i.e., when the galvanometer
shows no deflection, IrR = ^nlpM where Ir is the
current in the resistor, Ip the current in the primary,
M is the computable mutual inductance, and n is
* The unit of emf may also be determined in cgs electrostatic (es) units, with the
electric constant (the so-called permitivity of free space) taken as unity, using absolute
electrometers. The accuracy of the methods, however, is much lower than the em
approach, the order of magnitude being I pan in 10.000. The unit of emf in es units
is, therefore, arrived at indirectly by converting em values to es values using the ex.
perimental factor, (2.99792S 0.0000031x 10" cm sec'1, the speed of propagation of
an electrical disturbance (within experimental error, the speed of light in free space).
As an example. Sir William Thomson (Lord Kelvin) [12] found the emf of a Daniel!
cell to be 0.00374 cgs es unit from the attraction between two parallel disks connected
to the opposite poles of a Daniel) cell. This value of Thomson is equivalent to
1.121 X 10" cgs em units or 1.121 volts.
An atomic standard for the volt, although most desirable, has not been forthcoming
to date. The Stark effect has been proposed [13] but can be used only to measure
relative voltage and is of insufficient accuracy, with the limiting factor residing in the
measurement of the electric dipole moment 1 14].

the frequency of commutation in cycles per second

(2 n = a number of reversals per second of the rotary
reversing switch in the secondary). The circuit is
so arranged that during that portion of the cycle
when the primary of the mutual inductor is in series
with the resistor, /r = /,>; thus
R = 4nM.
(1)
Since M is calculated from dimensions with refer
ence to length standards and from the permeability
of the medium, and n is measured indirectly in
terms of the unit of time, R is given in terms of the
basic units of length and time and of the permeabil
ity of the space surrounding the windings of the
inductor. The overall precision (repeatability) of
the method is 5 parts per million (ppm). For
more details, reference [16] should be consulted.
The ohm in absolute measure may also be obtained
with similar accuracy using a self inductor [17],
or with lesser accuracy by measuring the relative
rotation of a coil and a magnet or the motion of a
coil in the earth's magnetic field. Although the
latter methods are now obsolete they are mentioned
here because of their historical importance in the
evolution of the electrical units.
The ohm in absolute measure may also be ob
tained using a computable capacitor [18]. In this
method the evaluation of resistance is based on a
nominally 1-pf capacitor whose value in es units
may be calculated with high accuracy from its di
mensions and thence in em units using the speed
of light. This capacitor is then used to calibrate
0.01-/if capacitors, the admittances of which are
then compared with that of a 104-ohm shielded a-c
resistor using a special bridge network. A con
ventional d-c step-down is then used to provide
"absolute" calibration of d-c resistors of lower
magnitude, specifically 1 ohm. For more details,
reference [18] should be consulted. The overall
precision (repeatability) of the method is about 2
ppm. Operationally, this method is less involved
than the inductance methods and may be used on
an annual basis to check on the constancy- of the
resistance standard.
The ampere in absolute measure is obtained in
terms of length, mass, and time with a current
balance [19]. In the current balance the electrodynamic force of attraction or repulsion, in the x
(vertical) direction, between two coils (one movable
and one stationary) through which a current is
flowing is balanced against the force of gravity,
mg, acting on a known mass, m, at a location where
g is the acceleration due to gravity. From this
force and the measured dimensions of the coils
(from which the rate of increase, dMjdx, of the mu
tual inductance between the stationary and mov
able coils may be calculated) together with the
permeability of space in which the coils act, the
current is expressed in terms of units of length,
mass, and time, where time enters into the calcula
tions through the acceleration of gravity. The
overall precision (repeatability) of this method is
about 6 ppm. The ampere may also be obtained

77-2

with an uncertainty of about 8 ppm by an electrodynamometer in which a torque rather than a force
is measured [20]. The combined precision (re
peatability) in determining the ampere in absolute
measure is about 5 ppm. For more details
references [19] and [20] should be consulted.
The ampere in absolute measure may be obtained
qualitatively by using a tangent or sine galvanometer
where the current is measured in terms of the de
flection of a magnetic needle, the radius of the
galvanometer coil, and the horizontal component
of intensity of the earth's magnetic field. Although
these measurements are now obsolete they are of
historical importance; see later. The current
balance methods are involved and accordingly are
used only about once every five or ten years. How
ever, in recent years, following the discovery of
nuclear resonance, a new method, although not an
"absolute" one, has been developed to check on the
constancy of the ampere as maintained at NBS in
terms of the physical standards of resistance and
emf. This method [21] is based on the determina
tion of the gyromagnetic ratio of the proton in
which the precession frequency of the proton is
measured in a magnetic field known in terms of
the "absolute" ampere and the constant of an ac
curately made solenoid. By this method drifts
in the national ampere of about 1 ppm over a
period of years can now be detected; also the
method may be used annually to check the con
stancy of the ratio of the national standards of emf
and resistance (see sec. 2.3). For more details on
this method, reference [21] should be consulted.
The volt in absolute measure is obtained at the
same time a measurement of the absolute ampere
is made. The usual procedure is to connect in
series with the coils of the current balance a re
sistor having a value, R, previously obtained in
absolute measure. The voltage drop, IR, in this
resistor is opposed to the emf, E, of a standard
cell. When the current, /, is held at such a value,
that IR balances E and at the same time the mass,
m, of the weight has been adjusted to balance the
electrodynamic force, the relation
E = R[mgl(dMldx)Yi* = RI

(2)

follows and gives the emf of the standard cell in

absolute measure. The experimental uncertainty
in the balance between the emf of the standard
cell and the IR drop in the standard resistor is less
than 0.1 microvolt. Considering the experimental
precisions of the absolute determinations of R and /,
<he precision in the determination of the volt in
absolute measure is about 7 ppm (rms).
The volt, as now maintained in the United States,
may exceed the absolute volt by as much as 18 ppm
or as little as 2 ppm. Driscoll and Cutkosky [19]
in 1958, as a result of measurements with a current
balance and a Pellat electrodynamometer reported
that
1 NBS ampere = 1.000010 0.000005 absolute
amperes.

In these measurements they used a value for gravity

17 ppm lower than that derived from Potsdam.
Recent studies have shown that 13 ppm is a better
correction of the value derived from Potsdam.
Using this correction,
1 NBS ampere = 1.000012 0.000005 absolute
amperes.
Cutkosky [18] in 1961, as a result of studies with
precision computable capacitors, found that 5
1 NBS ohm = 1.0000006 0.0000021 absolute ohms.
He also, by recalculating the results of Thomas,
Peterson, Cooter, and Kotter [16] obtained with a
Wenner mutual inductor, gave
1 NBS ohm = 0.999997 absolute ohm.
(J. L. Thomas has estimated the uncertainty in this
value to be 0.000005). From these relations for
the ohm and using the second relation for the
ampere,
1 NBS volt = 1.000013 0.000005 absolute volts,
or
1 NBS volt = 1.000009 0.000007 absolute volts.
The NBS volt, as now maintained by standard
cells at the National Bureau of Standards, is the
one disseminated. When additional data on the
above relations are obtained in the United States
and other countries, the International Committee
of the Bureau International des Poids et Mesures,
Sevres, France, which, by international treaty, has
authority to coordinate the standards of measure
ments in the field of electricity as well as of length
and mass, may recommend adjustment in the electri
cal units. Until then, the units as maintained and
disseminated by NBS serve to place laboratories
in the United States on the same standard basis,
accepted by all nations cooperating in the Treaty
of the Meter.
The emf of any or all standard cells could be
determined in like fashion. Obviously, such a
procedure would be cumbersome and time con
suming and if required would be unfortunate,
indeed. Absolute measurements are involved,
require painstaking work and, therefore, are
unsuitable for frequent or routine measurements of
emf. To circumvent the necessity for frequent
absolute measurements, standard cells 6 are con
structed, their emfs are determined in relation to
current and resistance in absolute measure, and
these cells are then used to maintain the volt; these
'This value differ* by 1.7 ppm from lhal (1.000002,1 published by Cutkosky [18).
He has since uncuvered an error uf (his magnitude in the published value.
* Standard resistors, usually hermetically sealed coils of annealed manganin wire
[22]. are used similarly to maintain the unit of resistance. Owing to the transitory
nature of an electric current no physical standards for an ampere have been, to date,
possible. Instead the standard amperr is given by the ratio of the values for the stand
ard cell and the standard resistor.

78-3

cells are also used to assign emfs to other cells that

may be constructed. In other words, the results of
absolute measurements are preserved in a physical
object, the standard cell.6 The validity and reali
zation of this approach depends on the possibility
of constructing standard cells, the emfs of which
are independent (or nearly so) of time; otherwise,
the unit of emf would be lost or drift in value in the
interim and a repetition of absolute measurements
would be imperative at frequent intervals. It is
obvious that this matter was and is of critical im
portance. Over the years extensive work has
been conducted to find that electrochemical system
which would exhibit constant emfs for long periods
of time, say decades. Fortunately, such an elec
trochemical system (see below under the Weston
Cell) has been found. The manner in which the
unit of emf is maintained by a group of standard
cells is discussed below under maintenance.

legal ohms) 7 (known in cgs em units) as standards

of resistance. Their results are summarized in
table 1. In each case, the results were converted
to a common international basis (see Appendix 1
for the defined international units). The emftemperature formula of Lord Rayleigh [24, 29] which
was then available was used to convert all the values
to a temperature of 15 C. The values were re
ferred to as "international" because international
agreement had been attained; it was also realized
at the time that additional work would be needed to
place the values on a basis truly representative of
the theoretical cgs em units.
In the years immediately following 1893 most
countries adopted 1.434 V for the emf of the
Clark cell at 15 C. Later work showed that cells
made with specially purified mercurous sulfate (see
below for a description of the Clark cell) had an
emf 0.0003 V lower than 1.434 V obtained for cells
made prior to the meeting of the Chicago Inter
2.2. History
national Electrical Congress. Germany, however,
in 1898 recommended and adopted 1.4328 V at
In 1893 the International Electrical Congress
15 C for the Clark cell based on the German values
meeting in Chicago chose the Clark cell, a cell
for
the ohm and ampere. Although the German
devised by Latimer Clark in 1872 [23], as the stand
value was not universally accepted, it was a more
ard of emf to which they assigned a value of 1.434
nearly correct value as later experiments showed.
international volts at 15 C in terms of the then
Also during the years 1893 to 1905 the standard cell
accepted standards for the ohm and the ampere.
devised by Edward Weston [30] was found to have
The Clark cell and this value for it were legalized
many advantages over the Clark cell and at an
as the standard of emf in the United States by an
informal
international conference called by the
act of Congress, July 12, 1894 (see Appendix 1).
Physikalisch-Technische Reichsanstalt at CharlotThe value, 1.434 international volts at 15 C, fol
tenburg in October 1905, the Weston TNormal Cell
lowed from the determinations made by Rayleigh
(see later for description) was first proposed as a
and Sidgwick [24], Carhart [25], Kahle [26], and
standard to be used for maintaining the volt and
Glazebrook and Skinner [27] who used current bal
was
officially adopted in 1908 at the London Inter
ances or silver coulometers (the electrochemical
national Conference on Electrical Units and
equivalent of silver having been determined by
Standards [31].
absolute methods) to determine current and B.A.
This London Conference went further and
coils 7 or Siemens mercury columns (or British
adopted provisionally 1.0184 V as the emf of the
Weston Normal Cell at 20 C and recommended for
r B.A. coil refer* lo British Association for Advancement of Science coil of silverplatinum alloy (66.6 percent Ag. 33.4 percent Pt); the B.A. ohm =0.9866 international
the
emf-temperature coefficient of the Weston
ohm. The Siemens ohm standard consisted of a column of chemically pure mercury
1 meter long and 1 square millimeter in cross-section at a temperature of0 *C. Carhart
Normal Cell the formula based on the measurements
and Kahle used Siemens units and then converted their results to British legal ohms.
of Wolff [32]. The London Conference still felt,
The British legal ohm was represented by B.A. coils, the resistances of which were
related to the 1884 "legal ohm," the resistance at 0 "C of a column of mercury 106
however, that further work was needed and recom
centimeters long and one square millimeter in cross-section; one British legal
ohm " 0.9976 international ohm; one British legal ohm 1.06 Siemens ohm.
mended that additional experiments be made.
Table 1. Absolute measurements of the electromotive force of the Clark cell

Experimenters

Temperature Resistance unit

Experi
menters'
Current method value (prior
to reduction)

International values
at observed
temperature

at IS'C

Bolts
volts
1.4344
1.4344
1.434
1.431
1.4342
1.4342
1.4342
1.4342
average
1.4342
rounded
1.434
value
- Carhart [28] later stated that the value 1.434 V was for a mean of two experiments, one at 18 'C and one at 17 "C; a correction for this does not alter the rounded value of 1.434 V
givenabove
the absolute
emf of the Clark
1 B.A.forohm
0.9666 international
ohm.cell.
r Carhart and Kahle converted their values from Siemens units to British legaf units, using the relation 1 British legal ohm = 1.06 Siemens ohm.
'1.118 mg silver deposited per second per ampere (international value of electrochemical equivalent of silver).
' - B.A. volt.
'British legal volt.
'1 British legal ohm =0.9976 international ohm.
Rayleigh and Sidgwick
Carhart
Kahle
Glazebrook and Skinner

Celsius
15.0
18.015.0
15.0

B.A. ohm
Siemens ohm r
Siemens ohmr
B.A. ohm

current balance
silver coulometer 4
silver coulometer *
silver coulometer 4

79-4

volts
1.4539
1.434 '
1.4377 '
1.4537

Accordingly, scientists from England, France, and

Germany met with United States scientists at the
National Bureau of Standards in 1910, and as a
result of their experiments with a large number of
Weston normal cells and silver-coulometer determi
nations adopted 1.0183 V as the emf of the Weston
Normal Cell at 20 C. Values derived from this
were later assumed to be significant to the fifth,
sixth, or seventh decimal as a basis of measurement
[33, 34]. This value (1.0183 V) was still character
ized in "international" units since the basis on
which the new measurements (1910) were made was
the same as it was in 1893; the precision of the
measurements was higher, however.
Values assigned to the reference groups of each
cooperating nation served as the fundamental basis
of all emf measurements from 1911 to 1948. By
1948, after interruptions caused by the two World
Wars and after improvements in techniques, an
accurate determination of the electrical quantities
in cgs em units was achieved, and on January 1,
1948, changes from international to absolute units
were officially made internationally.8 The legal
status of these new units in the United States is
exactly the same as that of the older ones because
the law of 1894 (see Appendix 1) mentions both sets
' Although "international" and "absolute" art frequently used in referring to the
units it is now best to consider these terms of historical interest only. There is only
one kind of volt in the em system of units, and the above terms had significance only
during the period during which efforts were being made to achieve the theoretical unit.
It is deemed advisable to speak, therefore, of only one kind of volt: when the term
"volt" is used, "absolute volt" is implied, i.e.. the volt has a direct relation to the

AMPERE

of units on an equivalent basis. However, in order

to remove the ambiguities of the old act, new
legislation was passed by the Congress in 1950 (see
Appendix 2). The changes [35] for the United
States were:
1 international volt (US) = 1.000330 absolute volts
1 international ohm (US)= 1.000495 absolute ohms
1 international ampere (US) = 0.999835 absolute ampere
1 international coulomb (US) = 0.999835 absolute coulomb
1 international henry (US) = 1.000495 absolute henries
1 international farad (US) = 0.999505 absolute farad
1 international watt (US) = 1.000165 absolute watts
1 international joule (US)= 1.000165 absolute joules
The conversion factors in other countries were
nearly the same as these. The emf of the Weston
Normal Cell at 20 C on the new basis then became
1.01864 V. This value is now of historical interest
only.
The relation of the fundamental units to the
measurement of power and energy in both the new
and former systems of units is shown diagrammati
cal!y in figure 1. The left half represents the funda
mental units and standards maintained by the
National Bureau of Standards, the right half the
units and standards used by the public.
2.3. Maintenance
The unit of electromotive force in the United
States was originally maintained (1897-1906) by
seven Clark cells, the mean emf of which was as
signed a value of 1.4337 V (this value was 0.0003 V
lower than the value recommended in 1893 by the

INT. OHM
Mtrcur- Ohm

MT OHM
W.rt StoodOfd

INT OHM
Stcondor* 3rd

INT AMPERE
Sdvtr Ainptrr.'

INT VOLT
Sotorolod Call

INT. VOLT
untoliiror.il CM

THE FORMER INTERNATIONAL ELECTRICAL UNITS OF

OHM

OHM
Wir. StonOord

1911

TO

1947

OHM
Soco.dor, $10
WATT
IPo.orl

AMPEItE

VOLT
Soturolod Coll

VOLT
Unol.roloO COM /

\
JOULE
(Enoroy )

THE PRESENT ABSOLUTE

ELECTRICAL UNITS

IN USE

SINCE JANUARY 1st 1948

Figure 1. Diagram showing the relation of the present and former systems of electrical units to the
basic mechanical units of length, mass, and time.
80-5

Chicago International Electrical Congress because

specially purified mercurous sulfate had been used
in the preparation of the cells; it was 0.0009 V
higher, however, than the value then recommended
by Germany). In 1906 the standard consisted of
both Clark and Weston Normal cells and after 1908
of Weston cells only; an emf of 1.019126 V was as
signed to the mean of the Weston cells at 20 C
based on a direct comparison with the Clark cells
(on the German basis the emf of the Weston
cells was 1.018226 V which was only slightly lower
than 1.0183 V found in 1910 by the International
Committee which met at the National Bureau of
Standards). After 1911, this latter unit was main
tained in each country until 1948 when "absolute"
units were adopted. The number of cells con
stituting the national standard has varied from time
to time; when a cell shows a steady change from its
previously steady value it is removed from the
group (see later for criterion and procedure).
Today, the National Reference Group of Standard
cells consists of 44 saturated Weston (or cadmium
sulfate) cells, all of which have been made at the
National Bureau of Standards from highly purified
materials and assembled under controlled condi
tions (see later for details). The National Standard
is based on the mean emf of these 44 cells. The
emf of any one cell in the group is equal to the mean
emf, less the average deviation in emf of all 44
cells from the emf of a selected reference cell in
the group, plus the deviation in emf of the individ
ual cell from that of the selected reference cell,
Ec (in volts) = Em-=^ + (Ec- Er)

(3)

where Ec = the emf of an individual cell, Em = the

mean emf of the 44 cells, 2A = the algebraic sum
mation of the differences between the emf of a se
lected reference cell, Er, and the emfs of all of the
other cells in the group. This reference group (of
cells) consists of three parts, of 11 "neutral" cells
made in 1906, 7 "acid" (0.05N) cells made in 1933,
and 26 "acid" (0.026N) cells made in 1948. The
last two groups of cells were added to the reference
group in 1937 and 1955, respectively. The meaning
of the terms "neutral" and "acid" appears later.
This National Reference Group of cells is also sup
plemented by a group of cadmium sulfate cells
made with 98 percent deuterium oxide (heavy
water) [36], the emf of which is about 380 piV lower
than that of cadmium sulfate cells made with nor
mal water. These cells are discussed in Appendix
3, but suffice it to say here that a study of the ratio
of the emfs of cells made with normal and heavy
water offers an auxiliary check on the stability of
the national unit of emf.
Obviously, all of a group of "identical" cells
may increase or decrease in emf with time without
departures from the original assigned mean becom
ing evident. Therefore, an alternative type of
standard cell of approximately the same emf as
* Measurements of large emf differences of the order of magnitude of 0.3 V to 0.S V
with the precision required imposes a problem, though not an insurmountable one.

the Weston cell9 but of different composition

would be most valuable, for if changes in emf with
time in two different systems occurred, they would
not be likely to follow the same pattern. Thus,
studies of the ratio of emfs of two different systems
over a period of years would give valuable insight
into the stability of the standard. It is for this rea
son that the National Standard or Reference Group
was eventually designed to include "neutral" and
"acid" cells 10 and to be supplemented by cadmium
sulfate cells made with heavy wateF. The ratio of
or the difference between the emfs of "neutral" and
"acid" cells is followed in the course of maintaining
the unit of emf.
A cell is removed from the reference group when
its emf has drifted by more than 1.0 fi\ from its
previously steady value. When a cell is removed,
the mean emf of the group is "recaptured" by
reverting in the records to the time the cell had been
added to the reference group, calculating a new
mean for the reduced group (less the cell removed)
at that time, and finally carrying the new mean
forward. In some cases it has entailed going back
as much as 20 years. Since the cells have closely
agreeing emfs the removal of one cell has only a
minor effect on the mean emf of the group; this
effect has generally been below 1 /x-V. The effect
on the emf of the mean of removing one cell obviously
is smaller the larger is the number of cells in the
group (for practical reasons there is a limit to this
number; if too high a number, it might not be
possible to measure the emfs of all the cells within
any one day this would then increase the problems
associated with maintenance).
It is difficult to provide incontrovertible evidence
regarding the long-term stability of the volt main
tained with saturated standard cells. A consider
able body of evidence indicates, however, that it is
very unlikely that the unit of emf preserved with the
National Reference Group of Standard Cells has
changed by any significant amount in the last 53
years. This evidence follows:
(1) In terms of measurements with silver coulometers (or voltameters) and standard resistors prior
to 1948 silver coulometers were used in defining the
international ampere (see Appendix 1). To date,
however, there are no international specifications
that enable the coulometer to be used unambigu
ously as a means of reproducing the international
ampere. Even so, experiments repeated with a
given type of coulometer under the same conditions
after a lapse of a number of years can serve to
establish the same current to high accuracy. From
1910 to 1912, inclusive, [37] a series of experiments
were made at the National Bureau of Standards with
a Smith form of coulometer and the results were
expressed in terms of the emf of the saturated
Weston cell at 20 C; the average of 55 experiments
gave 1.018274 V for the Weston cell at 20 C. In
M Celli thai differ much more in their chemical composition than "neutral" and
"acid" cells or "normal water** and "beary water" cadmium sulfate cells would be
even more desirable for this purpose. However, to dale, the prime requirement of
constancy in emf has not been as well realized in other systems as it has been in cells
of the cadmium sulfate type.

81-6

1931 a series of coulometer experiments were con

ducted at the Physikalisch-Technische Reichsanstalt by representatives from Germany, Great
Britain, and the United States [34, 38, 39]; the
results obtained by the United States with the Smith
form of coulometer gave 1.018273 V for the Weston
cell at 20 C.U Assuming that the unit of resistance
had not changed during this interval of time the
results indicate that the drift in the unit of emf was
not over 0.05 /aV per year.
(2) In terms ofmeasurements with current balances
and standard resistors in 1934 Curtis and Curtis
[40] using a current balance originally used by Rosa,
Dorsey, and Miller [41] found that 1 NBS interna
tional ampere (given by ratio of NBS units of emf
and resistance) was equal to 0.999928 absolute
ampere, whereas the results of Rosa et al. in 1911
gave 0.999926 absolute ampere. Accordingly,
assuming that the unit of resistance had not changed,
these results indicate that any apparent drift in the
unit of emf could not have exceeded 0.1 p.V per
year. Also, since 1938 all checks of the NBS unit
of emf have agreed within the precision ( 7 ppm)
of absolute measurements made with current
balances or electrodynamometers and self or
mutual inductors; these checks indicate that the
unit of emf is constant to 0.2 to 0.3 fiV per year
if the unit of resistance has remained unchanged.
(3) In terms of the gyromagnetic ratio of the
proton in these experiments the precession
frequency, as stated above, is measured in a
magnetic held known in terms of the absolute
ampere and the constant of an accurately made
solenoid. The solenoid current, when compared
with that given by the ratio of the NBS standards of
emf and resistance, has been found to have the
same value to better than 0.1 /u. A during the interval
from January 1960 to March 1963.12 If the unit of
resistance remained constant during this period,
the unit of emf has also remained constant to better
than 0.1 fx\ per year. The construction and prop
erties of standard resistors and standard cells are
radically different; the possibility that the two
systems should drift in such a manner as to maintain
a constant ratio seems remote. The accuracy of
absolute resistance measurements has increased
significantly in recent times. Modern absolute
resistance measurements coupled with reference to
atomic constants should, in the foreseeable future,
supply convincing evidence regarding the stability
of both resistors and cells.
(4) In terms of "neutral" and "acid" cells the
difference between the average emfs of the "neu
tral" and "acid" cells in the National Reference
Group of Standard Cells has increased by 8.5 /iV in
26 years, with the increase being only 1.0 jiV during
the last 10 years [42]. This comparison was made
for 7 "neutral" and 11 "acid" cells, or for cells of
" These results published by Vinal (34) were reported to the fifth decimal and
Vinal said "The value for Washington in terms of B.S. unit is. however, exactly the
same as the result of the bureau's own experiments on the voltameter [37] published
in 1913" (actually 1914). These data have been recalculated, retaining the sixth
decimal, and are given in Appendix 4.
11 The author is indebted to Forest K. Harris and Raymond L. Driscoll for this
information.

this type in the National Reference Group since

1937. Since the emfs of "neutral" and "acid" cells
have remained relatively constant during the last
10 years the earlier drift in their difference may be
attributed to an aging effect exhibited by the "acid"
cells (the "neutral" cells were 26 years old at the
start of the comparison between the "neutral" and
"acid" cells). For the last 10 years the average an
nual change has been 0.10 fi\.
(5) In terms ofinternational comparisons in 1948
the units of emf of the United States (USA) and the
Bureau International des Poids et Mesures (BIPM)
agreed whereas in 1960 the USA unit was 1.9 yxV
smaller than the BIPM unit. Assuming that the
BIPM unit had remained constant during this time
the USA unit has changed at a rate of 0.16 /u.V
per year. Assuming that the BIPM and USA units
changed at the same rates but in opposite directions
the USA unit then changed at a rate of 0.08 p.V
per year. Additional data on international compari
sons are given in section 2.5.
(6) In terms of customers' cells customers'
saturated cells, checked in terms of the National
Standard of emf, show, on the average, an annual
variation of 1.2 /iV per year, but no drifts in emf in
one direction or the other. This relative stability
refers only to cells that are about 3 or more years of
age; saturated standard cells usually exhibit an
aging effect of several microvolts during the first 3
years after their construction.
(7) In terms of newly prepared cells saturated
standard cells freshly made from new materials
usually agree with old or aged cells within 5 /xV.
Of course, in this type of comparison the cells must
be made with amalgams of the same percentage of
cadmium and with cadmium sulfate solutions of
the same acidity with respect to sulfuric acid. The
point here is that saturated cadmium cells, made at
different times to essentially the same specifica
tion, are highly reproducible. Cells made in the
past cannot have drifted seriously in emf if similar
cells of recent construction agree closely in emf with
them.
(8) In terms of the relative emfs of cells within a
group of cells the differences in emf between the
cells in the National Reference Group of Standard
Cells have remained remarkably constant for dec
ades. Although, as stated above, all "identical"
cells may increase or decrease in emf without evi
dent departures from their mean emf, constancy
in the difference between the emf of individual cells
in a group of cells with time nevertheless, in com
bination with items (1), (2), and (3) above, increases
confidence in the constancy of the mean emf of a
group of cells.
Although the above remarks apply to the stability
of standard cells in the National Reference Group
of Standard Cells they should also apply to any
standard cell of the saturated type providing it is a
quality cell properly maintained. Cells made with
impure materials or poorly assembled will invaria
bly show much less stability in emf. In section 2.5
the need for adjustments in the assigned values to

82-7

standard cells, although infrequent, is discussed.

The need for these adjustments is not clear. It
may be that the cells were of poor quality, were
affected by transport, or had not come to equilibrium
after temperature changes which are involved in
international comparisons.
The standard cells of the National Reference
Group are housed in slowly stirred oil baths main
tained at 28 C under diffuse light in an air-condi
tioned room maintained at 25 C 1 C; the relative
humidity automatically remains below 50 percent.
The temperature of the baths is maintained at 28 C
within 0.01 C on a long-term basis and within 0.001
C during measurements using a Gouy controller [43].
In a Gouy controller, a steel piano wire extending
into the mercury of the mercury-toluene regulator is
connected to a wheel which revolves at a slow rate
whereby the wire is made to periodically make and
break contact with the mercury. The design of
the bath and its temperature control are described
in Appendix 5. A special mineral oil, having the
characteristics listed in Appendix 6, is used as
fluid in the constant-temperature baths. Tests
made to determine the temperature of the oil at
various locations within the baths indicated a uni
form temperature within 0.001 C, i.e., no hot or
cold spots prevail within the bath or in the vicinity
of the standard cells. This design follows closely
that described many years ago by Wolff and
Waters [44]. A bath for use with saturated stand
ard cells has also been described recently by P. H.
Lowrie, Jr. [45].
The cells are supported about 2 in. below the
surface of the oil on seasoned mahogany strips
about 16 mm wide (slightly smaller than the distance
between the limbs of NBS H-shaped cells) and about
10 mm thick and either 24 or 48 cm long, with
grooves into which the cross-arms of the cells fit
snugly. Some racks carry 18 cells equally spaced,
except for a somewhat wider space at the middle of
the rack where the rack is supported; others carry
9 cells. Hard rubber, bakelite, or lucite strips

Figure 2. Oil baths used to house the National Reference

Group of Standard Cells.

(widened U-shaped), about 40 mm long, 10 mm wide,

and 10 mm thick in the center and 13 mm at each
end, are mounted between cells across the underedge of the mahogany strip. In the top of each
strip, at each end, is inserted a short copper rod
provided with a pair of holes, one 1 mm and the
other 2 mm in diameter and both about 9 mm in
depth; these holes serve as mercury cups, one for a
cell terminal and the other for external connections.
These two copper cups are spaced at the same
distance apart on each strip, so that any cell can be
put in the electric circuit by a stabber consisting of
a pair of stiff copper wires mounted in a lucite block
(copper wires are amalgamated on the tips). This
is a copper-copper connection through the mercury
and under the oil. From this point on all contacts
are copper-copper to avoid thermal emfs. The
other end of the stabber connection goes to a post
position. There are two post positions, one for the
Reference Cell and one for the Unknown Cell.
The leads of the cells go from the posts through
conduits to another post at the emf-measuring
instrument (see later), where the two cells are
placed in series opposition by joining the negatives
and the difference is measured.
A photograph showing the oil baths that house
the National Reference Group of Cells is shown in
figure 2. Other baths of intermediate size are also
available for housing cells on test. The tempera
ture of the baths is measured with a platinumresistance thermometer and a Mueller bridge,
having a sensitivity of 0.0001 ohm, corresponding
to 0.001 C.
2.4. Dissemination
Comparisons of the emf of standard cells, such
as may be required in the dissemination of the unit
of emf, i.e., the transfer of the unit from a standard
izing laboratory, must or should be carried out by
procedures of the highest precision involving a
minimum of uncertainty. In this case, comparisons
are made between standard cells, the emf of which
have been determined in absolute units and other
standard cells, the emf of which have not been so
determined. The former may be called reference
cells and the latter unknown cells. These intercomparisons may be made with high accuracy with
potentiometers by the opposition method (prefer
ably called the difference method). In this method
the two cells (reference and unknown) are connected
in series with their emfs in opposition and the
difference in emf between the two is measured
with a potentiometer using a galvanometer of high
sensitivity. Because the difference in emf between
the two cells is small, only a moderate percentage
accuracy in its determination is required to give the
emf of the unknown cell accurately in terms of the
known reference cell. If the difference between
the two cells were 100 p.\ the accuracy of the
measurements need be only 1 percent to give the
difference to 1 p. V or only 0.1 percent to give an
accuracy of 0.1 p.\.
This method is not direct reading, however, and

83-8

the operator must determine which of the two cells

has the higher emf. The greatest source of error
in measurements by the difference method is the
presence in the circuit of spurious emfs, such as
those from thermoelectric action. At the National
Bureau of Standards a special comparator, designed
by Brooks [46], is used. This comparator is direct
reading and compensates for parasitic thermal emfs.
The reference and unknown cells are connected in
series-opposition in series with a galvanometer and
an auxiliary source of a few microvolts and adjust
ments are made in the auxiliary source until the
galvanometer indicates a null balance. The
D'Arsonval galvanometer (see Gi, figure 3) used with
the comparator has a free period of 8 sec, an external
damping resistance of 1200 ohms, and a sensitivity
of 5 mm per microvolt at one meter lamp and scale
distance when operated slightly underdamped.
With the Brooks comparator, differences in emf
as large as 2100 fiV may be read with a precision
of 0.1 fi\. The Brooks comparator contains an
opposing circuit for the detection and compensa
tion of thermal emfs in the galvanometer circuit,
contains no sliding parts in the main circuit, and
has a "read-out" device by which the measured
difference in the emf is added algebraically to the
emf of the reference cell thereby giving directly
the entire emf of the unknown cell.
A circuit illustrating the principle of the Brooks
comparator is shown diagrammatically in figure 3
and the exterior and interior of the comparator are
shown in figures 4 and 5, respectively. The upper
most of the three sections shown in figure 3 relates
to the adjustment of the current in the comparator
to the proper value. A current h from battery B\
(No. 6 dry cell) is regulated by the rheostat R\ to a
definite value in the usual manner by reference to
an auxiliary standard cell. All parts with exception
of the battery are housed in the comparator. In
the middle section, the substitution resistances con
nected to tap points numbered 3 to 16 are selected
by the central dial as needed. Each step is 100 /iV.
It is necessary to interpolate between successive
steps to complete the process of measuring the emf
of the unknown cell, X (lower section). This is
done automatically by the circuits attached to
points 17 and 18. Bz and B3 are No. 6 dry cells,
also. The milliammeter A3 measures the current
/:) when an exact balance is indicated by the galva
nometer Ci as a result of varying the resistance R3.
The milliammeter A 3 is calibrated to make its scale
read directly in microvolts. A similar circuit at
tached at 1 and 2 performs a like function for the
reference cell N (lower section) whose emf value
must be set exactly if the readings for the unknown
cell X are to be correctly determined. The lowest
section of figure 3 shows the circuit wherein the un
known, X, and the reference cell, N, are connected
in series opposition. This circuit also contains the
galvanometer, Ci, of sensitivity, given above.
Thermal emf compensation is achieved by a
simple copper slide wire connected in series with

the main galvanometer, G\. About 15 fiA from a

No. 6 dry cell (not shown in the figure) enters this
slide wire at its central point and leaves by a slider.
If the slider is set at the central point of the copper
wire there is no current in the copper wire and,
therefore, no potential drop. By setting the slider
away from the central point, a small adjustable
potential drop may be introduced into the galva-

FlCURE 3. Diagrammatic plan of circuits illustrating the

principle of the potentiometric part of a Brooks standard-cell
comparator.

Figure 4. Photograph of a Brooks comparator (top view).

FlCURE 5. Photograph of the interior of a Brooks comparator.

84-9

nometer circuit to neutralize any parasitic emf in

the galvanometer and the wires connecting it to the
comparator. By depressing a "shunt" key any ap
preciable parasitic emf can be detected; if present
it will maintain a corresponding deflection of the
galvanometer and when the "shunt" key is de
pressed the galvanometer coil will assume its opencircuit zero position. To neutralize this undesired
emf the slider is manipulated until no motion of
the galvanometer coil ensues when the "shunt" key
is depressed. For additional details the original
paper by Brooks [46] should be consulted.
Although compensation for thermal effects is
provided in the instrument, thermal effects may also
arise at the cell terminals. These are eliminated
by making connections to the cells through mercury
cups immersed two inches below the surface of oil
maintained at the same temperature as the cells,
as was mentioned above under maintenance (sec
tion 2.3). In those cases (for example, unsaturated
cells or saturated cells in air boxes; see later) where
the terminals of the unknown cell cannot be im
mersed in oil, thermal effects are kept at a minimum
by using like metal connections in a constanttemperature room. Other types of standard cell
comparators have been described by Miller [47],
Vincent [48], and by Spinks and Hermach [49].
The comparator (portable potentiometer) of Spinks
and Hermach also includes two saturated and two
unsaturated standard cells in a temperaturecontrolled enclosure.
In the dissemination of the unit of emf, i.e., the
comparison of unknown and reference cells, working
groups of standard cells are used instead of the
National Reference Group of Standard Cells. Of
course, the emfs of the cells within these working
groups are known in terms of the National Reference
Group. These comparisons are made with a preci
sion of 0.6 fi\. One working group is maintained
in the Standard Cell Laboratory in Washington,
D.C., the other at the NBS Boulder Laboratories in
Boulder, Colo. [50]. A total of 10 comparisons of
the emf of the unknown (customer) cells with the
working groups is made, one per day, over a period
of ten days, at a specified temperature for saturated
cells and at an ambient temperature of 25 C (at
Washington) and of 23 C (at Boulder) for unsatu
rated cells.13 The best saturated cells are cali
brated with an uncertainty of 0.0001 percent; the
best unsaturated cells with an uncertainty of 0.005
percent. The results are transmitted to customers
in Reports of Calibration.
2.5. International Comparisons
Since 1932 the units of emf of various nations have
been intercompared at specified intervals at the
Bureau International des Poids et Mesures (BIPM),
Sevres, France. At the present time this interval
is 3 years. For many years the comparisons were
rather sporadic, and they were interrupted by the
11 The National Bureau of Standards provides emf calibrations of unsaturated
standard ceils only for public utilities and others having operations of such a nature as
to require calibrations by the National Bureau of Standards.

.,ool_l1910 1915
I
I
1
i
I
1
1
I
1920
1925
1930
1935
1940
1945
1950
Figure 6. International comparisons of the unit of electromo
tive force in international volts prior to 1 948.
two World Wars (see fig. 6). Prior to 1948 the
intercomparisons were made in "international volts"
and since 1948 in "absolute volts" (see figs. 6 and 7).
These intercomparisons are effected by standard
cells maintained by the participating countries and
by the International Bureau and are conducted at
20 C. As a rule each country submits 4 to 10 cells
to BIPM for the intercomparisons; at the present
time the cells are carried to BIPM by messenger.
The first comparison involved the measurements
at the National Bureau of Standards in 1910, at
which time the units of the participating countries,
France, Germany, Great Britain, and the United
States were identical (see fig. 6). From 1911 to
1932 the major intercomparisons were between the
units as maintained in the United States and in
Great Britain. In preparing figure 6 it was assumed
that the mean value for these two nations remained
constant and this mean is made the axis of abscissas.
A few comparisons were also made of the United
States unit with those of France, Germany, Japan,
and Russia and are shown in figure 6 in terms of the
mean unit of the United States and Great Britain.
In 1931 it was apparent that there was an increasing
discrepancy between the unit as maintained in
Germany at the Physikalisch-Technische Reichsanstalt (or PTR) in Charlottenburg and those
maintained in the United States at the National
Bureau of Standards (or NBS) in Washington, D.C.,
and in Great Britain at the National Physical
Laboratory (or NPL) in Teddington. Consequently,
at the invitation of the PTR arrangements were
made to have an international committee carry out
experiments with the silver coulometer at PTR and
thus reestablish the international volt. Accord
ingly, P. Vigoureux of NPL and G. W. Vinal of NBS
took silver coulometers to the PTR together with
standard cells and standard resistors, the values of
which had been carefully determined at their
respective laboratories. At the PTR, these stand
ards were compared with those of the PTR and an
extensive series of experiments with different types

85- 10

of silver coulometers were made. As a result of

these measurements, subsequently published [34,
38, 39], Germany increased its unit by 82 ppm (see
Appendix 4 for United States data).
As a result of extensive work on silver coulometers
by Kolossov [51], Russia changed its unit in 1934 by
87 ppm.

1950

1955

I960

Figure 7. International comparisons of the unit of electromo

tive force in absolute volts since 1948.
International comparisons at BIPM began in 1932
[52] and after 1934 the data in figure 6 are plotted
on the assumption that the "Mean International
Volt" as defined in 1935 by the Consultative Com
mittee on Electricity [53] has remained constant.
In December 1934 the six countries listed in figure
6 submitted cells to BIPM. France, on finding its
unit out of line from the others, increased its unit
by 76 ppm in 1935. The Consultative Committee
on Electricity suggested that the mean unit of the
five remaining countries be adopted by all thus as
suring immediate unification. However, only
Russia accepted this suggestion and reduced its
unit by 13 ppm. In 1937, 1939, and 1945-48, BIPM
carried forward its mean value obtained in 1935 for
the six countries after correcting for the Russian
3. Early Sta
Many types of galvanic cells have been proposed
as standards of emf and some of these have ap
peared in a variety of forms. Although Faraday
and others of his time used various galvanic cells,
such as Grove and Bunsen cells, in their investiga
tions, the Daniell cell [54] represented by
(-) Zn(s) | ZnS04, H,SO(aq) || CuS04(aq) | Cu(s) (+)
was the first cell seriously used as a standard of
emf, as it exhibited less gassing than its predeces
sors. Here a single vertical line is used to indicate
the interface of two distinct phases, a double
vertical line to indicate a liquid-liquid junction,
s = solid, and aq = aqueous solution. However,

adjustment; this adjustment made the BIPM unit

2.2 ppm lower than its 1935 unit; the data in figure 6
are so reported. For the last comparison in inter
national volts the National Physical Laboratory sub
mitted cells in 1945 and the comparison with BIPM
was completed in 1946; the other countries sub
mitted cells in 1948 and the comparisons with
BIPM were done the same year. Russia reported
its values in absolute units using the international
relation: 1 mean international volt = 1.00034 ab
solute volts. In 1939 the Russian unit had been
23.1 fi\ below the mean international unit. For
comparisons with other countries and assuming
that this difference still applied, BIPM converted
the Russian absolute values to international values
using a conversion factor of 1.000317.
On January 1, 1948, "absolute" units were
adopted but international comparisons on this basis
were not made until 1950. Intercomparisons in the
"absolute" units, obtained since 1950, are given in
figure 7 (note that the scale in figure 7 is much larger
than in figure 6); Australia and Canada were new
additions. In 1950 the German value was from East
Germany while in 1953, 1955, and 1957 both East
and West Germany took part in the international
comparisons; since 1957 the German value is for
West Germany. The data shown in figure 7 are the
deviations from the BIPM mean value. In 1950
Russia had apparently corrected their "absolute"
values of 1948 by the average deviations of the
various countries, Russia excepted, from the BIPM
unit, thus, ignoring the fact that their unit had been
23.1 fi\ below the mean international unit in 1939:
this procedure, then placed their unit in 1950 above
the BIPM unit by approximately this magnitude.
In 1955 Russia made an adjustment in its unit of
about 13 ^iV. In 1960 the spread between the 8
countries was 10.2 p.V with the units of Australia.
Great Britain, and Russia being high. The spread
between the other 5 countries was 3.3 fi\. In
1960, values for Italian cells were included with the
values listed with the French cells; France had first
made a direct comparison of the Italian cells with
their cells at the Laboratoire Central d'Electricite,
Fontenay-aux-Roses, France.
dard Cells
this type of cell did not exhibit a long-term stability
in emf. Being a two-fluid cell, the solutions dif
fused into each other causing local action at the
electrodes and a steady decrease in emf. Some
what better results were obtained when saturated
solutions, no acid, and amalgamated zinc were used,
but, even so, the cells did not exhibit the per
manency required in a standard. Nevertheless,
for over 35 years, Daniell cells were used as a
standard of emf. Absolute electrical measure
ments gave 1.07 to 1.14 V for the emf of freshly
prepared cells; the actual value depended on the
concentration and acidity of the solutions used.
Absolute measurements were not needed to show
the lack of constancy in the emf of the cells; this

86-11

behavior was evident from comparing new cells

with old ones. The cell reaction is
Zn(s) + CuSCMaq) - ZnS04(aq) + Cu(s).

(4)

The Daniell cell has an emf-temperature coefficient

of about +34 /uV/deg C at room temperature (18 to
30 C).
Daniell cells were made in various forms. The
original form consisted of a glass jar containing a
porous pot of unglazed earthenware in which a
zinc plate or rod was placed. Outside and around
the pot a cylindrical sheet of copper was placed.
The outer jar was filled with a concentrated solu
tion of copper sulfate and the porous pot with dilute
sulfuric acid, or zinc sulfate, or zinc sulfate acidified
with sulfuric acid [55]. Important modifications
were the gravity [56] and Fleming-Thomson cells
[57]. In the gravity cell the less dense solution
of zinc sulfate was placed over the more dense so
lution of copper sulfate so that the porous cup was
eliminated. Zinc in circular or crow-foot form and
copper in leaf form were used as electrodes in their
respective salt solutions. In practice the cell was
kept on closed circuit to curtail the mixing of the
two solutions. In the Fleming-Thomson cell a
U-tube was used containing a solution of copper sul
fate in one arm and a solution of zinc sulfate of
the same density in the other arm. Rods of copper
and zinc were supported in their respective salt
solutions.
In 1872 Latimer Clark [23] proposed a cell which
had a profound effect on work pertaining to the
electrical units. His cell, represented by
(-)Zn(i) | ZnSO* 7HtO(c) | ZnSCMsat aq) | ZnSO,
7H,0(c) | HglS04(s) | Hg(/K+)
where c = crystals, s = solid, / = liquid, and sat
aq = saturated aqueous solution, was a one-fluid
cell free of a liquid-liquid junction and exhibited
relatively good stability in emf. Because of the
historical importance of the Clark cell (see sec. 2)
it is discussed in more detail in section 4. Owing
to the success that was obtained with the Clark cell
several other one-fluid cells were proposed as
standards, the more important being those sug
gested by De la Rue [58], Helmholtz [59], Gouy [60],
and Weston [30]. Of these four the first three ex
hibited larger drifts in emf and had higher internal
resistances than Clark cells, while the last one had
many advantages over the Clark cell. As a result,
Weston cells are now used almost exclusively as
standards of emf. Weston cells had an emftemperature coefficient about l/30th of that of the
Clark cell, better emf stability, and an emf closer
to unity than the Clark cell.
Early types of standard cells are listed in abbre
viated form in table 2 together with their nominal
emf and emf-temperature coefficient. As is evident
from the table the Weston cell is similar to the
Clark cell except that Cd replaces Zn. In all of
. 315-411 O - 68 - 7

these cells, the original Clark cell excepted, amal

gamated anodes were generally recommended and
H-shaped containers (see below) were frequently
used. In general, saturated solutions were used
and excess crystals of the stable salt added to the
cells at the electrode surfaces. Weston [61] and
Carhart [62] proposed the use of an unsaturated
solution of zinc sulfate in the Clark cell in order to
decrease its emf-temperature coefficient. The
original Weston patents also covered the unsatu
rated form of Weston cell.
Table 2. Early types of standard celts
(represented in abbreviated form)
Name

Dale

System

Approximate
emf, volls

Approximate liEltll,
volt per deg i'.

Daniell
Clark
De la Rue
Helmholtz
Gouy
Weston

1836
1872
1878
1882
1888
1892

Zn|ZnSO,|CuSO.|Cu
Zn|ZnSO.|Hg,SO.|Hg
Zn|ZnCl,|AgCl|Ag
Zn|ZnCI,|H||CI|Hg
Zn|ZnSO,|HgO|Hg
Cd|CdSO<|Hg,SO.|Hg

1.04-1.14 In)'
1.433 (IS "C)
1.03 (nl
1.000 (IS *C)
1.3V0 (12 *C)
1.01864 (20 "O

+ 0.000034 (n)
-0.0012(15 *C1
-0.001 (n)
+ 0.000087 (nl
-0.0014 In)
-0.000041 (20 "CI

1n represents nominal value at room temperature (about 18 to 30 *C).

Somewhat later Bronsted [63] proposed the cell:
()Pb, HgtOIKzSO^PbSO^cJlKsSCMsat aq)|Hg2S04(i)|Hg(/K+)
as a standard, stating that at 22 C it had an average
emf of 1.0481 V and an emf-temperature coefficient
of 0.0001 V/deg C. Henderson and Stegeman
[64], however, stated that the Bronsted cell did not
exhibit steady emfs with time and that better re
sults were obtained when Na2S04 was used as the
electrolyte. However, their cell at 25 C had a
lower emf of 0.96463 V and a higher emf-tempera
ture coefficient of + 0.000174 V/deg C. These cells
have not been used probably because of their high
emf-temperature coefficients. Vosburgh, Guagenty,
and Clayton [65] proposed the cell:
(-)Cd, Bi.
Hg(3p)

C6SO.-8l3H,CMc),
sat aq sol of the
two salts in 0.1 M Hg,SO.U) Hg(/X+)
CdSO.Na,SO.-2H.O(c) acetic acid or in
0.01 M H,SO,

which is a modified Weston cell in which a 3-phase

(3p) amalgam (8.9 percent Cd, 11.1 percent Bi, 80
percent Hg) and a 0.1 M acetic acid or 0.01 M H2SO4
solution saturated with the two salts, as indicated,
are used. The cells were sealed with a nitrogen
atmosphere inside. At 25 C they reported an
average emf of 1.0184 V and an emf-temperature
coefficient of +0.000013 V/deg C which is about 2
to 10 times that of unsaturated Weston cells but
less than one-third of that of the saturated Weston
cell at 20 to 30 C but of opposite sign. Data avail
able on this cell indicate that it does not have the
long-term stability exhibited by saturated Weston
cells, but compares most favorably with the un
saturated type [66].

87-12

4. The Clark Cell

As stated above, the original Clark cells con
sisted of a zinc anode and a mercury-mercurous
sulfate cathode in a saturated solution of zinc sul
fate containing crystals of ZnSCh 7H2O. Clark
used a very simple construction as shown by (a) in
figure 8. Various modifications of this general con
struction were used in commercial cells; one such
modification being the British Board of Trade cell
illustrated by (b) in figure 8. In all of these the
zinc anode was in contact with the paste consisting
of mercurous sulfate, a saturated solution of zinc
sulfate, and crystals of ZnSC4 7H2O.
Clark determined the emf of his cell in absolute
units with (a) a sine galvanometer and (b) an electrodynamometer and a B. A. resistor (see footnote 7,
p. 4). With these he obtained 1.45735 V and
1.45761 V, respectively, at 15.5 C, the mean of
which he rounded to 1.457 V as the absolute emf
of his cell. Clark's cell, although much better as a
standard than the two-fluid cells, did not show the
permanency in emf hoped for. The cell tended to
gas at the zinc anode and the emf showed large
variations owing mainly to concentration gradients

that developed, during slight changes in ambient

temperatures, within the compact paste.
In 1884 Rayleigh and Sidgwick [24] introduced
two modifications in the Clark cell that resulted in
substantial improvements. They substituted zinc
amalgam (percentage not stated) M for zinc and used
an H-shaped container in which the zinc anode could
be kept entirely under solution and out of contact
with the mercurous sulfate paste. Their design is
illustrated by (c) in figure 8. The zinc amalgam
reduced the rate of gassing at the anode and by
being under solution and out of contact with the
mercurous sulfate paste the zinc anode exhibited
more nearly constant potentials. Rayleigh and
Sidgwick obtained 1.453 V for the absolute emf of
their cells at 15 C using a current balance (now
known as a Rayleigh balance) and a B.A. resistor.
n y 2f. I S,^V\C\ Pr"cnled ,h'r PP" in June 1884; in the paper published
in Dec. 1884 they added ihe note "Some H-cells have been set up Mr. Thxelfall with
amalgamator knowncomposilion. varying from 1/32 line to 1/5 zinc by weight. The
duration of the teat has as yet been scarcely adequate, but it appears-a that
that the 1
quantity of zinc is sufficient." In general, Clark cells were made, at a later date
with 7 to 10 percent amalg,

Zn
sol

sol

sj^j
-w Hg
(a)

(b)

(c)

sol
S
p
Hg

0-Hg
()

(f )
Figure 8. Various forms of standard cell containers.
(aI Original Clark cell
Zn unc
(b) Board of Trade (British) Clark cell
Hg mercury
<c) Rayleigh-Sidgwick H-shaped Weston cell
P mercurous sulfate paste
(d) Sealed H-shaped cell
C cork
(e) Hulett H shaped cell
C
glue
(f) Modified Hulett container
sol solution
(g) Wright-Thompson container
A
amalgam
(h) Cooper container
S salt crystals
88-13

(d)

These cells tended to leak in time, and Callendar

and Barnes [67] recommended their hermetic seal
ing, (d) in figure 8. Other types of cell containers,
also used for Weston cells, have been proposed and
are illustrated in figure 8. Hulett [68] proposed a
shorter crossarm (e) and today some cells are made
with no crossarm but with a partition at the base of
a single tube (f). Wright and Thompson [69] pro
posed the inverted-Y form, hermetically sealed, and
Kahle [70] the same form but with a cork or groundglass stopper (g); these are more difficult to fill.
Cooper [71] proposed a modified Kahle type (h)
which required no support in a thermostaticallycontrolled bath; his form could rest on a flat surface
without support. The Cooper cell could also be
used in water as well as in non-conducting oil
because the cell terminals were not exposed but
protruded above the bath fluid. Today, the H-shaped
container, hermetically sealed, is used most widely;
some single-tube types, either like (0 or in a modified
form of the original design (a) are also made.
The reaction in the saturated Clark cell made
with amalgamated anodes is:

500,

(6)

for the dependence of the emf on temperature. At

15 C and 25 C this relation gives - 0.00120 V/deg C
and -0.001076 V/deg C, respectively, for dE/dt.
Kahle [74] and Jaeger and Kahle [75] who also
studied the temperature dependence of Clark cells
obtained results in substantial agreement with
eq (6).

Table 3. Composition of the liquid phases of 10 percent zinc and

cadmium amalgams
Percentage
Temperature,
c
of line

Percentage
of cadmium
2.50
3.70
5.00
6.40
7.80

1JS
1.60
1.99
2.40
2.90

0
10
20
30
40

=
m I (ZnS<V7H ,OMc) + 2Hg(0 +(*- l)Zn, yHg(2p) (5)

, = ,s.-0.0012(/- 15 )-0.0000062(t- 15 Y

40 50 60
Percent Zinc
FIGURE 9. Phase diagram for the zinc-mercury system.

*Zn,(yHgX2p) +HgjS04<s) + (ZnS(VmH2OXsat aq)

m7

where x = moles of Zn associated with y moles of

Hg in the amalgam, 2p = 2-phase, and m is the num
ber of moles of water associated with 1 mole of
ZnS04 in the saturated solution. Ten-percent
amalgams are most commonly used; they are of two
phases, solid and liquid, with the solid phase being
pure zinc (see fig. 9) [72]. The composition of the
liquid phase of a 10 percent zinc amalgam at various
temperatures is given in column 2 of table 3.
The solubility of zinc sulfate in water changes
considerably with temperature. Above 39 C the
stable sulfate is ZnS04'6H20; below 39 C it is
ZnS04-7H20. The solubility of ZnSCv7H20 in
water from 0 to 39 C is given in table 4. In satu
rated solution zinc sulfate hydrolyzes to give a
solution containing 0.004 N sulfuric acid and having
a pH of 3.35 at 25 C [73]. This concentration of
acid is sufficient to prevent the hydrolysis of the
mercurous sulfate used in the positive electrode (see
later).
The emf of the Clark cell decreases as the
temperature is increased; the decrease is about 0.1
percent per degree C. For the range 0 to 28 C
Callendar and Barnes [67] gave, in volts:

Table 4. Solubility o/ZnS04-7H20 in water

Temperature
ZnSO. in
c
lOOOg H.O,
0
5
10
15
20
25
30
35
39-

grams
419
446
477
509
542
579
620
663
701

temperature, ZnSO,'7HiO-ZnSO<'6H>0.
The Clark cell has three advantages over the
Weston cell, namely, (1) zinc sulfate hydrolyzes to
produce sufficient sulfuric acid to prevent the hy
drolysis of the mercurous sulfate paste; it, thereby
supplies its own buffering action, (2) the solid phase
of the amalgam is a single component (zinc) whereas
in the Weston cell the solid phase is a solid solution
of cadmium and mercury; the emf of the Clark cell
is less dependent, therefore, on the composition
of the amalgam than is the Weston cell, and (3) the
Clark cell shows less emf-temperature hysteresis
than the Weston cell. However, the disadvantages
of the Clark cell over the Weston cell well outweigh
its advantages. These are: (1) The Clark cell
tends to gas at the amalgam electrode. This gas,
as it forms slowly over the surface of the amalgam,
may dislodge crystals of ZnS04 7H20 with adhering
solution from over the amalgam, and may eventually
cause an open-circuit in the electric circuit unless a

89-H

constriction, as discussed later, is used in the con

Weston cells because the change in solubility of
tainer at the level of the crystals. (2) Zinc amalgam
cadmium sulfate with temperature is much less
than that of zinc sulfate. (5) The Clark cell has
slowly alloys with the platinum lead at the base of
an emf which is considerably higher than unity
the amalgam electrode. Since this alloy occupies
which, although not necessarily a disadvantage,
more space than the platinum, strains are produced,
makes its use as a standard less convenient than a
and breakage of the glass container follows. (3)
Weston cell.
The Clark cell has an emf-temperature coefficient
The emfs of saturated Clark cells are decreased
which is about 30 times that of the Weston cell.
by addition of sulfuric acid to the electrolyte.
(4) Owing to the relatively large change in the solu
Some data based on measurements of Hulett [76]
bility of zinc sulfate with changes in temperature
crystals of ZnSO* 7H2O tend to pass from the
are given in table 5.
positive to the negative limb of the cell. During a
TABLE 5. Effect of sulfuric acid on the emf ofClark cells at 25 "C
temperature rise the volume of electrolyte over the
positive electrode becomes saturated more quickly
Normality of >ulfuric acid'
because it is usually a smaller volume than that over
Change in emf.
the negative electrode. As a result a concentration
Before
saturation
AfterZnSO,
saturation
with

7H.0
with
ZnSO.

7H.O
difference between the limbs results and diffusion
volt
of electrolyte to the negative limb takes place until
0.2024
0.1074
-0.00028
the concentration difference is dissipated. This
1.012
0.5400
-0.00233
2.024
1.1488
-0.00448
phenomenon can be prevented by making the vol
ume above the negative and positive electrodes
" normality of acid in a saturated solution of pure zinc chloride at 25 "C is
equal. A similar phenomenon does not occur in
/V[73L
5. The Weston (or Cadmium Sulfate) Cell
5.1. General
The Weston (or cadmium sulfate) cell is, for all
intents and purposes, the only electrochemical
system used today as a standard cell. Accordingly,
the words "standard cell" when used today in
variably mean cadmium sulfate cells, and the re
mainder of this Monograph, appendices excepted,
relates to this type of cell. The cell is made in
two general types, saturated and unsaturated where
these terms refer to the state of the electrolyte.
The saturated type is the precision cell used in the
maintenance of the unit of emf. It may be made
in a highly reproducible form and exhibits a con
stant emf for long periods of time. However, for
high precision it must be maintained at a constant
temperature owing to its relatively high emf-tem
perature coefficient. Most saturated cells must
be hand carried, although some recent types have
been made shippable by locking the electrodes
in place with inert and porous septa. The un
saturated cell is less stable than the saturated type
in that its emf decreases slowly with time and is
regarded as a reference of d-c voltage known within
0.005 percent. It is usually made in a shippable
form with a septum over each electrode. It has
a low emf-temperature coefficient and accordingly
is used widely in ambient temperatures as an emf
reference where 0.005 percent (0.05 mV) accuracy
suffices.
The saturated cell is also known as the Weston
Normal Cell (or Element). It consists of a cadmium
amalgam anode and a mercury-mercurous sulfate
cathode in a saturated solution of cadmium sulfate
with crystals of CdSO, 8/3H20 over the surface
of both electrodes. This cell may be represented
by:
(-)Cd, Hg(2p)|CdS04-8/3|H,0(c)|CdS04<8at aq)|
CdS(V8/3 H,0(c)|Hg!S04(5)|HgaX-l-)

where the symbols have the same significance as

given above. The cell reaction is:
xCd, yHg(2p)-r-Hg>SO,(5)+ 8/^.. (CdSO, mHjOXsat aq)
171 O/O
= ^ (CdS04 8/3H.CX.C) + 2Hg(/) + (*- l)Cd, rHg(2p). (7)
where x moles of Cd are associated with y moles
of Hg in the amalgam and m is the number of moles
of water associated with 1 mole of CdSC>4 in the
saturated solution. At the end of the reaction the
amalgam may be two phases (liquid and solid) or
may be a liquid phase only, depending on the ex
tent of the reaction and the relative amounts of
amalgam and mercurous sulfate used in preparing
the cell. When not discharged, as is normal when
the cell is used as an emf standard, the amalgam
remains in two phases. Cells are usually made
with 10 or 12V2 percent amalgams (see later).
The unsaturated cell differs from the saturated
type only in that an unsaturated solution of cad
mium sulfate and no crystals of CdSC4 8/3H2O
are used. It is customary to use a solution that is
saturated at 3 or 4 C, the temperature range at
which the salt exhibits a minimum solubility; the
solution is then unsaturated at higher temperatures.
The emf of the unsaturated cell at ambient room
temperature is about 0.05 percent higher than that
of the saturated type. The cell reaction is simply:
*Cd, yHg(2p) + HgzSO 4(5) = CdS04(aq)
+ 2Hg(Z) + (x - l)Cd, yHg(2p) (8)
unless the reaction is continued until crystals of
CdS04 8/3H2O are formed; then the reaction is
the same as in the saturated cell. Again, when
the cell is not discharged, as is normal when the
cell is used as an emf standard, the solution remains
unsaturated and the amalgam in two phases.

90-15

Both saturated and unsaturated standard cells

have been made as "neutral" or as "acid" cells; the
saturated "neutral" type is also known as the Weston
Normal Cell. These terms refer to the degree of
acidity of the electrolyte with respect to sulfuric acid
in the cell. If an aqueous solution of pure cadmium
sulfate, to which no sulfuric acid is added, is used
in the preparation of the cell, the cell is called a
"neutral" type, even though the pH of a saturated
aqueous solution of cadmium sulfate, owing to
hydrolysis, is 4.00 at 25 C [73]. If sulfuric acid is
added in sufficient amount to make the acidity 0.03
N to 0.1 N, the cell is called an "acid" type. The
purpose of adding the acid to the electrolyte is to
prevent hydrolysis of the mercurous sulfate used in
the cell; more details are given later. Today, most,
if not all, cells are made of the "acid" type.
5.2. Preparation and Properties of Materials
The procedures employed at the National Bureau
of Standards for the preparation and purification
of materials for use in the construction of standard
cells are described in the next few sections. In
addition some properties of the materials, as they
relate to standard cells, are given. Only four ma
terials are required. These are mercury, cadmium,
sulfuric acid, and water, all of which may be puri
fied by distillation. To these four materials may
be added a fifth, CdS04 - 8/3 H2O, to avoid preparing
it from cadmium and sulfuric acid; it may be puri
fied by repeated crystallizations from conductivity
water.
Mercury of good grade, after washing in dilute
nitric acid, then in distilled water, is dried and
then distilled in a Hulett [77] still in a stream of
dry air. In this method, the mercury is distilled
near 200 C in a partial vacuum, air being drawn
through the condenser at a rate to maintain the
air pressure at 25 mm, corresponding to an oxygen
partial pressure of 5 mm. Any metallic vapor,
except mercury, is oxidized and collects as a scum
on the distillate which is removed by filtering
through a fine pinhole in filter paper. Metals more
noble than mercury remain in the boiler of the still.
Finally, the mercury is redistilled in a vacuum
still. Procedures for purification of mercury of
various grades are outlined in reference [78].
Cadmium of electrolytic grade is sublimed under
reduced pressure. The sublimation is done at
about 350 to 400 C in an evacuated Pyrex glass
tube with an external electrical heating jacket.
The distilled cadmium crystallizes out on the cooler
parts of the tube but frequently adheres tightly
to the glass walls. It is usually necessary to break
the glass walls to remove the distilled cadmium
adhering to the walls.
Sulfuric acid of reagent grade is twice distilled
in an all-Pyrex still at a temperature of 270 to 290 C
with the middle fraction being retained. Glass
beads are placed in the boiler to reduce bumping
and the distillation is carried out in a hood.
Water is repeatedly distilled in a Barnstead or
comparable still until the conductivity of the water

becomes as low or lower than 1 X 10~8ohm_1 cm-1.

This water is frequently called "conductivity water."
Cadmium sulfate (CdS04 8/3 H20) of high grade
is obtained by several recrystallizations of the CP.
salt. These recrystallizations are carried out from
aqueous solutions below 43.6 C [79] since above
this temperature the stable salt is the monohydrate,
CdSG*4 H2O. Cadmium sulfate of high grade may
also be prepared from pure cadmium and redis
tilled sulfuric acid solutions or from pure cadmium
nitrate and redistilled sulfuric acid solutions,
followed by recrystallizations of the salt from
aqueous solutions in each case. The latter method
is preferred rather than the use of very impure
cadmium sulfate as starting material.
Both hydrates, CdS04 8/3 H20 and CdS04 H20,
are highly soluble in water; the solubility of the for
mer increases with temperature while that of the
latter decreases. The solubilities of each as a
function of temperature are given in table 6 [79, 80,
81]. The solubilities are expressed as the number
of grams of the anhydrous salt in 1000 g of water.
Saturated solutions of CdSO 8/3 H20 begin to
freeze at 17 C and become completely frozen at
24 C; these temperatures, then, represent the
lower limit for the use of cadmium sulfate cells (see
later). It has been reported [82] that CdS04 8/3
H20 shows a transition to CdS04 7H20 at 4 C.
However, cells tested in this temperature range do
not exhibit an abrupt inflection in emf at 4 C indica
tive of a phase change but a smooth maximum in
the range of 3 to 4 C indicative of a minimum sol
ubility in cadmium sulfate (see table 6).
The amalgam may be prepared either electrolytically by depositing cadmium in mercury or by
heating the two metals together. In the former,
crystals of purified CdS04 8/3H20 are placed on
the surface of mercury (contained on the bottom of
a beaker) and covered with distilled water acidified
with a few drops of sulfuric acid. Mercury is made
Table 6. Solubility of CdSO, 8/3H.O and CdSO, H.O in
aqueous solution
(given as grams of CdSO* in 1000 grama of water)
CdSO, 8/3
Temperature,
c
HrO (as CdSOJ
-18
-IS
-10
- 5
0
3
S
10
IS
20
25
30
35
40
42
43.6'
45
50
SS
60
65
70
72

767.6
765.3
762.1
759.2
757.5
757.2
757.5
7S9.4
761.4
763.7
766.9
771.0
776.5
7833
787 2
790.5
793.5
806.1

" transition temperature.

16

CdSO. HrO
(as CdSOJ

786.5
770.6
754.7
738.9
722.0
707.0
700.0

the cathode and a platinum foil suspended in the

acidified solution at the top serves as anode. Usu
ally only the amount of CdSC>4 8/3 H2O needed to
prepare an amalgam with the desired percentage
of cadmium is used. Since the crystals dissolve
slowly, a current of low magnitude is used initially
until nearly all the cadmium is deposited. The
current is then increased to about 5 times the initial
value for 30 minutes, and then, with the current
still flowing, washing with distilled water is started.
Water is successively added and decanted until the
current decreases to less than 0.01 A for an im
pressed voltage of 1 10 V. The remaining water is
then siphoned off and the last drops of water re
moved from the surface of the amalgam by filter
paper. A cadmium rod may be used as anode in
an alternative procedure; the amount of cadmium
deposited in the mercury is determined from the
current and time by Faraday's law.
The method in which the two metals are heated
together is more convenient. The proper weights
of the two metals are heated together in a covered
porcelain casserole, in a hood, until completely
melted and homogeneous. A slight scum may
form on the surface. This is brushed aside and
the liquid amalgam immediately transferred to an
electrically heated covered buret and introduced
directly to the cell through an electrically heated
delivery tube. Any small amount of scum on the
surface of the amalgam will disappear as soon as
acidified cadmium sulfate is added to the cell. This
is the method now used at the National Bureau of
Standards.
No difference in the properties of standard cells
containing amalgams prepared by these methods
has been observed.
Amalgams exhibit less erratic behavior and cor
rode less rapidly, if at all, than pure unamalgamated
metals when in contact with aqueous solutions of
their salts. Amalgamation raises the hydrogen
overvoltage of metals and reduces stresses within
the metal. It is for these reasons that amalgams
make better anodes for standard cells than the
pure metals. A 10 or 12V2 percent amalgam is
generally used commercially in constructing cad
mium sulfate cells. A 10 percent amalgam is now
used by the National Bureau of Standards. These
are weight percentages. At normal temperatures
amalgams of these precentages consist of two
phases, a liquid and a solid phase. The solid phase
is a solid solution of cadmium and mercury. The
phase diagram [83] for the cadmium-mercury sys
tem is given in figure 10. The composition of the
liquid phase of a 10 percent amalgam is listed for
various temperatures in column 3 of table 3.
Mercurous sulfate may be prepared in a number
of ways. Shortly after Weston patented his cell
much research was devoted to methods of preparing
and storing mercurous sulfate and cells made with
various preparations were extensively studied.
Some samples were white, i.e., devoid of free mer
cury while others were gray in appearance owing
to the presence of finely dispersed mercury. Some

samples were thoroughly washed with absolute

alcohol and anhydrous ether and dried in vacuo
while other samples were kept moist with either
dilute sulfuric acid or saturated solutions of cad
mium sulfate acidified slightly with sulfuric acid.
Mercurous sulfate may be prepared in a number
of ways as follows:
(1) By the reaction between sulfuric acid and
mercurous nitrate [84].
(2) By the action of fuming sulfuric acid on mer
cury [84].
(3) By the action of dilute nitric acid in sulfuric
acid on mercury [44, 84].
(4) By the reduction of mercuric sulfate by mer
cury [44].
(5) By the reduction of mercuric sulfate by sulfurous acid [44].
(6) By the reduction of mercuric sulfate by form
aldehyde [85].
(7) By the recrystallization of commerical mer
curous sulfate from sulfuric acid [84].
(8) By a-c electrolysis [86].
(9) By d-c electrolysis [44, 68].
Of these methods standardization has been on
the last one which is now used by the National
Bureau of Standards. By d-c electrolysis uniform
samples of mercurous sulfate of high purity and
reproducible grain size are obtained. Such sam
ples are free of all foreign materials except sulfuric
acid, water, and mercury, all of which are used in
standard cells, and especially in the mercurous sul
fate paste. In this method [44, 68] mercury anodes,
platinum cathodes, and a 1:6 sulfuric acid-water so
lution are used and electrolysis is carried out in
a darkened room (see sec. 7.7). Mercury is placed
on the bottom of one or two shallow glass vessels.
These vessels are then placed on a glass stand one
above the other, as shown in figure 11, in a deeper
and larger dish ^3 filled with the sulfuric acid so
lution; the acid solution extends over the shallow
vessels. The upper dish contains a central glass
tube through which passes the shaft for the stirrer
for the lower dish; obviously this is unnecessary if
only one shallow dish were used. A platinum-foil
cathode is placed near the top of the solution. The
solution is vigorously stirred at 70 to 200 rpm and

Percent Cadmium
FIGURE 10. Phase diagram for the cadmium-mercury system.

92-17

FIGURE 11. Views of the electrolytic preparation of mercurous sulfate.

Left vessel stirrer in operation, cloudiness due to mercurous sulfate suspended in solution.
Center and right vessels stirring stopped, mercurous sulfate at bottom of vessels and on
surface of mercury in dishes.
the current density at the mercury surface is main
tained at 1 to 2 A per 100 cm 2. When the electro
lyte becomes saturated with mercurous sulfate the
solid salt appears on the surface of the mercury
and must be swept off into the outer dish by stir
ring in order to keep the surface of the mercury
anode clean. In this stirring finely divided mercury
becomes mixed with the mercurous sulfate to give
a gray product. Electrolysis is continued until the
desired amount of gray mercurous sulfate has been
prepared; the current is then cut off but stirring is
continued for several hours. Photographs of the
electrolytic production of mercurous sulfate are
shown in figure 11.
For "neutral" cells the mercurous sulfate is then
washed repeatedly with a saturated solution of
cadmium sulfate and then stored under such a
solution until used. For "acid" cells the mercurous
sulfate is stored under the electrolysis solution
until needed at which time it is washed thoroughly
with the solution of the type to be used in the cells.
Mercurous sulfate is the oxidizing agent or the
"depolarizer" used in Weston cells. It is highly
insoluble in water and in dilute solutions of sul
furic acid. The solubility of mercurous sulfate,
expressed in terms of mercury, in various concen
trations of sulfuric acid is listed in table 7 at 0 and
28 C [87]. In aqueous solutions mercurous sulfate
hydrolyzes according to the reaction:

showed that it did not change appreciably with

temperature, being 0.00198 N at 0 C and 0.00216 /V
at 28 C. Although, theoretically, a concentration
slightly exceeding the equilibrium concentration
could be used to prevent the hydrolysis of mer
curous sulfate, a higher concentration is recom
mended. In initial studies [90], solutions 0.03 /V
to 0.06 N with respect to sulfuric acid were used,
in part because mercurous sulfate exhibits a mini
mum solubility in the range from 0.04 to 0.08 /V.
(Hulett [91] and Sir Frank Smith [92] found the
minimum solubility of mercurous sulfate to be at
approximately 0.04 N, while Craig, Vinal, and Vinal
found it to be at 0.06 N at 28 C and 0.08 /V at 0 C.)
Now, the lower acidity value is preferred since cells
with the lower acidity have shown the greater
stability in emf. Cells with 0.10 /V acid, for example,
tend to gas as a result of the action of the acid with
the cadmium amalgam.
Concentrations of sulfuric acid somewhat higher
than the equilibrium value are chosen, not only
because of the decrease in the solubility of merTable 7. Solubility of mercurous sulfate in aqueous solutions
of sulfuric acid

2Hg2S04(5) + H20(Q = HgjO Hg2S04(s)

+ H2S04(aq)

(9)

to form a basic salt, Hg20 Hg^SC^, and an equi

librium amount of sulfuric acid. Gouy [88] and
Hager and Hulett [89] found this concentration
to be 0.002 /V while Craig, Vinal, and Vinal [87]
93-18

Crams of Hg per 100 ml

Molarity of
H.SO.
0.002
.004
.006
.008
.010
.020
.030
.040
.050
.060
.070
.080
.090

Grams of Hg per 100 ml

Molarity of
H.SO.

<=o*c

l = 28'C

0.0290
.0239
.02 IS
.0203
.0197
.0182
.0179
.0178
.0178
.0178
.0179
.0180
.0181

0.0463
.0395
.0360
.0346
.0338
.0318
.0313
.0317
.0322
.0327
.0332
.0337
.0341

0.100
.200
.300
.400
.500
.600
.700
.800
.900
1.000
2.000
3.000
4.000

l=0'C

I-2BT

0.0183
.0198
.0212
.0224
.0233
.0239
.0244
.0247
.0248
.0249
0240
.0139
.0078

0.0344
.0379
.0403
.0423
.0438
.0451
.0461
.0467
.0470
.0470
.0409
.0294

curous sulfate but because the acid decreases the

rate of solubility of the glass container by the elec
trolyte. It is well known that glasses are more
soluble in neutral salt or alkaline solutions than in
acidic ones [93, 94]. The final solution is titrated
against sodium hydroxide using methyl purple as
an indicator, which has a pH transformation interval
of 4.8 to 5.4 [95].
5.3. Containers for Standard Cells
Saturated standard cells as made at the National
Bureau of Standards are of the H-form shown in
figure 12. Photographs of a cell container and a
completed cell constitute figure 13. The container
is made of Kimble Standard Flint glass, a chem
ically resistant soda-lime glass, having an average
linear thermal expansion coefficient of 92x10"7
per degree C. (The coefficient of linear thermal
expansion is the increase in length per unit length,
measured at 0 C, per degree Celsius.) Since this
linear thermal expansion coefficient approximates
that of platinum, 89xl0~7 per degree C, vacuumtight seals are obtained at the platinum leads.
No lead-containing sealing in glass is used in
sealing in the platinum leads at the bottom of each
limb as the solution may extract the lead leading
to cell deterioration. On the average the height
of the cell is about 92 mm, the diameter of the ver
tical limbs about 16 mm, the diameter of the crossarm about 11 mm, and the distance between limbs
about 22 mm. A constriction is made near the
base of both limbs, as shown, to lock in part of the
crystals of CdSCu 8/3H20. The constriction
may be a complete circumferential indentation or
may consist of several knobs directed inward.
The constrictions are so placed that crystals of
CdSCs 8/3H20 are both below and above them.
These locked crystals prevent the displacement
of the materials in the cell limbs, and those at the
anode help prevent an opening of the circuit within
the cell by any gas that might form on the amalgam.
The platinum leads are secured by cotton thread
to the side of each limb at the constrictions and
cemented in place by collodion; platinum leads
have a tendency to break mechanically at the glass
seal if not securely held in place. For permanent
records a number is etched on each cell on the
outside wall of the container with hydrofluoric acid.
Prior to filling with cell constituents the container
is thoroughly cleaned with nitric acid, rinsed with
distilled water, steamed, and then dried at 110 C.
At various times attempts have been made to
use Pyrex for cell containers. However, since
Pyrex has an average linear thermal expansion
coefficient of 32 X 10-7 per degree C tungsten which
has a linear thermal expansion coefficient of 43 X 10-7
per degree C must be used for external electrical
connections to assure vacuum-tight seals. Tung
sten, however, is more brittle than platinum and a
number of cells have had to be discarded because
of breaks at the tungsten-Pyrex seal.
Fused silica, owing to its extreme inertness, has
frequently been suggested for standard-cell con-

SATURATED SOLUTION

Cd S04-8/3H2O

10% Cd AMALGAM
Figure 12. Sketch of a saturated standard cell of the cadmium
sulfate type (National Bureau of Standards type).

FIGURE 13. Photograph of container and completed saturated

standard cell of cadmium sulfate type (National Bureau of
Standards types).
tainers. Fused silica, however, has two major
drawbacks. It has such a low linear thermal ex
pansion coefficient (5 X 10~7 per degree C) that the
electrical leads have to be brought into the cell in
a special way. Also, a very high temperature must
be employed to seal the cell; the chemicals within
the cell may thereby be affected. These two draw
backs have been solved by the cell design [96] shown
in figure 14. The main features of this design are
(1) the use of two seals at each arm and (2) a graded
seal between the fused silica and Pyrex at the top
of each limb of the container. The use of the
graded seals makes possible the sealing of the cell
at a low temperature after filling. Transparent
silica is used to facilitate the filling of the cells.
The two seals at each arm are spaced about 7 cm

94-19

ABC"
D-

PYREX
GRADED SEAL
TRANSPARENT FUSED SILICA
S6ALE0 WITH FUSED SILICA
SHRUNK AROUND PLAT-

Ficure 14. Cross-sectional sketch of standard-cell container

made offused silica.
apart. The lower seal consists of platinum wire
No. 28 sealed in silica, and although this seal is
not vacuum-tight it is sufficiently tight to retain
the cell constituents. The upper vacuum-tight
seal is a Houskeeper [97] type and consists of
molybdenum ribbon sealed in silica (this type of
seal cannot be used as the lower seal since molyb
denum is chemically attacked by the cell con
stituents). The molybdenum ribbon and platinum
wire are welded at a point about 2 cm below the
upper seal. Silica is shrunk around the weld, the
molybdenum ribbon, and platinum wire. Copper
terminals are used and are silver-soldered to the
molybdenum ribbon. Errors that might arise from
thermoelectric effects are prevented by completely
immersing the cells in oil at a constant tempera
ture. Dimensions of the cell are included with
the figure.
Although it has been generally thought that inter
actions between the glass container and the cell
electrolyte cause "aging" or drifts in the emf of
standard cells, available data have not shown sig
nificant differences between cells made in soft
glass, Pyrex, or fused silica on a long-term basis.
Plastic containers have also been proposed [98].
These must prevent vapor transport or a drying out
of the cells may occur.
5.4.

Assembly and Mounting of

Standard Cells

Two long-stemmed funnels, one sliding through

the other, are employed at the National Bureau of
Standards for introducing materials into the con
tainer. The inner funnel carries the material and
the outer one serves to prevent the material from
coming into contact and "mussing up" the walls
of the container. After introducing the material
the stem of the inner funnel is drawn up into the
stem of the outer funnel; both funnels may then be

withdrawn from the cell without any of the material

touching the container walls.
Mercury is first placed in the bottom of one limb
and the amalgam in the bottom of the other limb,
each to a depth of about 6 mm. The amalgam is
added while warm and in a single liquid phase; on
cooling the amalgam becomes two-phased, solid and
liquid. The mercurous sulfate is then placed in a
Gooch or similar crucible, washed free of the solu
tion under which it was stored with dilute sul
furic acid and then washed with solution of the
type to be used in the cell, and then while moist
introduced over the mercury to a depth of about 13
mm. The mercurous sulfate should be mixed with
a small amount of mercury (partially done during
the electrolytic preparation of mercurous sulfate)
and finely divided crystals (fineness of granulated
sugar) of CdSO* 8/3H2O prior to introduction to
the cell; this mixing may be done prior to the wash
ing procedure. The mixing and washing of the
mercurous sulfate paste hastens the attainment of
chemical equilibrium within the cell after its
assembly.
Crystals of CdS04 8/3H20, of a size that will
pass through a tube of 4-mm bore, are then added
to both limbs of the cell to a depth of about 10 mm
at the negative electrode and about 8 mm at the
positive electrode. Finally, a saturated solu
tion of CdS04 8/3H20 is added to a level sbghtly
above the crossarm, and the cell is then hermetically
sealed.
In some cells, especially of larger size, large
crystals (about 10 to 15 mm in diameter) of CdSCs
8/3H20 are used. Larger crystals have an advan
tage over smaller crystals in that any gas that may
form at the electrode surface (especially the nega
tive one) will not become entrapped by the crystals
whereby an open circuit might be produced. How
ever, cells with large crystals tend to come to
equilibrium, after a temperature change, more
slowly than those made with small crystals.
Unsaturated cells are made similarly except that
no crystals of CdS04 8/3H2O are used and they
are made portable (shippable) by inserting cork
or plastic rings, covered with linen, over the elec
trode surfaces. In some cells ceramic discs, either
locked in place or supported by ceramic rod which
protrudes through stoppers in each limb, are used
[99,100]. The unsaturated cell is the commercial
type used widely in the United States for work
requiring no greater accuracy than 0.005 percent;
it is not made at the National Bureau of Standards.
It is used for pyrometer work, in pH meters, re
cording instruments, etc., and is usually housed
in nontransparent copper-shielded cases for general
laboratory work. A copper-shielded case is em
ployed to assure a uniform temperature at both
Limbs of the cell (see later). Saturated cells are
not mounted in cases since they are intended for
immersion in temperature-controlled oil or air
baths. Commercial saturated cells are usually
mounted in groups of 3, 4, or 6 on special racks

95-2 0

for convenience in use. NBS saturated cells are

mounted in oil as described above in section 2.3.
In practice, saturated standard cells are main
tained at a constant temperature in thermostat
ically controlled oil baths or in portable thermo
statically controlled air boxes. The latter are
generally made after a design first proposed by
Mueller and Stimson [101]. The cells are housed
in a thin-walled aluminum box which rests in a larger
thick-walled aluminum box, the temperature of
which is controlled by a mercury-in-glass thermoregulator. The aluminum boxes are thermally in
sulated and are enclosed in a wooden box which
also contains an a-c relay, a transformer, and a
pilot light. The box is operated on 110 V 60-cycle
a-c line. The leads from the individual cells are
brought to binding posts on the outside of the box.
These boxes are designed to operate at tempera
tures above room temperature; the choice of tem
perature depends on the location where the boxes
will be used or on the size of the box. As a rule
these boxes operate at some temperature between
28 and 37 C. The temperature of the cells in
the air box is measured by a mercury-in-glass
thermometer provided with the box. In some
boxes, a well is added to provide for a platinumresistance thermometer in which case the tempera
ture is so measured. A detailed description of a
Mueller-Stimson box is given in Appendix 7.

been "aged" (see later). If reproducibility of 0.O1

mV in emf is considered adequate, all of these fac
tors are inconsequential.
Cells within any one group, although made at the
same time of the same materials, may have emfs
that differ by as much as 0.005 raV (or 5 /iV).
These differences may not only be present initially
but may persist for years; i.e., individual cells,
although differing in emf may show high stability in
emf. Although diffusion is a slow process, acid or
solution concentration inequalities within any one
cell cannot explain these differences. Instead they
must arise (1) from slight differences in the com
position of the amalgam between cells (see fig. 15,
especially the horizontal sections; this figure is
discussed in sec. 6.3) even though precautions are
taken to keep the amalgam homogeneous during
the filling process, (2) from slight differences in the
acidity of the solutions between cells produced
during the filling process or in interactions with the
glass containers, and/or (3) to slight differences
between cells in the grain size of the mercurous
sulfate used in the positive electrodes.

5.5. Electromotive Forces of Newly

Made Cells
Cells when first made will, except in the most
rare cases, exhibit emfs that will differ consider
ably from the final steady values. Values steady
to 0.01 mV or 10 fi\ are usually obtained within a
few days but one to three years may be required
before values steady to 0.1 fi\ are attained. In
some cases the emfs decline in their approach to
equilibrium, in other cases they increase. There
are two main reasons for this "aging" process,
namely, (1) equalization of acid throughout the cell
and especially within the mercurous sulfate paste,
and (2) attainment of solution saturation. Although
attempts are made to take care of both of these
during cell assembly, inequalities in acid concentra
tion may readily occur during the filling process
owing to evaporation; and since the preparation of
solutions of cadmium sulfate which are truly sat
urated is an extremely slow process unsaturated
solutions may inevitably be used in preparing cells.
Cells that "age" with a decrease in emf probably
have been made with unsaturated solutions, even
though crystals of CdS04 8/3H20 are present, or
have too much acid in the negative limb while those
that increase in emf during "aging" probably have
excess acid in the positive limb. Also, the reaction
between the electrolyte and the glass container
probably contributes to "aging" since as the glass
neutralizes some of the acid in the cell the emf will
increase. However, this reaction eventually ceases
as is shown by the emf stability of cells which have

Percentage of cadmium
FIGURE 15. Relation of the electromotive force of the cadmium
sulfate standard cell to the percentage of cadmium in the
amalgam and the temperature.

96-21

6. Effect of Variations in Components on the Electromotive Force of Standard

Cells
6.1. Concentration of Solution
The emf of cadmium sulfate cells depends on the
concentration of the cadmium sulfate solution.
Over a limited range of concentration near satura
tion the emf increases about 0.0017 V for a one
percent decrease in cadmium sulfate content. For
more dilute solutions the change is somewhat
greater. In table 8 the emfs at 25 C corresponding
to various concentrations of cadmium sulfate, near
saturation, are given. These data are based on
results of Vosburgh and Eppley [102] for cells con
taining 12 Vi percent amalgams and 0.023 N H2SO4;
their data reported in international volts have been
converted to absolute volts here. For neutral solu
tions the emfs listed here would be 18 piV higher.
Table 8. Electromotive forces of cadmium sulfate cells at 25 C
as a function of electrolyte concentration
CdSO.
AIM
42.39
42.63
42.77
42.90
42.94
saturated solution.

Emf

CdSO.

Emf

V
1.021289
1.020019
1.019673
1.019315
1.019121
1.019043

percent
42.98
43.06
43.12
43.22
43.402'

V
1.018990
1.018883
1.018704
1.018603
1.018392

Differences between the emf of "neutral" and

"acid" cells according to the Smith formula are
listed in table 9. Sir Frank Smith [103] also
investigated the effect on the emf of cadmium
sulfate cells if the acid were confined to one or the
other of the electrode compartments. For these
effects, valid to 4 N, he gave:
Aegative(volts) = 0.01090* - 0.00125*2

A PO(jmve(volts)= -0.01 150* + 0.00120* 2 (13)

the summation of which gives the Smith equation
above for the effect of acid on the emf of the cell
as a whole. These differences for various normali
ties of acid are given in table 10. These relations
show that more acid at the negative electrode
increases the emf while more acid at the positive
electrode produces a decrease in emf. It is neces
sary, then, for high reproducibility and stability in
emf that the acidity be the same and remain the
same at the two electrodes.
Table 9. Differences of emf of "acid" cells from the standard
value of Weston "neutral" cells

6.2. Acidity of Solution

The addition of sulfuric acid slightly decreases
the emf of a cadmium sulfate cell. For the satu
rated cell several equations have been proposed
relating the change in emf to acid concentration.
For acid concentrations up to 4 N Sir Frank Smith
[103] gave:
A (volts) =- (0.00060* + 0.00005*2)

(10)

Normality of
H30.

Difference of
emf

Normality of
H.SO.

Difference of
emf

0.01
.02
.03
.04
.OS
.06
.07
.08
.09
.10
.20

y
-6
-12
-18
-24
-30
-36
-42
-48
-54
-60
-122

0.30
.40
.50
.60
.70
.80
.90
1.00
2.00
3.00
4.00

P-V
-185
-248
-312
-378
-444
-512
-580
-650
-1400
-2250
-3200

TABLE 10. Differences of emf of cells from the standard value

of Weston "neutral" cells if the acid is confined to the negative
or positive limb of cell

where AE is the difference in emf of acid cells from

neutral ones and * is the normality of the sulfuric
acid solution before it is saturated with CdS04-8/3
H20. For acid concentrations up to 0.4 N only, the
National Physical Laboratory [104] gave the linear
relation
A(microvolts) = 615*.

(12)

(11)

Obata [105] and Ishibashi and Ishizaki [106] also

gave linear relations with the coefficients being,
respectively, 855 and 833; Vosburgh [107] found
Obata's equation to be valid to 1.49 N. In these
last two, x refers to the normality of the acid in a
saturated solution of CdS(V8/3 H20. All of these
formulas agree closely if applied properly. For low
acidities, the acidity of a solution of sulfuric acid
after saturation with CdS04-8/3 H20 is 0.767 of that
before saturation.
97-22

Differencea of rmf, microvolts

Normality of
H.SO.
0.01
.02
.03
.04
.05
.06
.07
.08
.09
.10
.20
.30
.40
.50
.60
.70
.80
.90
1.00
2.00
3.00
4.00

Negative limb

Positive limb

+ 109
+ 218
+ 326
+ 434
+ 542
+ 650
+ 757
+ 864
+ 971
+ 1078
+ 2130
+ 3157
+ 4160
+ 5138
+ 6090
+ 7018
+ 7920
+ 8798
+9650
+ 16800
+ 21450
+ 23600

-115
-230
-344
-458
-572
-686
-799
-912
-1025
-1138
-2252
-3342
-4408
-5450
-6468
-7462
-8432
-9378
-10300
-18200
-23700
-26800

6.3. Composition of the Amalgam

The emf of cadmium sulfate cells depends on the
composition of the amalgam as may be seen by the
data (converted to absolute volts) of Sir Frank Smith
[108] and shown plotted in figure 15. This figure
clearly shows that the useful range of amalgam com
position is from about 8 to 14 percent cadmium for
normal temperatures. The useful range of amalgam
composition for standard cells is limited to the
horizontal part of the curves; here the emf is very
insensitive to or not critically dependent on the
exact percentage of cadmium in the amalgam. In
this range the amalgam consists of two phases,
solid and liquid, with the solid phase being a solid
solution of mercury and cadmium. For low and
high percentages of cadmium the emf is very sensi
tive to the exact amount of cadmium in the amal
gam; this sensitivity is more marked at lower tem
peratures. In figure 16 is shown the range over
which amalgams of various cadmium content may
be safely used. The range for use for a 12V2 per
cent amalgam is about 12 to 62 C while for a 10
percent amalgam the range is from about 8 to
51 C. It is for this reason that a 10 percent
amalgam is used in cells made at the National
Bureau of Standards. The significances of the
lower dotted lines A and B appearing in figure 16
were discussed under the section on cadmium
sulfate; the significance of the upper dotted line is
discussed in the next section.
6.4. Crystal Phases of Cadmium Sulfate
As was stated above, cadmium sulfate over a
normal temperature range exists in two different
hydrates, CdSO, 8/3H20 and CdS04 H20 with the
transition temperature being 43.6 C, below which
the former hydrate is the stable form. The emfs
of saturated standard cells made with these two
different hydrates differ except at the transition
point where the emfs are the same. Cells can be
used in the metastable range for short periods of
time. In either case, at a particular temperature,
cells made with the more stable hydrate have the
higher emf. These facts are illustrated by the data
[109] given in table 11. These cells were "neutral"
cells made with 10 percent amalgams. When a cell
made with a particular hydrate is carried over into
the metastable range its emf will correspond at first
to that for a supersaturated solution of the meta
stable hydrate and then slowly rise in value as the
7. Characteristics
7.1. Emf-Temperature Coefficient
Unsaturated standard cells have a very low emftemperature coefficient. Although it is frequently
stated that unsaturated standard cells have a dE/dt
of 0.00001 V/deg C, the emf-temperature coeffi
cient is a function of temperature, is dependent on
the concentration of the electrolyte, and increases

6
8
10 12 14 16 IB 20
Percentoqe of Cadmium in Amalgam
FIGURE 16. Range oftemperature over which cadmium amalgams
of various cadmium content may be used in standard cells.
hydrate is converted to the more stable form. This
transition in phase is slow unless the two hydrates
are present; then the transition is relatively fast.
Accordingly, overheating of normal cells may have
only transient adverse effects if the overheating is
only momentary; otherwise the adverse effects may
persist for long periods of time. In any case normal
type saturated cells made with crystals of CdS04-8/3
H20 should not be heated above 43.6 C; the upper
dotted line in figure 16 corresponds to this upper
temperature limit.

Table 11. The emfs of Weston Normal Cells made with crystals
o/CdSOa-8/3 H20 orcrystals o/CdSCVH.O*
Temperature

CdSO,-8/3 H.O
cells

CdSO,H,0
cell*

c
20
25
30
35
40
43.4"
45
50

volts
1.01866
1.01844
1.01817
1.01785
1.01752
1.01729
1.01716
1.01682

volu
1.01333
1.01417
1.01501
1.01587
1.01672
1.01729
1.01757
1.01842

'Original data of ref. [ 109] were in international volts; are converted here to absolute
volts.
*A better value for the transition temperature is 43.6 "C (79).
Standard Cells
as the cell ages. In table 12, values of dE/dt are
given for temperatures from 15 C to 45 C for
various weight percents of CdS04. The emf values
corresponding to cells made with the various con
centrations of CdS04 are given at 25 C as an aid
to users in determining the emf-temperature coeffi
cient of a particular cell. For example, if a cell had
at 25 C an emf of 1.019043 V its dE/dt would be

98-23

Table 12. Emf-temperature coefficients of unsaturated standard

cells
(In microvolts per degree Celsius)
CdSO. Emf, 25 'C
15 *C
<*>
41.84
42.39
42.63
42.77
42.90
42.94
42.98
43.06
43.12
43.22
'43.402

20 *C

25 *C

dEldl
30 #C

35 *C

40 *C

45 "C

volts
1.021289 -12.0 -13.2 -13.5 -12.9 -11.5 -9.2 -6.1
1.020019 -3.7 -5.3 -6.0 -5.7 -4.4 -2.2 + 1.0
1.019673 -1.2 -3.0 -3.8 -3.6 -2.S -0.4 + 2.7
1.019315 + 0.2 -1.3 -2.0 -1.6 -0.2 + 2.2 + 5.6
1.019121 + 1.6 -0.1 -0.7 -0.4 + 1.0 + 3.5 + 6.9
1.019043 + 1.9 + 0.3 -0.4
0
+ 1.3 + 3.8 + 7.3
1.018990 + 2.3 + 0.6
0
+ 0.1 + 1.7 + 4.2 + 7.6
1.018883 ++ 3.8
3.2 + 1.5 + 0.8 + 0.9 + 2.5 + 4.9 + 8.4
1.018704
+ 2.0 + 1.3 + 1.6 + 3.0 + 5.5 + 8.9
1.018603 ' + 4.8 + 2.9 + 2.2 + 2.5 + 3.9 +6.4 + 9.9
'1.018392 -30.4 -40.6 -49.4 -56.6 -62.4 -66.6 -69.4

this composition corresponds to a solution saturated at 4 *C and gives a cell with

aotC/niofirro al25'C.
* supersaturated.
' percentage of CdS04 in a saturated solution at 25 *C; dEldt values given are for
solutions saturated with CdSO, - 8/3 HiO at the respective temperatures.
* -contains 0.03 N H,SO- 0.4 piV/deg C at 25 C, while if it had an emf of
1.018704 V its dEldt would be +1.3 /xV/deg C.
Inspection of the data of table 12 also shows that a
cell made with a solution containing 42.98 wt percent
of CdS04 has a zero emf-temperature coefficient at
25 C. Vosburgh and Eppley [102] showed that this
percentage corresponds closely to a solution satu
rated at 4 C (the solutions of Vosburgh and Eppley
also contained 0.023 N r^SO*). Since a solution
saturated in the range from 3 C to 4 C leads to
cells with negligible emf-temperature coefficients
such solutions are frequently used in the construc
tion of unsaturated standard cells.
Saturated standard cells exhibit a much larger
emf-temperature coefficient than unsaturated standare cells owing to the change in solubility of CdS04*
8/3H2O with temperature. They exhibit a maxi
mum emf at 3 to 4 C, the temperature range of the
minimum solubility of CdS04 8/3H2O (see table 6).
Two formulas relating the emf to temperature have
been proposed. Wolff [32] working with 12V2 per
cent amalgams obtained:

valid from 20 to 40 C; Vigoureux and Watts used

the 6 percent amalgam for the lower temperatures
but the 10 percent amalgam gives the same results
as long as the amalgam contains liquid and solid
phases; thereafter the emf decreases below that
given by eq (15). The cell becomes completely
frozen at -24 C; eq (15) gives -1139.5 /aV for
E-24 c Ezo c- However, when the cell becomes
completely frozen the emf is approximately 1.007 V.
The emf of "neutral" saturated standard cells, as
a function of temperature, based on the above
formulas, is given in figure 17. Both formulas
reproduce the maximum in emf at 3 C to 4 C, the
temperature range of the minimum solubility of
CdS04 8/3H20. Since the acidity of the electro
lyte does not appreciably affect dEldt, the emfs of
"acid" cells would parallel the curve shown. The
displacements in the curves would depend on the
acidity employed (see sec. 6.2). In this figure the
value at 20 C, the temperature at which interna
tional comparisons are made, is labeled Emf
standard.
The differences between eq (14) and (15) are not
large. In table 13 the differences in emf at various
temperatures from the emf at 20 C as given by
these two formulas, are given. Since 10 percent
amalgams are now used widely in preparing satu
rated standard cells eq (15) (or 18 below) should be
used to calculate the change in their emf with
temperature.
Above 43.6 C where CdS04 8/3H20 transforms
to CdSCVHgO the emf-temperature relation is given
by [109]:

1.01920,

E, (volts) = 43.6 + 0.000173 - 43.6)

,
.
,
1
1
,

(16)
1

Et = 20= - 0.00004060(t* - 20) - 0.000000950(f - 20)2

+ 0.000000010(f-20)3

(14)

in international volts where Et is the emf at tempera

ture t and 20 is the emf at 20 C. This equation
is known as the International Temperature Formula.
It is valid from 12 to 40 C but may be used to 0 C
as long as the liquid and solid phases of the amalgam
are present. Vigoureux and Watts [110] using 6
and 10 percent amalgams, to extend the temperature
range, obtained:
Et = Eur ~ 0.00003939(1 - 20) - 0.000000903(t - 20)*
+ 0.00000000660(1 - 20)3 - 0.000000000150U - 20)4
(15)
in international volts. These two equations are also
applicable in absolute volts. This later equation is

1.017201*
-20

'
-10

'
0

'
'
IO
+20
Temperoture, *C

<
+30

' *
+4C

1
+S0

Figure 17. Relation of the electromotive force of the cadmium

sulfate standard cell to the temperature.

99-24

Table 13. Differences in emf of saturated standard cells from

the value at 20 C
Tempera Vigoureux and Inter- Tempera Vigoureux and
Watts formula
ture Watts formula national ture
formula
c
-20
-19
-18
-17
-16
-15
-14
-13
-12
-11
-10
-9
-8
-7
-6
-5
-4
-3
-2
-1
0
+1
2
3
4
5
6
7
8
9
10
11
12

microvolts microvolts
-675.6"'
-57S.8"-'
- 482.0b
-394.2*-"
-312.1"
-235.6'
- 164.5"
-98.6"
-37.8"
+ 18.2'
69.3"
115.8"
157.9"
195.6
229.2
258.7
284.3
306.0
324.1
338.7
352.2
350.0
357.8
360.0
364.8
362.4
366.7
363.9
362.4
365.6
358.0
361.7
354.9
350.8
340.8
345.4
328.3
333.3
318.5
313.2
301.1
295.6
281.3
275.7
259.0
253.5

c
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35'
36
37
38
39
40
41
42
43
43.6

International
formula

Acidity (H,SO) of solution in cell

Temperature
neutral "

microvolti
229.0
202.3
173.5
142.7
109.9
75.2
38.5
0.0
-40.3
-82.2
-126.1
- 171.6
-218.8
-267.7
-318.1
- 370.2
-423.9
-479.2
-536.1
-594.6
-654.6
-716.3
-779.6
-844.4
-911.0
-979.1
-1049.0
-1120.6
(-1193.8)
(- 1268.9)
(- 1345.7)
(- 1392.3)

microvolts
234.3
207.4
178.1
146.7
113.1
77.4
39.7
0.0
-41.5
-84.9
-130.1
- 177.0
-225.5
-275.7
-327.4
-380.6
-435.2
-491.1
-548.4
-606.9
-666.6
-727.4
-789.3
-852.1
-915.9
-980.6
-1046.1
-1112.4
(-1179.3)
(- 1246.9)
(-1315.1)
(- 135S.8)

E, = E2g> - 0.00005390U - 28)

- 0.0000007 10350 - 28)2
+ 0.000000010U-28)3

(17)

and
E,=Eve- 0.000052899(t - 28)
- 0.00000080265(* - 28)2
+ 0.0000000018130 - 28J3
-0.00000000014970 -28)"

0.03/V

O.OSN

0.MW

c
volts
volts
volts
volts
20
1.018636
1.018612
1.018596
1.0185S6
25
1.018417
1.018393
1.018377
1.018337
28
1.018266
1.018242
1.018226
1.018186
30
1.018157
1.018133
1.018117
1.018077
32
1.018041
1.018017
1.018001
1.017961
35
1.017856
1.017832
1.017816
1.017776
37
1.017725
1.017701
1.017685
1.017645
-Actually 0.0O092jV; cadmium sulfate hydrolyxes to produce HiSO, of this acidity
at 25 "CPS).
In table 14 the nominal values of the emf of
"neutral" and "acid" saturated cells are given at a
series of temperatures at which cells are most
frequently calibrated at the National Bureau of
Standards. Exact agreement with these nominal
values cannot be expected since the emfs of satu
rated cells are very sensitive to the acidity of the
electrolyte, to the amalgam composition, and to the
extent of the solubility of the glass container in the
cell electrolyte; even so they have high emf stability.

*for first few hours after a change from a higher temperature; electrolyte starts
to freer* at 17 "C.
'6
percent
amalgam. were obtained by extrapolation.
cvalues
in parenthesis
and unlike the cells made with CdS(V8/3 H2O the
emf-temperature coefficient is positive. The equa
tion also applies for the metastable monohydrate
down to 20 C. Vinal and Brickwedde used 43.4
in eq (16) but later results [79] showed 43.6 to be a
better value and is used here.
Since 28 C is frequently used as a maintenance
temperature for saturated standard cells, it is con
venient in such instances to have the emf-tempera
ture relation expressed in terms of this temperature.
Equations corresponding, respectively, to eq (14)
and (15) but with 28 C as the reference temperature
are:

in volts.

Table 14. Nominal emfs of saturated standard cells at some

common temperatures
(Cells made with 10% cadmium amalgam)

(18)

Frequently, the question arises as to the effect of

small changes in temperature on the emf of satu
rated standard cells. Table 15 lists the changes in
emf that are produced by changes in temperature of
only 0.001 or 0.01 C. The + sign for the emf
refers to changes in emf produced by a decrease in
temperature while the sign for the emf refers
to changes in emf produced by an increase in
temperature. These data show that temperature
must be controlled more accurately at the higher
temperatures for comparable control of emf. They
also show that for precisions of 0.5 ptV the tem
perature must be controlled to slightly better than
0.01 C; for 0.05 /xV to better than 0.001 C.
The emf-temperature formulas given above refer
to the saturated cell as a whole, assuming that all
parts of the cell are at the same temperature. The
separate limbs of the cell have much larger co
efficients, the negative limb having a negative co
efficient and the positive limb a positive coefficient.
The emf-temperature coefficient of the whole cell
is the summation of those of the two limbs. The
emf-temperature coefficients for each limb from 0
to 37 C are given in table 16; those given in paren
thesis were obtained by interpolation or extrapola
tion of the data of Sir Frank Smith [103]. The sum
mation of those for the two limbs are given in
column 4; these agree well with those calculated
by International Temperature Formula. It is ob
vious, in view of these data, that saturated standard
cells must be kept at a uniform temperature for
high accuracy and precision. In table 17, are given
the errors that would arise in the emf of a cell if
the positive limb were at a slightly higher or slightly
lower temperature than the negative limb. Also
given are data for the case in which the tempera
ture of the negative limb were slightly higher or
lower temperature than the negative limb. It is for

100-25

TABLE 15. Effect of small changes in temperature on the emf of

saturated standard cells at various temperatures
Temperature

Table 17. Errors produced in emf of saturated standard cells

if the temperature of the positive limb (or negative limb)
differed from that of the negative limb (or positive limb), by
small amounts, at various temperatures

0.001 -C 0.01 *C
Temperature

c
microvolt microvolt
70.04
70.4
20
70.05
7 0.5
25
70.05
70.5
28
70.06
7 0.6
30
70.06
7 0.6
32
70.06
7 0.6
35
70.07
70.7
37
+ sign for the emf refer* to decrease in temperature.
tign for the emf refers to an increase in temperature.
Table 16. Temperature coefficient of positive and negative
limbs of saturated standard cells and of complete cells at
various temperatures

Temperature
difference 0"C

20 *C

25 "C

28 *C

30 *C

32 *C 35 "C

37 "C

*v
+ 3.1
+ 1.6
+0.3
-0.3
-1.6
-S.1

MV
+ 3.1
+ 1.6
+ 0.3
-0.3
-1.6
-3.1

*v
+ 3.1
+ 1.6
+ 0.3
-0.3
-1.6
-3.1

v
+ 3.1
+ 1.6
+ 0.3
-0.3
-1.6
-S.1

-3.7
-1.9
-0.4
+ 0.4
+ 1.9
+ 3.7

-3.8
-1.9
-0.4
+ 0.4
+ 1.9
+ 3.8

-3.8
-1.9
-0.4
+ 0.4
+ 1.9
+ 3.8

-3.8
-1.9
-0.4
+ 0.4
+ 1.9
+ 3.8

Poa tive limb

+0.01
+0.005
+ 0.001
-0.001
-o.oos
-0.01

(A"
+ 2.9
+ 1.5
+0.3
-0.3
-1.S
-2.9

<v
+ 3.1
+ 1.6
+ 0.3
-0.3
-1.6
-3.1

MV
+ 3.1
+ 1.6
+ 0.3
-0.3
-1.6
-3.1

(*V
+ 3.1
+ 1.6
+ 0.3
-0.3
-1.6
-3.1

Negative limb

Temper Negative limb Positive limb

ature

+ 0.01
+ 0.005
+ 0.001
-0.001
-0.005
-0.01

Complete cell
International Vigoureux
Observed temperature and Watts
formula
formula

c
vottldeg
vottldeg
vottldeg
vottldeg
-0.000288 + 0.000290 + 0.000002 + 0.000009
0
5
-0.000297 + 0.000291 -0.000006 -0.000005
10
-0.000315 + 0 000295 -0.000020 -0.000019
15
-0.000333 + 0.000302 -0.000031 -0.000031
-0.000350 + 0.000310 -0.000040 -0.000041
20
25
-0.000366 + 0.000314 -0.000052 -0.000050
28
(-0.O0O372C (+0.000314) (-0.000058) -0 000054
30
-0.000374 + 0.000314 -0.000060 -0.000057
32
(-0.000376) (+0.000314) (-0.000062) -0.000059
35
(-0.000380) (+0.000314) (-0.000066) -0 000062
37
(-0.000382) (+0.000314) (-0.000068) -0.000064
" Values in parenthesis obtained by interpolation or extrapolation.

vottldeg
+ 0.000009
-0.000006
-0.000019
-0.000030
-0.000039
-0.000048
-0.000053
-0.000056
-0.000059
-0.000064
-0.000067

this reason that copper shields are used around the

portable unsaturated cells (see sec. 5.4). Park [111]
and Eppley [112] investigated the effect of service
temperature conditions on the emf of unsaturated
standard cells and Park showed that errors in emf
arising from thermal inequalities could be reduced
to a minimum by enclosing the cell in a copper
shield.

-2.9
-l.S
-0.3
+ 0.3
+ 1.5
+ 2.9

-3.5
-1.8
-0.4
+ 0.4
+ 1.8
+ 3.5

-3.7
-1.9
-0.4
+ 0.4
+ 1.9
+ 3.7

-3.7
-1.9
-0.4
+ 0.4
+ 1.9
+ 3.7

on cooling
+ 20H-

on heating

-40

7.2. Emf-Temperature Hysteresis

In general, if saturated or unsaturated standard
cells are subjected to slowly changing temperatures
their emf will follow closely the relations given
above. If, on the other hand, the cells are sub
jected to abrupt temperature changes, deviations
from the true emf will occur. These deviations are
generally referred to as hysteresis. On cooling,
the cells show first too high an emf and then a slow
decrease in emf to the equilibrium value. On
heating, the cells show first too low an emf and then
a slow rise in emf to the equilibrium values. The
magnitude of the hysteresis is given by the per
centage deviation from the equilibrium value and
is usually greater when the temperature is de
creased than when it is increased. These general
relations are illustrated in figure 18 for abrupt heat
ing of new cells from 25 to 30 C and for abrupt
cooling from 30 to 25 C for unsaturated standard
cells having a dE/dt of 5 /nV/deg C in this tempera
ture range. The magnitude of the hysteresis in
each case is given by the distance marked by an "A"
divided by the emf of the cell at the starting tem-

10

20

30
hours

50

GO

Figure 18. Relations showing the emftemperature hysteresis of

unsaturated standards when abruptly heated or cooled 5 C.
perature (this emf may be taken as unity). This
figure clearly shows that the magnitude and dura
tion of hysteresis obtained on cooling is nearly
double that found on heating; in either case the
magnitude of the hysteresis is less than 0.01
percent.
For unsaturated cells the magnitude and duration
of hysteresis depend on the type of construction,
age, acidity and concentration of the electrolyte,
on the purity of the materials, and on the rate and
magnitude of the temperature change. To these
must be added the size and change of solubility with
temperature of the crystals of CdSOa 8/3 H2O
for the saturated cell. Therefore, it is not possible
to give quantitative data that may be applied as

101-Z6

corrections to the emf of cells under diversified

conditions involving abrupt temperature changes.
Furthermore, the magnitude of hysteresis for un
saturated cells increases with age; a cell 10 years old
exhibits about 10 times the hysteresis of a new cell.
In general, for a 5 C abrupt change in temperature
the hysteresis of new unsaturated cells ranges from
0.01 to 0.02 percent (0.0001 to 0.0002 V) on cooling
and from 0.005 to 0.01 percent (0.00005 to 0.0001
V) on heating. The cells usually recover their
original emf within 1 to 2 days after cooling and
within 10 to 12 hours after heating. For older
cells, hysteresis may persist for days or even
months. Proportionate hysteresis is to be expected
for larger or smaller temperature intervals than 5
C. In general, saturated cells show less hysteresis
than unsaturated ones. In some cases, saturated
standard cells do not exhibit the "overshoot"
typical of hysteresis. Instead they approach
steady emfs at a new temperature at an exceedingly
slow rate. This phenomenon here called "lag" is
sometimes referred to as "negative hysteresis"
and is attributed to the slowness in the precipita
tion of CdS04 8/3 H2O on cooling or to the slowness
of dissolution of CdSO< 8/3 H20 on heating.
Many explanations have been given for hysteresis
but no single factor is alone responsible. The
difference between the heat capacities of the two
limbs of the cell, changes in solubility of CdSCv8/3
H20 (for the saturated cell) with temperature, septa
(in the unsaturated cell), and the disturbances of
equilibrium conditions within the cell during tem
perature changes must all contribute to hysteresis.
In any case standard cells should be maintained at
constant temperature if at all possible. Tempera
ture fluctuations may be kept at a minimum for
portable unsaturated cells by placing them in tem
perature-lagged boxes or in Dewar flasks. A view
of a temperature-lagged box used at the National
Bureau of Standards in the testing of unsaturated
cells is shown in figure 19.
7.3. Temperature Range
The range of temperature over which standard
cells may be used is dictated by the composition of
the amalgam, the transition temperature of the
cadmium sulfate hydrates, and the freezing point of
the electrolyte. The range over which amalgams
of various cadmium contents may be safely used
was shown above in figure 16. For 12V2 percent
amalgams this range is 12 to 62 C; for 10 percent
amalgams it is 8 to 51 C. Both of these amal
gams may be used for a short time (2 to 3 hr) below
these temperatures (12V2 percent amalgam to 0C;
10 percent amalgam to 20 C) or as long as the
amalgams consist of two phases, solid and liquid.
However, for work of the highest precision the cells
should be confined to the temperature ranges shown
in figure 16.
A 6 percent amalgam should be chosen for lower
temperatures; its useful range is 24 to 28 C. For
high temperatures amalgams of high cadmium con-

Ficure 19. Temperature-lagged box used at the National Bureau

of Standards in the testing of unsaturated standard cells.
tent should be used, viz, a 14 percent amalgam will
give cells that could be used, from 24 to 67 C. It
may be possible to use unsaturated standard cells
to temperatures of the boiling point of the electro
lyte (slightly above 100 C) by using amalgams of
high cadmium content (above 20 percent) but these
cells would not be suitable for precision work much
below 80 C.
At temperatures above 43.6 C, even though the
higher-percentage amalgams are satisfactory, satu
rated cells must be made with crystals of CdSCv
H20 since this is the stable form of solid cadmium
sulfate above 43.6 C. For unsaturated standard
cells no such problem exists and attention need be
given only to the composition of the amalgam.
The lower limit of use is 24 C where the cell
becomes completely frozen. Freezing begins at
17C. Cells completely frozen at 24 C will
behave normally after thawing if freezing has not
caused a fracture in the cell. The time required
for the cell to recover its normal emf may be long,
however. The internal resistance of standard cells
at 10 C is about 6 times that at 25 C and this
increase in internal resistance may limit the use of
the cell.
Significant temperatures in the use of standard
cells are summarized in table 18. In practice, cells
should probably be subjected to a somewhat lesser
range of temperatures than shown; 16 to 40 C is
a good range for saturated cells and 4 to 40 C for
unsaturated ones.
7.4. Emf- Pressure Coefficient
The effect of pressure on the emf of a galvanic
cell at constant temperature is given by
(dE\

~(AV)k

where AV is the volume change in cubic centi

meters at atmospheric pressure per faraday, n is

102-27

Table 18. Significant temperatures in the use ofstandard cells (also temperature limitsfor use ofstandard cells)

Temperature
Remarks

67
62
51
43.6
40
28-37
28
24
20
12.1
4
3-4
-8
-17
-24

152.6
143.6
123.8
110.5
104
82.4-98.6
82.4
75.2
68
53.8
39.2
37.4-39.2
17.6
1.4
-11.2

Upper limit for 14 percent amalgam, unsaturated cell

Upper limit for 1 2 Va percent amalgam, unsaturated cell
Upper limit for 10 percent amalgam, unsaturated cell
Transition temperature for CdSO-8/3 HjO = CdSCvH0
Upper practical limit recommended for standard cells
Suitable range for thermostated air boxes
Temperature at which primary standard of United States
is maintained; also upper limit for 6 percent amalgam
Lower limit for 14 percent amalgam
Weston Normal Cell has a nominal emf of 1.0186360 abso
lute volts
Lower limit for 12Vi percent amalgam
Lower practical limit for unsaturated cells
Emf of Weston Normal Cell is at a maximum; solubility
of CdSCv8/3 H2O is a minimum
Lower limit for 10 percent amalgam
Freezing begins
Cell completely frozen; will recover normal emf in about
a week after thawing, if not cracked

the number of equivalents involved in the reaction,

F is the faraday, and k is a conversion factor,
0.101325, which converts cubic centimeter-atmos
pheres into joules, i.e., volt coulombs. The
volume change may be calculated from the atomic
or molecular weights and the densities of the reactants and products of the reaction; for saturated
standard cells this reaction is given by eq (7). At
20 C, the accepted densities of Cd, HgzSCu,
CdSCv8/3 H2O, Hg, and a saturated solution of
CdS04 are, respectively, 8.648, 7.56, and 3.090 g
cm"3 and 13.5463 and 1.6119 g ml-' [113]. Using
these data, accepted atomic weights on the C12
scale [114] and 96,487 coulombs/gram-equivalent
[115] ,s for the faraday equation 19 gives 6.4 ptV/atm
for (dEldP)T- This value agrees excellently with
6.1 piV/atm found experimently by Ramsey [116]
(this value is an interpolated value, Ramsey found
6.02 jiV/atm at 20.4 C and 7.6 pV/atm at 19 C).
The effect of pressure on an unsaturated cell
would be nearly the same, since unsaturated cells
are made with solutions that are nearly saturated.
A priori no effect on the emf of sealed cells
would be expected providing the pressure was
insufficient to fracture the cell. Recent experi
ments by Catherine Law and D. N. Craig of this
Bureau, in which the external pressure on the cell
was increased to approximately 5 atm, has con
firmed this. Standard cells have an air space
above the electrolyte which would act as a cushion
to absorb pressure change even if the elasticity of
the glass container were such as to transmit pres
sure change to the cell components.
" In the original paper [1 15] values based on the physical and chemical scales of
atomic weights were given. The value given here is based on an atomic weight of
aOver relative to the C" scale of atomic weights, namely, 107.870.
315-411 O - 68 - 8

7.5. Internal Resistance

The internal resistance in the vicinity of 25 C
ranges from 100 to 500 SI for unsaturated cells and
from 500 to 1000 SI for saturated cells manufactured
in the United States. The internal resistance
increases at the rate of about 2 percent for a
decrease in temperature of one degree C; this fact
should be taken into account in calculating the IR
drop in a cell (see table 19). The resistances of
the positive and negative limbs of symmetrical
H-shaped cells are approximately equal. As a cell
ages its internal resistance increases slightly. The
internal resistance of a standard cell may be
estimated by momentarily placing a 10-M11 resistor
across the cell terminals and reading the emf, Er.
The value of the internal resistance is then given by
R = 107(E0 - Er)IEr ohms,

(20)

where E0 is the open-circuit emf. The cell recovers

its initial emf within a few minutes after the 10-Mfl
resistor is removed. If a cell exhibits high internal
resistance or insensitivity it may contain a gas
bubble at the anode which may be frequently re
moved by tapping the cell when inclined 45.
If this treatment is ineffective a new cell is
recommended.
7.6. Effect of Current
Standard cells are not intended to serve as
sources of electric current. Even so, the question
frequently arises about the effect of current on
standard cells, especially the unsaturated type.
When standard cells are charged or discharged the

103-28

Table 19. Effect of current on the emf of

unsaturated standard cells at 25 C

The magnitude of V, obtained on discharge at

25 C is given by
Ve (in volts) :

0.161 It
FW

-1.67X 10"6

It
W
(23)

where W is now the weight of the electrolytic solu

tion in grams and the other symbols have the mean
ing given above. For unsaturated cells the weight
of solution is about 100 and 20 g, respectively, for
cells having internal resistances of 100 and 500 ft.
Equation (23) then becomes

owing to changes in amalgam ( Ka) and electrolyte

\V,) composition (after one year)
V,

V.

V.

Fa. (in volts) = -1.67xl0-*/f

io-
10-*
1010 -
io-
10 >

0.05
0.5
5.3

for the 100-ft cell and

0.03
0.26
2.6
26.3

0.02
0. 1 J

Ve (in volts) =- 8.35 X lO"8 It

Changes arising from electrode po

larization (after one year)

10""
io-
10 -
10"'
io-
10 -*

*y
0.1
0.7
4.3
26.
1S9.
970.

(24)

(25)

for the 500-ft cell.

The magnitudes of electrode polarization 16 during
discharge may be calculated, within a few micro
volts, by the equation

f-V
0.3
2.0
12.5
76.
466.
2844.

log AE (in microvolts) = 1.6048 + 0.786 log

(26)

"dash means that the change is

less than 0.01 /V.
rise or drop in emf is initially dictated by the IK
drop and subsequently by a voltage change, Vc,
associated with the chemical changes in the cell,
and by electrode polarization and changes in internal
resistance. For the unsaturated cell Vc is made up
of two parts: Va, the change in emf associated with
the change in the composition of the amalgam during
charge or discharge and Ve, the change in emf
associated with the change in electrolyte content of
the solution during charge or discharge. When
unsaturated cells have been discharged until their
solutions are saturated the latter no longer applies.
For saturated cell, Vc consists of Va only.
The magnitude of Va obtained on discharge at
25 C is given by
v ..
,
-0.001067/<
11(Wtw,n-Bft
Va (in volts) =
= 1-106 X 10 8 -jy
(21)
where / is the current in amperes, t is time in
seconds, W is the total weight of the amalgam in
grams, and F is the faraday (96487.0 coulombs).
For NBS saturated cells the total weight of amalgam
is generally about 10 g. In this case eq (21) reduces
to
Va (in volts) = - 1.106 X 10~9 It.
(22)
For unsaturated cells having internal resistances of
100 and 500 ft the weight of the amalgam is approxi
mately 67 and 20 g, respectively.

where A is the change in emf, in microvolts,

arising from electrode polarization and changes in
internal resistance, C is quantity of electricity in
coulombs (ampere seconds) and A is apparent
(geometric) electrode area in square centimeters
[117]. For unsaturated cells having internal
resistances of 100 and 500 ft, A is about 5.5 cm2 and
1.4 cm2, respectively. Some miniature types having
internal resistances of about 1000 ft and apparent
surface areas of less than 1 cm2 are also available.
The length of time standard cells will sustain a
discharge depends on the amount of material in the
cells; most unsaturated cells of United States
manufacture contain sufficient material to yield
from 700 to 5000 coulombs if the current is kept
below about 2 X 10-5 A per cm2 of electrode area.
When an external load is removed, the cells re
cover their initial emf, provided the discharge has
not been a prolonged one. The time required for
recovery depends on the severity of the discharge.
For example, if an unsaturated cell (internal re
sistance as 100 ft) is discharged for 5 min at 6 X 10~6
A cm-2 it will recover its original emf within 5 fiV
in 30 min and will completely recover after 6 hr.
At a higher current density of 6 X 10~4 A cm-2 the
emf will be about 180 fi\ below its original value
after 30 min, 5 /xV after 6 hr, and several days will
be required for complete recovery. If the cell
were discharged at a current density of 6 X 10~4 A
cm-2 to a low cutorT voltage of 0.001 V, recovery will
be exceedingly slow requiring several months and
16 Includes any change in internal resistance that may occur as a cell is discharged.

104-29

full recovery will not be attained after this prolonged

period because the normal emf will have declined
(see below). The time of recovery, therefore, is
seen to depend on the rate and extent of the
discharge.
In table 19 changes in emf caused by various cur
rents of low magnitude are illustrated for cells hav
ing electrode cross-sectional areas of 5.5 cm2 and
1.4 cm! and internal resistances of 100 and 500 CI.
The changes arising from internal resistance are
initial changes whereas those arising from changes
in amalgam composition, Va, electrolyte composi
tion, Ve, and electrode polarization (see footnote 16),
AE, are functions of time of discharge (for illustra
tion, a period of 1 year is chosen for these). The
total changes in emf during a discharge is the sum
of the four effects. On a current density basis it
should be noted the changes for the 100-fl and 500ft cells are nearly identical since the electrode area
for the former cell is approximately five times that
of the latter.
Standard cells may be short-circuited momentarily
without permanent damage to the cells. The cells
will recover their original emf within a few minutes
after taken off short circuit. If kept on short
circuit they will be completely discharged within
V2 to 2 days depending on the size and internal
resistance of the cell, and will not recover their
initial emf. The short circuit current is given by
the ratio of the open-circuit emf and the internal
resistance of the cell. For cells having internal
resistances of 100 ft and 500 ft, the short-circuit
(flash) current will, therefore, be 1X10-2 and
2 X 10" 3 A, respectively.
7.7. Effect of Light
Mercurous sulfate is sensitive to light and changes
in color at a slow rate through tan, to gray-brown,
to dark brown, and finally to black. Although
standard cells having discolored mercurous sulfate
may have normal emfs [118] they exhibit slower
approach to equilibrium values after temperature or
other changes. Standard cells should, therefore,
be mounted in nontransparent cases or kept in the
dark and used only for short periods at a time under
diffuse light.
7.8. Effect of Shock
Mechanical shocks insufficient to fracture or

break or scramble the components of unsaturated

standard cells have no lasting effects on the cells.
Unsaturated standard cells packaged in excelsior
and shipped by common carrier to the National
Bureau of Standards have been observed to perform
satisfactorily. When subjected to shocks of 10 to
40 g for durations of 6 to 18 msec unsaturated cells
exhibit large transient changes in emf ranging from
4,200 to 31,000 /liV [119]. After the shock the cells
immediately recover their original emf within 2 fiV.
The transient emfs observed during shock probably
arise from a disturbance of the mercury and
amalgam surfaces during the period of shock.
On the other hand, the usual types of saturated
cells of United States manufacture should not be
subjected to sudden shock, should not be shipped
by common carrier, should be transferred by
messenger, and should not be tilted more than 45.
Some new saturated standard cells of novel design
are stated to be portable, i.e., may be shipped by
common carrier. However, studies over a period
of time will be required to ascertain the long-term
stability of their emf.
7.9. Effect of Vibration
Vibrations at frequencies from 10 to 1,000
Hz (c/s) with accelerations of 1 to 10 g have no
lasting effects on the emf of unsaturated standard
cells [119]. During the vibration, however, rather
large a-c voltages of the same frequency are
generated. For an unsaturated cell having an in
ternal resistance of 500 fl this a-c voltage ranges
from about 25 /xV at 1 g and 1000 Hz to 9900 /*V
at 10 g and 50 Hz. Furthermore, there is a decrease
in the d-c emf ranging from 3 /xV at 1 g and 1000 Hz
to about 200 /aV at 10 g and 100 Hz. In general at
frequencies above 100 Hz the waveform of the a-c
voltage is sinusoidal whereas below 100 Hz it is
nonsinusoidal owing to the resonance of the various
components of the cell. In most cases the a-c and
d-c effects of the vibration appear and disappear
instantaneously when vibration is started or stopped.
In some instances the d-c change may be rapid in
the initial moments of vibration and then build
up slightly in an exponential manner for 2 or 3 min.
In these instances when the vibration is stopped the
d-c emf decays in the same fashion as it was built up.

8. Life of Standard Cells

Saturated standard cells have an exceedingly long
electrolyte, these cells on the average reach an
emf of 1.01830 V within 23 to 37 years, providing
life. Some cells at the National Bureau of Stand
ards have been in use for nearly 60 years, and they
they are maintained at 25 C or thereabouts, and are
have retained their emfs within a few microvolts;
not subjected to abuse, such as discharging or
see section 2.3. On the other hand, unsaturated
charging current; a practical life time for these cells
cells at room temperature (about 25 C) decrease in
is probably 12 to 18 years. When unsaturated
emf at a rate of about 20 to 40 piV per year [48].
standard cells reach an emf of 1.01830 V or lower
This decrease in emf is equivalent to a corrosion
the cells generally behave erratically (largely be
rate for the amalgam of about (6.8 to 13.6) X 10~7 A
cause the electrolyte may become supersaturated
cm-2. Since the emf of new unsaturated standard
on cooling), have large emf-temperature coeffi
cells generally range from 1.01900 to 1.01940 V,
cients, and show excessive emf-temperature hys
teresis. The life of the cell is considerably reduced
depending on the concentration and acidity of the
105-30

if the cell is stored at higher temperatures. The

rate of decrease in emf is approximately doubled
for every 12 C increase in temperature [120]; thus
at 37 C the life of an unsaturated cell would be one
half that given above. Cells of the miniature type
with short diffusion path between electrodes will
have a shorter life.

Earlier unsaturated standard cells decreased in

emf at room temperature at a rate of 70 to 85 fiV
per year [121, 122, 123] and, therefore, had an aver
age theoretical life of 10 to 15 years or a practical
life of 8 to 12 years. The improvement in life noted
in modern cells has resulted mainly, if not entirely,
from the use of improved septa in the cells.

9. References
[1] Giovanni Ciorgi, Unita razionali di electtromagnetisimo,
Atli dell' Assoc. Elettr. Ital. 5, 402 (1901); Proposal
concerning electrical and physical units. Trans. Intern.
Electrical Congr. 1, 136 (St. Louis, Mo., 1904).
[2] Wilhelm (E.) Weber, Messungen galvanische Leitungswiderstande nach einem absoluten Maasse, Ann. Physik.
(Pogg. Ann.) 2, 82 , 337 (1851); see also ref. [4].
[3] Reports of the Comm. on Electrical Standards of the Brit.
Assoc. Adv. of Science (reprinted by Cambridge Univ.
Press., 1913), Second Report, p. 60, Newcastle-on-Tyne,
1863.
[4] J. C. Maxwell, A treatise on electricity and magnetism,
3d ed., Vol. 2, Ch. 10, p. 263, Clarendon Press (Oxford,
1892).
[5] Ref. [3], First Report, p. 9, Cambridge, 1862.
[6] Report of the 43d meeting of the British Association for
the Advancement of Science, Bradford, First report of
the Committee for the Selection and Nomenclature of
Dynamical and Electrical Units, Brit. Assoc. Report,
p. 222, 1873.
[7] Oliver Heaviside, The relations between magnetic force
and electric current. I., The Electrician (London) 10,
6 (1882).
[8] F. B. Silsbee, Establishment and maintenance of the
electrical units, NBS Cir. 475 (1949); Systems of elec
trical units, NBS Mono. 56, (1962); J. Res. NBS 66C,
(Eng. & Instr.), 137 (1962).
[9] C. F. Gauss, "Intensitas vis magneticae terrestris ad
mensuram absolutam revocata," read before Royal
Society of Gottingen on December 15, 1832, Commentationes Societatis regiae Scientarum Gottingensis
recentiores, VIII, pp. 3^4 (1832-1837), Gottingen, 1841;
reprinted in Carl Friedrich Gauss Werke, V, pp. 79-118,
(Gottingen, 1867; 2nd printing, Gottingen, 1877), German
translation in PoggendorfTs Ann. Physik., XXVIII,
241-273, 591-615 (1833); second German translation by
E. Dorn, Ostwald's Klassiker der exakten Wissenschaften
No. 53, Leipzig, 1894.
[10] Sir William Thomson (Lord Kelvin), On the mechanical
theory of electrolysis, Phil. Mag. (4) 2, 429 (1851);
Practical electricity by W. E. Ayrton and T. Mather,
p. 205, Cassell and Co., Ltd., London (1911); Brit. Assoc.
reports on electrical standards, p. 137, Cambridge Univ.
Press, 1911.
[11] A. E. Kennelly, Historical outline of the electrical units,
J. Eng. Education 19, 229 (1928) reviews these develop
ments.
[12) Sir William Thomson (Lord Kelvin), Papers on electro
statics and magnetism, Macmillan and Co., Ltd., London,
p. 246, 1884.
[13] J. W. Trischka, U.S. Patent No. 2,959,683, Electric reso
nance voltage standards, Nov. 8, 1960; for theory see
C. H. Townes and A. L. Schawlow, Microwave spectro
scopy, Ch. 10, McGraw-Hill Book Co., Inc., New York,
1955.
[14] Y. Beers and G. L. Strine, The measurement of voltage by
the use of the Stark effect, IRE Trans. Instrumentation
III, Nos. 3 and 4, 171 (1962).
[15] F. Wenner, A proposed modification of the Kirchhoff
method for the absolute measurement of resistance, by
title only. Science 29, 475 (1909); see ref. [16] for details.
[16] J. L. Thomas, C. Peterson, I. L. Cooler, and K. K. Kotter,
An absolute measurement of resistance by the Wenner
method, J. Res. NBS 43, 291 (1949) RP2029.
[17] H. L. Curtis, C. Moon, and C. M. Sparks, A determination

[18]
[19]

[20]
[21]
[22]
[23]

[24]

[25]

[26]
[27]
[28]
[29]
[30]

[31]
[32]

[33]

[34]
[35]
[36]

106-31

of the absolute ohm using an improved self-inductor,

J. Res. NBS 21, 375 (1938) RP1137.
R. D. Cutkoskv, Evaluation of the NBS unit of resistance
based on a computable capacitor, J. Res. NBS 65A
(Phys. and Chem.) 147 (1961).
R. L. Driscoll and R. D. Cutkosky, Measurement of current
with the National Bureau of Standards current balance,
J. Res. NBS 60, 297 (1958) RP2846: R. W. Curtis, R. L.
Driscoll, and C. L. Critchfield, An absolute determination
of the ampere, using helical and spiral coils, J. Res.
NBS 22, 485 (1939) RP1200; see also H. L. Curtis.
Electrical measurements, Ch. XIX, McGraw-Hill Book
Co., Inc., New York, 1937.
R. L. Driscoll, Measurement of current with a Pellat-type
electrodynamometer, J. Res. NBS 60, 287 (1958) RP2845.
P. L. Bender and R. L. Driscoll, A free precession determi
nation of the proton gyromagnetic ratio, IRE Trans.
Instr. 1-7, Nos. 3 and 4, 176 (1958).
J. L. Thomas, Precision resistors and their measurement
NBS Circ. 470 (1948).
Latimer Clark, On a voltaic standard of electromotive
force, Proc. Royal Soc. (London) 20, 444 (1872); On a
standard voltaic battery, Phil. Trans. Royal Soc. I,
164, 1 (1874).
Lord Rayleigh and Mrs. H. Sidgwick, On the electro
chemical equivalent of silver, and on the absolute
electromotive force of Clark cells, Phil. Trans. IK. 175,
411 (1884).
H. S. Carhart, Relation between the electromotive force
of a Daniel! cell with the strength of the zinc sulfate
solution, Am. J. Sci. (3) 28, 374 (1884); Ref. [3], nine
teenth report, Edinburgh, 1892.
K. Kahle, Beitrage zur Kenntniss der elektromotorischen
Kraft des Clark'schen Normalelementes, Z. Instrumentenkunde 12, 117(1892).
R. T. Glazebrook and S. Skinner, On the Clark cell as a
standard of electromotive force, Proc. Royal Soc.
(London) 51, 60 (1892).
H. S. Carhart, The various determinations of the e.m.f. of
the Clark cell, Phys. Rev. 12, No. 3, 129 (1900).
Lord Rayleigh, On the Clark cell as a standard of electro
motive force, Phil. Trans. II, 176, 781 (1885).
Edward Weston, Normalelement, German Patent 75,194
(Jan. 5, 1892); Improvements in voltaic cells, British
Patent 22,482 (Feb. 6, 1892); Voltaic cell, U.S. Patent
494,827 (Apr. 4, 1893).
Report of International Conference on electrical units and
standards, Oct. 12, 1908; given as Appendix 2 in ref. [8].
F. A. Wolff, The temperature formula of the Weston
standard cell, Bull. BS 5, 309 (1908); data were submitted
in October 1908 to the London International Conference
on electrical units and standards.
Announcement of a change in the value of the Interna
tional Volt, NBS Circ. 29, p. 9, footnote 1, Dec. 31, 1910;
records of NBS carried results of emf of saturated stand
ard cells to 6 decimals from 1911 until 1934 and to 7
decimals since then.
G. W. Vinal, International comparison of electrical stand
ards, BS J. Res. 8, 729 (1932) RP448.
Announcement of changes in electrical and photometric
units, NBS Circ. 459, 1947.
L. H. Brickwedde and G. W. Vinal, Electromotive force of
saturated Weston standard cells containing deuterium
oxide, J. Res. NBS 20, 599 (1938) RP1094; Relation of
electromotive force to the concentration of deuterium

oxide in saturated standard cells, J. Res. NBS 27,

479 (1941) RP1435.
[37] E. B. Rosa, G. W. Vinal, and A. S. McDaniel, The silver
voltameter Part IV, Third series of quantitative ex
periments and special investigations, BS Bull. 10,
475 (1914), Table 3 (experiments made from 1910 to 1912,
inclusive).
[38] P. Vigoureux, International determination of the electro
motive force of the Normal Weston Cell in International
Volts, Coll. Res. Nat. Phys. Lab. 24, 79 (1932).
[39] H. von Steinwehr and A. Schulze, Neubestimmung der
EVIK des Internationalen Westonelements, Z. Instru
ment 52 , 249 (1932).
[40] H. L. Curtis and R. W. Curtis, An absolute determination
of the ampere, BS J. Res. 12, 665 (1934) RP685.
[41] E. B. Rosa, N. E. Dorsey, and J. M. Miller, A determina
tion of the International Ampere in absolute measure.
BS Bull. 8, 269 (1911).
[42] W. J. Hamer, L. H. Brickwedde, and P. R. Robb, Standard
cells and the unit of electromotive force, NBS Circ. 524,
Electrochemical constants, Ch. 12, 1953; results ob
tained since 1953 extend and modify the data given in
fig. 12.2 of this ref.
[43] M. Gouy, Sur une etuve a temperature constante, J. Phy
sique (3) 6, 479 (1897).
[44] F. A. Wolff and C. E. Waters, Clark and Weston standard
cells, BS Bull. 4, 1 (1907).
[45] P. H. Lowrie, Jr., Controlled temperature oil baths for
saturated standard cells, NBS Tech. Note 141, 1962.
[46] H. B. Brooks, The standard-cell comparator, a specialized
potentiometer, BS J. Res. 11, 211 (1933) RP586.
[47] J. H. Miller, A simplified standard cell comparator, AIEE
Communications and Electronics, No. 14, 413 (Sept.
1954).
[48] G. D. Vincent, The construction and characteristics of
standard cells, IRE Trans. Instr. 1-7, Nos. 3 and 4,
221 (1958).
[49] A. W. Spinks and F. L. Hermach, Portable potentiometer
and thermostatted container for standard cells, Rev.
Sci. Instr. 26, No. 8, 770 (1955).
150] Barbara A. Wickoff, The establishment and maintenance
of the unit of voltage at the National Bureau of Stand
ards-Boulder, ISA Preprint No. 28.2.63, 18th Annual
Conference and Exhibit, Chicago, Sept., 1963.
[51] A. C. Kolossov, Determination by the method of the silver
voltameter of the electromotive force of the International
Weston Normal Cell, Proces Verbaux, Comite inter
national des Poids et Mesures 16, 150 (1933) Annexe
No. 19.
[52] Appendix 3 of ref. [8] and p. 16 of ref. [8].
[53] Proces Verbaux, Comite International des Poids et Mesures
17, 94 (Resolution 4), 1935.
[54] J. F. Daniell, On voltaic combinations, Phil. Mag. Ill, 8,
421 (1836).
[55] W. R. Cooper. Primary batteries: their theory, construc
tion and use, 2d ed., Ch. IX, D. Van Nostrand Com. Ltd.,
London, 1916.
[56] G. W. Vinal, Primary batteries, pp. 32, 232, 233, John
Wiley & Sons, New York, 1950.
[57] Normal- Darnell-Element der Edison and Swan United
Electric Light Company (Fleming and W. Thomson),
Dingler's polytechnisches J. 258, 319 (1885); also Re
ports of Paris International Electrical Congress, 1882.
[58] W. De la Rue and H. W. Miiller, On a new form of constant
battery, J. Chem. Soc. 21, 488 (1868); Experimental
researches on the electric discharge with the chloride
of silver battery, Phil. Trans. Royal Soc. 1, 169, 55 (1878);
this second paper deals with the cell as a standard and
compares the cell with a Clark standard cell.
[59] H. Helmholtz, Die thermodynamik chemischer Vorgange,
Sitzungsber. der Kbniglich Preussischen Akad. der
wissenschaften zu Berlin, p. 27, 1882; see also ref. [75],
p. 11, 74; W. Ostwald, Das kompensations-elektrometer,
Z. phys. Chem. 1, 403, 1887.
[60] M. Gouy, Sur une pile etalon, Compt. Rendu. 104, 781
(1887), J. de phys. (2) 7, 532 (1888); Uber ein Normalelement, Z. physik. Chem. 2, 978 (1888).

[61] Edward Weston, Galvanic battery, U.S. Patent 310,004

(Dec. 30, 1884).
[62] H. S. Carhart, An improved standard Clark cell with low
temperature coefficient, Phil. Mag. 28, 420 (1889), Am.
J. Sci. (3) 38, 402 (1889); Relation between the electro
motive force of a Clark cell and the density of the zinc
sulphate solutions, (Proc.) Am. Ind. Elect. Eng. 9, 615
(1892).
[63] J. N. Bronsted, Studien zur chemischen affinitat. V. Die
bildung des kaliumbleisulfats, Z. physik. Chem. 77, 315
(1911).
[64] W. E. Henderson and G. Stegeman, A lead standard cell
and a determination of the potential of the lead electrode,
J. Am. Chem. Soc. 40, 84 (1918).
[65] W. C. Vosburgh, M. Guagenty, and W. J. Clayton, Satu
rated standard cells with small temperature coefficients.
II, J. Am. Chem. Soc. 59, 1256 (1937).
[66] W. C. Vosburgh and R. G. Bates, Constancy of modified
Weston standard cell over long periods, J. Electrochem.
Soc. Ill, 997 (1964).
[67] H. L. Callendar and H. T. Barnes, On the variation of the
electromotive force of different forms of the Clark stand
ard cell with temperature and strength of solution, Proc.
Royal Soc. (London) 62, 117 (1897).
[68] G. A. Hulett, The constancy of standard cells and a con
stant temperature bath, Phys. Rev. 32, 257 (1911); The
construction of standard cells and a constant tempera
ture bath. The Electrician 67, 129 (1911), see also ref.
[91].
[69] C. R. A. Wright and C. Thompson, On the determination
of chemical affinity in terms of electromotive force.
Part VII. On the electromotive force of Clark's mercurous-sulphate cell; on the work done during elec
trolysis. On the e.m.f. of Clark's cell, Phil. Mag. V, 16,
25 (1883).
[70] K. Kahle, Instructions for preparing Clark standard cells,
The Electrician 31, 265 (1893).
[71] Ref. [55], p. 336.
[72] International Critical Tables, Vol. 2, p. 436, 1927.
[73] W. J. Hamer, Hydrolysis of aqueous solutions of cadmium
and zinc sulfates at 25 C, unpublished, NBS.
[74] K. Kahle, Zur behandlung des silvervoltameters und seine
verwendung zur bestimmung von Normalelementen, Ann
Physik. (or Wied. Ann.) (3) 67, 1 (1899), Z. Instrk. 18,
230 (1898).
[75] W. Jaeger, Die Normalelement und ihre anwendung in der
elektrischen messtechnick, Wilhelm Knapp, Halle a.S.,
1902.
[76] G. A. Hulett, Equilibria in standard cells, Phys. Rev. 27,
337 (1908).
[77] G. A. Hulett, Die destination von amalgamen und die
reinigung des quecksilbers, Z. physik. Chem. 33, 611
(1900); see also G. A. Hulett and H. D. Minchin, Dis
tillation of amalgams and purification of mercury, Phys.
Rev. 21, 388(1905).
[78] C. L. Gordon and E. Wichers, Purification of mercury and
its physical properties, Ann. N.Y. Acad. Sci. 65, 369
(1957).
[79] L. H. Brickwedde, Solubility of cadmium sulfate in HjOD20 mixtures, J. Res. N,BS 36, 377 (1946) RP1707.
[80] F. Mylius and R. Funk, Uber die hydrate des cadmiumsulfats, Ber. deut. chem. Ges. 30, 824 (1897).
[81] Ph. Kohnstamm and E. Cohen, Physikalisch-chemische
studien am Normalelement von Weston, Ann. Physik.
(or Wied. Ann.) (3) 65, 344 (1898).
[82] International Critical Tables, Vol. 1, p. 120, 1926.
[83] International Critical Tables, Vol. 2, p. 429, 1927.
[84] F. E. Smith (Sir Frank), On the preparation of a cadmium
ceU, The Electrician 55, 857 (1905).
[85] G. Deniges, Sur de nouvelles classes de combinaisons
mercurico-organiques et sur leurs applications, Ann.
chim. phys. [7] 18, 382 (specifically 397) 1899, see also
ref. 102.
[86] F. Laporte, Report to the International Committee on elec
trical units and standards, p. Ill, Washington, D. C,
Jan. 1, 1912.
[87] D. N. Craig, G. W. Vinal, and F. E. Vinal, Solubility of

107-32

mercurous sulphate in sulphuric-acid solutions, J. Res.

NBS 17, 709(1936) RP939.
[88] M. Gouy, Action de l'eau sur le sulfate mercureux, Compl.
rendu. 130, 1399 (1900).
[89] O. B. Hager and G. A. Hulett, The hydrolysis of mercurous
sulphate, J. Phys. Chem. 36, 2095 (1932).
[90] L. H. Brickwedde, Method of making saturated cells at
the National Bureau of Standards, Compt. rendu, de la
Quatorzieme Conference, Int. Union Pure and Applied
Chem. p. 105, July 1947.
[91] G. A. Hulett, A study of the materials used in standard
cells and their preparation. Part 1. Mercurous sulphate
and standard cells. Trans. Am. Electrochem. Soc. 6,
109 (1904), Trans. Int. Elec. Cong., St. Louis 2, 109
(1904).
[92] F. E. Smith (Sir Frank), A dictionary of applied physics,
p. 267, ed. Sir Richard Glazebrook, Macmillan and Co.,
Ltd., London, 1911.
W) P. H. Walker and F. W. Smither, Comparative tests of
chemical glassware, J. Ind. Eng. Chem. 9, 1090 (1917).
Tech. Pap. BS 10 (1918) T107.
[94] E. Wichers, A. N. Finn, and W. S. Claybaugh, Compara
tive tests of chemical glassware, J. Res. NBS 26, 537
(1941) RP1394.
[95] H. Fleischer, Chemical indicator, U.S. Patent 2.416,619
(Feb. 25, 1957); Sensitive indicator for volumetric deter
mination of boiler feedwater alkalinity, Ind. Eng. Chem.,
Anal. Ed. 15, 742 (1943); the author did not specifically
call his indicator "methyl purple" but the indicator he
suggested later became to be known by this name; see
also. Water Works Manual, 1947^8 Ed.
[96] G. W. Vinal, L. H. Brickwedde, and W. J. Hamer, New
quartz containers for standard cells at the National Bu
reau of Standards, Compt. rendu, de la Quinizieme Con
ference, Int. Union Pure and Applied Chem., p. 92,
Amsterdam, Sept. 1949.
[97] W. G. Houskeeper, The art of sealing base metals through
glass. J. Am. Inst. Elec. Eng. 42, 954 (1923).
[98] C. A. Dyer, Primary electrolytic cell or group of cells (to
Minneapolis-Honeywell Regulator Co.), U.S. Patent
2,647,155 (July 28, 1953).
[99] R. O. Heinrich, Standard voltaic cell, U.S. Patent 631,044
(Aug. 15, 1899).
[100] H. Eicke and A. Kessner, Eine stabfiirmige Bauart des
Weston-Normalelementes, Electrotech. Z. (A) 74, 623
(1953).
[101] E. F. Mueller and H. F. Stimson, A temperature-control
box for saturated standard cells, J. Res. NBS 13, 699
(1933) RP739.
[102] W. C. Vosburgh and M. Eppley, The temperature co
efficients of unsaturated Weston cells, J. Am. Chem. Soc.
45, 2268 (1923).
[103] Ref. [92], p. 268.

[104] Report of Natl. Phys. Lab., Electrician 75, 463 (1915).

[105] J. Obata, Cadmium standard cells containing acid elec
trolyte, Proc. Math. Phys. Soc. (Japan), (3), 2, 232 (1920).
[106] V. Ishibashi and T. Ishizaki, Further studies on acid stand
ard cells. Researches Electrotechnical Lab., No. 318,
1931.
[107] W. C. Vosburgh, Conditions affecting the reproducibility
and constancy of Weston standard cells, J. Am. Chem.
Soc. 47, 1255 (1925).
[108] F. E. Smith (Sir Frank), On cadmium amalgams and the
Weston Normal Cell, Proc. Phys. Soc. (London) 22,
11 (1910); Phil. Mag. 19, 250 (1910); Natl. Phys. Lab.
Coll. Res. 6, 137 (1910).
[109] G. W. Vinal and L. H. Brickwedde, Metastability of cad
mium sulfate and its effect on electromotive force of
saturated standard cells, J. Res. NBS 26, 455 (1941)
RP1389.
[110] P. Vigoureux and S. Watts, Temperature coefficient of
Weston cells, Proc. Phys. Soc. (London) 45, 172 (1933).
[Ill] J. H. Park, Effect of service temperature conditions on the
electromotive force of unsaturated portable standard
cells, BS J. Res. 10,89(1933) RP518.
[112] M. Eppley, International standard of electromotive force
and its low-temperature-coefficient form. Trans. AIEE
50, No. 4 1923 (1931).
[113] International Critical Tables, Vol. 1, CdSO, 8/3 HtO
(p. 120), Hg,S04 (p. 121); Vol. 2, Cd (p. 456), Hg(p. 458);
Vol. 3, sat. sol. CdSO, interpolated (p. 66).
[114] Table of atomic weights, 1961 (Based on carbon-12),
Chem. and Eng. News., p. 43, Nov. 20, 1961.
[115] D. N. Craig, J. I. Hoffman, C. A. Law, and W. J. Hamer,
Determination of the value of the faraday with a silverperchloric acid coulometer, J. Res. NBS 64A (Phys.
and Chem.) 381 (1960).
[116] R. R. Ramsey, The change of volume in Clark and cadmium
cells and its relation to change of electromotive force
due to pressure, Phys. Rev. XVI, No. 2, 105 (1902).
[117] Based on measurements made at Natl. Bur. Standards
by Phyllis R. Robb and W. J. Hamer, 1950-52; see also
ref. [120].
[118] G. A. Hulett, Mercurous sulphate and the standard cell,
Phys. Rev. 22, 321 (specially 336, footnote 1) 1906.
[119] R. J. Brodd and W. G. Eicke, Jr., Effect of vibration and
shock on unsaturated standard cells, J. Res. NBS 66C
(Eng. and Instr.) 85 (1962).
[120] G. D. Vincent, Effect of current drains on cadmium stand
ard cells, J. Electrochem. Soc. 104, 712 (1957).
[121] G. W. Vinal, D. N. Craig, and L. H. Brickwedde, Standards
of electromotive force, Trans. Electrochem. Soc. 68,
139 (1935).
[122] Ref. [56], p. 211.
[123] F. X. Lamb. Aging of standard cells, NBS Circ. 524,
Electrochemical constants, Ch. 11, 1953.

10. Appendix 1
U.S. Law of 1894, 53d Congress, 28 Stat., Ch. 131, p. 102 (Public-No. 105)
An Act To define and establish
Be it enacted by the Senate and House of Repre
sentatives of the United States of America in
Congress assembled, That from and after the pass
age of this Act the legal units of electrical measure
in the United States shall be as follows:
First. The unit of resistance shall be what is
known as the international ohm, which is sub
stantially equal to one thousand million units of
resistance of the centimeter-gram-second system
of electro-magnetic units, and is represented by
the resistance offered to an unvarying electric cur
rent by a column of mercury at the temperature of
melting ice fourteen and four thousand five hundred
and twenty-one ten-thousandths grams in mass, of
a constant cross-sectional area, and of the length

the units of electrical measure

of one hundred and six and three tenths centi
meters.
Second. The unit of current shall be what is
known as the international ampere, which is onetenth of the unit of current of the centimeter-gramsecond system of electro-magnetic units, and is
the practical equivalent of the unvarying current,
which, when passed through a solution of nitrate
of silver in water in accordance with standard
specifications, deposits silver at the rate of one
thousand one hundred and eighteen millionths of
a gram per second.
Third. The unit of electro-motive force shall be
what is known as the international volt, which is
the electro-motive force that, steadily applied to a

108-33

conductor whose resistance is one international

ohm, will produce a current of an international
ampere, and is practically equivalent to one thou
sand fourteen hundred and thirty-fourths of the
electro-motive force between the poles or electrodes
of the voltaic cell known as Clark's cell, at a tem
perature of fifteen degrees centigrade, and prepared
in the manner described in the standard
specifications.
Fourth. The unit of quantity shall be what is
known as the international coulomb, which is the
quantity of electricity transferred by a current of
one international ampere in one second.
Fifth. The unit of capacity shall be what is
known as the international farad, which is the ca
pacity of a condenser charged to a potential of one
international volt by one international coulomb of
electricity.
Sixth. The unit of work shall be the Joule, which
is equal to ten million units of work in the cen
timeter-gram-second system, and which is prac

tically equivalent to the energy expended in one

second by an international ampere in an interna
tional ohm.
Seventh. The unit of power shall be the Watt,
which is equal to ten million units of power in the
centimeter-gram-second system, and which is prac
tically equivalent to the work done at the rate of
one Joule per second.
Eighth. The unit of induction shall be the Henry,
which is the induction in a circuit when the electro
motive force induced in this circuit is one inter
national volt while the inducing current varies at
the rate of one Ampere per second.
Sec. 2. That it shall be the duty of the National
Academy of Sciences to prescribe and publish, as
soon as possible after the passage of this Act, such
specifications of details as shall be necessary for
the practical application of the definitions of the
ampere and volt hereinbefore given, and such speci
fications shall be the standard specifications herein
mentioned.

11. Appendix 2
Public Law 617 81st Congress
(Chapter 484 2d Session)
(S. 441)
AN ACT
To redefine the units and establish the standards
of electrical and photometric measurements.
Be it enacted by the Senate and House of Repre
sentatives of the United States of America in
Congress assembled, That from and after the date
this Act is approved, the legal units of electrical
and photometric measurements in the United States
of America shall be those defined and established
as provided in the following sections.
SEC. 2. The unit of electrical resistance shall
be the ohm, which is equal to one thousand million
units of resistance of the centimeter-gram-second
system of electromagnetic units.
SEC. 3. The unit of electric current shall be
the ampere, which is one-tenth of the unit of current
of the centimeter-gram-second system of electro
magnetic units.
SEC. 4. The unit of electromotive force and of
electric potential shall be the volt, which is the
electromotive force that, steadily applied to a con
ductor whose resistance is one ohm, will produce a
current of one ampere.
SEC. 5. The unit of electric quantity shall be
the coulomb, which is the quantity of electricity
transferred by a current of one ampere in one
second.
SEC. 6. The unit of electrical capacitance shall
be the farad, which is the capacitance of a capaci
tor that is charged to a potential of one volt by one
coulomb of electricity.
SEC. 7. The unit of electrical inductance shall
be the henry, which is the inductance in a circuit
such that an electromotive force of one volt is

induced in the circuit by variation of an inducing

current at the rate of one ampere per second.
SEC. 8. The unit of power shall be the watt,
which is equal to ten million units of power in the
centimeter-gram-second system, and which is the
power required to cause an unvarying current of
one ampere to flow between points differing in
potential by one volt.
SEC. 9. The units of energy shall be (a) the
joule, which is equivalent to the energy supplied
by a power of one watt operating for one second,
and (b) the kilowatt-hour, which is equivalent to
the energy supplied by a power of one thousand
watts operating for one hour.
SEC. 10. The unit of intensity of light shall be
the candle, which is one-sixtieth of the intensity
of one square centimeter of a perfect radiator,
known as a "black body", when operated at the
temperature of freezing platinum.
SEC. 11. The unit of flux of light shall be the
lumen, which is the flux in a unit of solid angle
from a source of which the intensity is one candle.
SEC. 12. It shall be the duty of the Secretary of
Commerce to establish the values of the primary
electric and photometric units in absolute measure,
and the legal values for these units shall be those
represented by, or derived from, national reference
standards maintained by the Department of
Commerce.
SEC. 13. The Act of July 12, 1894 (Public Law
Numbered 105, Fifty-third Congress), entitled "An
Act to define and establish the units of electrical
measure", is hereby repealed.
APPROVED July 21, 1950.

109-34

12. Appendix 3
Standard Cells with Deuterium Oxide
The electromotive forces of saturated standard
cells, made with deuterium oxide and normal water
follow:
Emf at 20 "C, volts
Percentage of DjO in the water
1.018603
0.02 "
1.018567
10
1.018531
20
1.018495
30
1.018459
40
1.018423
SO
1.018384
60
1.018344
70
1.018301
80
1.018255
90
1.018204
100 (extrapolated from
98 percent)
Normal water contains 0.02 percent deuterium oxide.
' Original data were reported in international volts and have here been
converted to absolute volts.
These cells were made with crystals of cadmium sulfate in
equilibrium with the respective solutions. This was accom
plished by saturating each solution with anhydrous cadmium
sulfate: the crystals that separated from the solution were then
in equilibrium with the saturated solution. The solutions were
then made 0.035 to 0.042 /V with respect to sulfuric acid; the above
emfs are reported on a uniform basis of 0.04 N.
Cadmium sulfate is about 8 percent less soluble in heavy water
than in normal water in the temperature range of 0 to 60 C.
This comparison follows:

Moles CdSOs per mole water

Temperature
"C

CdSO, -8/3 aq'

CdSCvaq"

normal water 4

heavy water

0
5
10
15
20
25
30
35
40
45
.50

0.06546
.06555
.06566
.06580
.06600
.06627
.06663
.06710
.06773
.06857
.06955

0.06026
.06033
.06042
.06055
.06073
.06097
.06130
.06174
.06232
.06311
.06415

45
50
55
60

.06797
.06659
.06522
.06385

.06326
.06181
.06037
.05893

" here aq means either HiO or DtO.

* contains 0.02 percent heavy water.
Studies of the changes in the difference between the emf of
cells made with normal and heavy water with time have given
supplemental evidence on the stability of the National Reference
Group of Standard Cells. These studies have shown that this
difference is within the spread, namely, 0.05 fi\/year, discussed
under item (1) on maintenance in section 2.3. However, cells
made with heavy water have not, to date, shown the same
internal consistency as those made with normal water, i.e.,
individual cells show greater day-to-day variations in emf than
normal water cells.

13. Appendix 4
by the United States in the Physikalisch-Technische Reichsanstalt at Chariot tenburg, 1931

Coulomeler (Smith form) r

Experiment Number"*
Item

A
B
C
D
E
F
C
H
J
K
L
M

4.09189
4.09172
7230.00
1.999989
0.OOO0O5
1.018301
1.018301
1.012582
1.018259
0.000011
1.018116
1.018187

10

4.09328
4.09317
7199.85
1.999973
0.000005
1.018429
1.018430
1.017114
1 018390
0.00001]
1.018299
1.018370

4.09600
4.09591
7200.06
1.999973
0.000005
1.018411
1.018402
1.018145
1.018694
0.000014
1.018234
1.018305

4.08556
4.08543
7199.84
1.999975
0.000005
1.018336
1.018336
1.015176
1.018296
0.000014
1.018229
1.018300

4.08615
4.08621
7200.01
1.999973
0.000005
1.018334
1.018334
1.015326
1.018293
0.000013
1.018220
1.018291

4.10120
4.10123
7200.16
1.999971
0.000005
1.018331
1.018330
1.019020
1.018288
0.000015
1.018238
1.018309

4.09495
4.09507
7200.06
1.999975
0.000005
1.018328
1.018328
1.017458
1.018283
0.000013
1.018266
1.018337

4.09105
4.09110
7199.84
1.999981
0.000005
1.018332
1.018330
1.016712
1.018283
0.000013
1.018070
1.018141

4.16814
4.16824
7320.05
1.999985
0.000005
1.018334
1.018332
1.018796
1.018286
0.000012
1.018144
1.018215

'Electrical
measurements
made
in PTR units;
- PTR)coulomeler,
ohm = 9 ppm;
(USA -notPTR)
volts = - 61 ppm.
Experiment
Number 6 was
conducted
with a (USA
porous-cup
therefore,
included.
A weight of silver deposit in one coulomeler, grams.
B weight of silver deposit in second coulomeler in series wilh A, grams.
C lime of electrolysis, seconds.
D resistance it of Mandard resistor, ohms.
E correction for resistance owing to leads, etc., ohm.
F average emf of control cell at 20 C, volts.
C emf of control cell at 20 *C at end of run, volts.
H IR m standard resistor, volts; / is current in amperes.
J emf of German reference cell at 20 *C, volls.
K correction 10 J to give mean value of Kleiner Stamm (primary volt standard of Germany), volt.
L German value of selected cell at 20 *C; given by

M-USA (USA units) value of selected cell at 20 *C; given by

*= 1.000070/. volts, since (USA-PTRI ohm =9 ppm and (USA-PTRl volts = -61 ppm
110-35

ave = 1.018202
ave= 1.018273

14. Appendix 5
Constant-Temperature Oil Baths
Constant-temperature oil baths of the type used
to maintain the National Reference Group of Stand
ard Cells at a constant temperature were shown in
figure 2. In this appendix a general description of a
typical bath is presented. A circular bath of stain
less steel or nickel-plated copper, about 75 cm in
diameter and 45 cm deep, may be used (all other
metal parts are also of steel or nickel-plated copper
to prevent the possibility of galvanic corrosion). It
is filled with oil (see Appendix 6) to a depth of about
35-40 cm. The oil is stirred by a centrally located
two-blade agitator which produces an upward and
rotary circulation. The agitator (9) is about 48
cm in length, 5 cm wide, 0.3 cm thick, has a pitch
of 20 and rotates at about 45 rpm. It is supported
on a shaft (1) which is supported in ball bearing
housings at the bottom and top of the bath (the one
at the top is supported in a metal slab (3), about 10
cm wide and 0.3 cm thick, which rests on the central
top edges of the bath. A drip cup (4) is placed on
the shaft at the position shown to collect any grease
that may flow down the shaft from the ball bearing
housing. To guide the direction of oil circulation
a metal baffle (5), 25 cm in height, is placed in the
bath in the position shown in figure Al. This baffle
extends to 10 -cm of the bottom of the bath and
slightly above the agitator. This baffle serves

also to protect the cells from direct influence of

the heating element and the cooling coils.
Heating is supplied by a heating coil (8) of bare
manganin wire wound on a Pyrex rod having a
diameter of 1 cm. This rod is supported about 5
cm above the bottom of the bath by glass supports
and circles the bath; its position is clearly shown
in figure Al. The heating coil has a resistance of
90 ohms and operates on 110-V d-c supply. A
bank of four incandescent lamps in parallel is
included in the heating circuit; two lamps operate
intermittently during control while two operate
continuously, i.e., some heating is provided con
tinuously and some intermittently. Cooling coils
(7), (11), as shown, are also provided. Through
these coils flow oil of the same type as used in
the bath. These coils are connected to similar
coils in the freezing compartment of a refrigerator
and the required cooling is attained by controlling
the rate of oil flow. Cooling by this means is re
quired only in emergencies or when tests of stand
ard cells are made at temperatures below room
temperature. The temperature of the bath is
determined by a platinum resistance thermometer
and a Mueller bridge; the thermometer is supported
at the position (2) shown in the figure. The lamps
in the circuit also aid an operator in determining
when the bath has attained a peak temperature
(one reading on Pt resistance thermometer) and a

FIGURE Al. Cross-sectional sketch of a constant-temperature oil bath used at the

National Bureau of Standards.
1 Agitator shaft
7 Cooling coils
2 Thermometer mount
8 Heater
3 -Center plate
9- Agitator
4 Drip cup
10 Mercury-toluene regulator
5- Baffle
11 Cooling coil connector
6 Cell rack
12 Fenwall thermuswitch
111-36

lower temperature (second reading on Pt resistance

thermometer) and thereby determine the mean
temperature of the bath (the period between peak
and low temperatures is about 15 sec).
A mercury-toluene regulator (10) is used as the
temperature-control element. It consists of a
reservoir in the form of a rectangle, 40 cm on an
edge, made of thin-walled Pyrex tubing, 1.5 cm in
diameter, containing the thermometric. substance
(toluene), the volume changes being transmitted by a
U-shaped connecting tube filled with mercury to a
vertical capillary as shown in the figure. The Ushaped connecting tube is provided with a down
ward extension from the toluene reservoir to allow
for expansion and contraction of the toluene thereby
preventing the toluene from reaching the capillary.
The capillary is about 1 mm in diameter and 8 cm
long and the adjustable electrical contact to the
mercury is made with a steel piano wire. The
permanent electrical terminal to the mercury is

made through a seal about 4 cm below the capillary

tube. This terminal is connected to the electro
magnet of a sensitive 150-fl relay, whereas the
adjustable terminal is connected to one terminal of
a long-life Lalande battery of 2.5 volts; the other
terminal of this battery is connected to the elec
tromagnet. The secondary of the relay is in the
110-volt d-c heater circuit. The regulator is sup
ported on the baffle about 10 cm from the bottom of
the bath. The capillary extends above the oil and
is therefore, exposed to the air; this exposure, how
ever, does not cause a significant error in regulation.
A Fenwall thermoswitch (12) is included in the heat
ing circuit as a precaution in case the regulator does
not function; this thermoswitch opens the heating
circuit at a preset value, usually 1 C, above the
temperature at which the bath is controlled. The
standard cells are mounted in the bath on the metal
rack (6) in the manner described in the text (sec. 2.3).
Other details on the bath are shown in figure Al.

15. Appendix 6
Oil for Constant-Temperature Baths
The quality of the oil in which standard cells are
submerged is of considerable importance, and for
this purpose a white mineral oil with the following
specifications is used:
Colorless, odorless, tasteless;
Viscosity, in poises at 25 C, 0.245*;
Specific gravity, 25 C, 0.846;
Flashpoint, 171 C (340 F);
Burning point, 207 C (405 F);

Acidity (mg KOH/gm oil), none;

Sulfur, none;
Oxidation number, 0;
Gum, on heating in oxygen, none;
Discoloration, on heating in oxygen, slight orange.
*Sufficiently low as to make the stirring effective.
In addition this oil, when new, has a resistivity
of about 3 X 10u ohm-cm or higher.

16. Appendix 7

Constant-Temperature Enclosure for

Saturated Standard Cells
The Mueller-Stimson temperature-control box
for saturated standard cells was described in gen
eral terms in section 5.4. In this appendix more
details are given. In this control box, the tem
perature of an outer aluminum case is automatically
controlled by a conventional mercury-in-glass
thermoregulator. A second aluminum case, inside
the first and thermally well insulated from it, con
tains five saturated standard cells. The inner
case assumes a temperature very nearly the time
average of that of the outer case. The inner case
is 35/8 in. long, 2*U in. wide, and 4 in. deep inside.
This case is a casting with V4-in. walls and 3/sin. bottom. A V4-in. -thick cover is secured to the
casting by numerous screws to improve thermal
contact. The inner case is spaced within the outer
one with V2 in. balsa wood on all six sides. The
outer case is also a casting, with 3/s in. sidewalls
and V2 in. bottom. A cover of sheet 3/s-in. alumi

num is secured to the top by screws. The endwalls

of the casting are plane on the inside and cylin
drical on the outside, the thickness at the edges
being 3/s in. and aU in. at the middle; this extra
thickness at the middle provides space for a vertical
hole, 15/32 in. in diameter and 6 in. deep, to ac
commodate the thermoregulator at one end and a
smaller hole to accommodate a thermometer at
the other end.
The outer case is insulated with a layer of balsa
wood 1 in. thick on the sides and bottom and 2 in.
thick on the top. The pieces of balsa wood fit
neatly into an outer wooden box. This box is of
sufficient size to provide a compartment at one end
to house a transformer, relay, and binding posts
for the necessary connections. These accessories
are mounted on a wooden panel which slides in
vertical grooves in the sides of the box. The out
side dimensions of the box are 8 by 13 in., by 10
in. deep, and its weight, complete, is 22 lb. A
horizontal section of the control box is shown in
figure A2.

112-37

FlCURE A2. Horizontal sketch of Mueller-Stimson temperaturecontrolled box.

The standard cells are mounted on balsa wood of
appropriate size. No. 28 (0.013 in.) insulated cop
per wire, 4 ft. in length, is used for connections to
the cell terminals. About one-third of this length
is formed in a helix which is kept inside the inner
case and another one-third is placed between the
inner and outer cases. The wires are brought
through the cases in smoothed saw cuts in the upper

edges of the sides and insulated additionally with

silk and glyptal whereby good thermal contact
with the metal cases is assured. These thermal
tiedowns and lengths of wire reduce heat conduc
tion to a negligible amount. Sheets of mica are
placed on the surfaces of the aluminum cases,
adjacent to the connecting wires, to protect them
against accidental electrical grounding. Outside
the outer case the wires are brought to binding
posts supported on a hard-rubber strip; the positive
terminal is placed on one side of the box, the nega
tive on the other.
The thermoregulator is an adjustable mercury-inglass type with a bulb 4 in. long about i5/a2 in. in
diameter, and 8 in. long. It is covered with grease
to assure good thermal contact to the reservoir
into which it fits.
The heating resistor, No. 38 (0.004 in.) constantan
wire, has a resistance of about 70 Q and is wound
in four turns on the sides and ends of the outer
aluminum case on silk fabric. One turn is placed
near the top edge and another near the bottom edge
of the case. It operates on 20 V. A small, quickacting 12-V a-c relay operating on about 0.05 A is
used. Power for the heater and relay is supplied
by a bell-ringing transformer rated at 50 W.

113-38

INSTRUMENTATION SYSTEMS AUTOMATIC CONTROL

Oil Baths for Saturated Standard Cells*

Patrick H. Lowrie, Jr.

National Bureau of Standards
Boulder, Colorado

versely affect the cells. As a result, more groups of

cells may be contained in a single bath when using
oil as the medium. In addition, provided the proper
precautions are taken, cells or groups of cells may be
introduced into or removed from the oil without dis
turbing the other groups in the same bath. This is
because there is no heat loss due to convection when
the lid of the bath is opened. Furthermore, better
INDUSTRY today is faced with the need for continu
temperature control can be achieved in an oil bath
ally increasing accuracy of measurement. One result:
than can be achieved in an air bath of equivalent size.
the increasing use of saturated standard cells in in
Much experience with oil baths has been gained in
dustrial standards laboratories. Because of their
the many years during which they have been in use
rather large temperature coefficient of emf, these
at NBS. (However, with the exception of very early
saturated cells' must be contained in a temperaturepapers,--3 the published information has been rather
co.itrolled environment to realize their inherent accu
sparse.) When it became necessary to establish and
racy. In answer to the growing demand for tempera
maintain an accurate standard of voltage at the NBS
ture-controlled baths, several companies are now
Boulder Laboratories, this experience was drawn
manufacturing baths appropriate for use with these
upon. Additional features were incorporated to ob
cells. Frequently, however, the user prefers to build
tain the desired temperature control and allow a
his own bath to include special features. In the design
large calibration workload to be handled convenient
of any enclosure of this type the two most important
ly. The accuracy desired in the measurement of cells
requirements are close temperature control and the
dictated a short-time temperature variation of no
control of gradients. No matter how elaborate the
greater than 0.001C, with day-to-day variations
design, if these two requirements are not met, the
held to no greater than 0.005C. It was desired that
bath will be inadequate.
temperature gradients be unmeasurably small be
The oil bath has many advantages over the air
cause of the extreme sensitivity of emf to tempera
bath. In oil the temperature gradients may be con
ture differences between the two legs of the saturated
trolled with a much lower fluid velocity than that
cells.4 Moreover, the baths had to exhibit a flexibility
required in air. Thus, larger oil baths may be built
nf design such that they could accept any type of
without requiring fluid velocities that might adnormally used cell rack.
*An bbr.vin.d vwiim of NBS T*chnic>l Not. 141.
Reprinted from ISA Journal, December, 1162
114-1
Limiting bath temperature variations to less than
0.002C per day, a modified on-off control system
helps achieve the inherent accuracies attributable to
standard cells.

Figure I. Standard cell oil baths at NBS, Boulder, Colo.

Figure 2. Top view of 28 C oil bath.

Figure 3. View of bath with cover removed.

115-2

General Description
The oil baths discussed here were designed as free
standing self-contained units. At present two baths
are in use (Figure 1). The rectangular bath is used
for measurements of 35 C; and the D-shaped bath,
containing the reference group of standard cells, is
used for measurements at 28C. Each bath consists
of three tanks: a main tank, in which the cells are
contained during measurement (and usually for 3 to
5 weeks before measurement), a drain tank, and a
pre-heater tank. The pre-heater is used to bring a
group of cells and its container, basket, etc., to the
approximate temperature of the oil in the main tank.
This is done to prevent temperature gradients in the
oil of the main tank, which otherwise would be caused
by the introduction of new groups of cells, and
would degrade the accuracy of measurements for
periods of one to two hours. The space between
the main tank and the outer wall of the bath is
filled with crushed cork. The bath time constant is
long enough to eliminate nearly all effects of changes
in room temperature over an eight-hour period. Ap
proximately four hours after a room temperature
change of 2C, the oil temperature will have changed
only by about 0.001 C. Each bath is covered with a
compound lid (Figure 2), designed so that only a
small section need be opened for measurement of the
cells. When cells are transferred from pre-heater to
main tank, or from main tank to drain tank, one
whole side is opened. The entire cover may be quickly
removed if the need arises. One-inch thick, and fabri
cated of aluminum channel faced with sheet alumi
num, the cover is filled with rigid polystyrene foam.
Figure 3 shows the D-shaped bath with the cover re
moved. Toward the rear of the main tank may be
seen part of the temperature-regulating apparatus. In
the left rear corner of the basket rack is the guard
regulator which serves to prevent a malfunction in
the main circuit from damaging the cells. The small
glass tube at the right rear corner of the basket rack
is the upper end of the main thermoregulator. The
temperature of the oil is measured with a platinum
resistance thermometer, the terminals for which can
be seen at the rear of the bath.
In general, the cell groups are contained in
baskets hung from the framework in the main tank.
As a protection against tipping, the baskets are de
signed to have the center of gravity of the cell group
below the point of suspension. The baskets are pro
vided with handles to facilitate their transfer to and
from the oil.
Temperature Control
The control of temperature within very close toler
ances requires careful coordination of both sensing
and heating elements. To allow for overshoot, the
sensing element must detect a somewhat smaller
change in temperature than the maximum permissible
temperature variation, and the heater must be de
signed to minimize overshoot. The sensing element is
a Pyrex thermoregulator (Figure 4). It consists of a
large sensing tube filled with toluene, to which is
attached a second mercury-filled tube in the form of
a "J". In operation, the toluene (which has a tem
perature coefficient of expansion approximately six
times that of mercury) expands with increasing tem
perature and forces the mercury upward in the
capillary section, causing it to make electrical contact
between the sealed platinum electrode and a piece of
platinum wire inserted in the capillary tube. The

contact closes a relay, which causes the heater

voltage to be reduced.
Though simple in principle, the regulator poses
many problems in practice, most of which can be
overcome by the methods described. One inherent
problem, of course, is its fragility; the J-tube is
more than 13 inches long and only Vt inch in diameter.
For this reason extensive experimental work was
done with thermoregulators of stainless steel (Figure
5), but these proved to be inconsistent in control. The
cause of the inconsistency was not determined with
certainty, but is believed to be the dimensional in
stability of the steel. Stainless steel regulators are
in use in less critical applications and appear to be
satisfactory for these purposes.
The first problem encountered after the regulator
has been fabricated is that of filling it. Since the
presence of air in the regulator would cause erratic
operation and temperature fluctuation with changes
in atmospheric pressure, it is necessary that the
toluene be boiled to liberate all dissolved air before
it is introduced into the sensing tube. (Toluene should
be treated with care as it is highly flammable, and
the fumes are toxic.) After the toluene has been
boiled, it is carefully poured into a flask connected to
the regulator through a three-way valve. The third
port of the valve is connected to a vacuum pump.
After the pump has evacuated the regulator, the
valve is turned, allowing the toluene to flow into the
regulator without contaminating the pump. Even
though all connections are tight and scaled with
vacuum grease, an air bubble probably will remain
after the regulator has been filled. The bubble can be
removed in the following manner. The regulator is
cooled in an ice bath and additional toluene is allowed
to flow into it. It is then carefully inverted to position
the bubble at the junction of the sensing tube and the
J-tube, and the sensing tube is gently warmed by
holding it with bare hands until the expanding
toluene has forced the bubble out. The easiest way
to introduce mercury is to heat the regulator to about
one degree higher than the highest temperature at
which it will ever be used, and then pour as much
mercury as possible into the J-tube. The regulator
is then allowed to cool to room temperature; mean
while mercury is added as needed.
In order to achieve a cyclic control of 0.001C,

use was made of a principle discussed by T. Deighton.r' The circuit using this principle is called an an
ticipator since it causes the regulator to act as
though it anticipated the increase in temperature that
would result in temperature overshoot, and to reduce
the heater current before this effect can occur. Such
anticipating action is controlled by a heater wire
wound directly on the sensing tube of the regulator.
This heater, being of much lower wattage than the
main heater, introduces very little heat into the oil.
It is supplied with current during the heating cycle,
but is completely disconnected during the cooling
cycle. The main heater is supplied continuously with
current, adjusted to produce on the heating cycle
slightly more heat than would be required to main
tain temperature, and on the cooling cycle slightly
less than would be required. Thus, while every effort
is made to minimize the temperature change in the
main heater (and still retain control), the tempera
ture of the anticipator heater is allowed to change
significantly during the cycle. As a result, the regu
lator senses a larger, faster temperature change than
actually exists, and overshoot is minimized.
The regulator operates a sensitive relay which in
turn controls two power relays. One of these changes
the voltage applied to the main heater; the other op
erates the anticipator. The relay contacts are pro
tected by zener diode spark suppressors. In order
to eliminate relay chattering, an R-C delay (about a
half second) circuit was added between the regulator
and the sensitive relay.
The heater is of an open-grid design to minimize
its heat capacity and obtain a short reaction time
in order to prevent overshoot. The heater, wound on
linen-base phenolic tubes reinforced with stainless
steel, can be seen below the cylindrical baffle in
Figure 5.
Control of Gradients
Originally considered was the possibility of using
a high-speed centrifugal-type impeller to circulate
the oil. This method, however, has two disadvantages:
it tends to promote the existence of gradients in a
large tank, and the impeller introduces considerable
heat into the oil because of friction. Since gradients
could not be tolerated, and since it was desired to
operate one of the baths only 5C above ambient
(without cooling coils if possible), the design was

Figure 4. Oil bath temperature sensing element is a

Pyrex thermoregulator.
Figure 5. Internal view of 35C oil bath.
116-3

changed to incorporate a large, low-speed, propellertype stirrer. The present circulation system is shown
in Figure 5. The four-bladed stirrer may be seen in
the center of the figure.
In operation, the rotation of the stirrer draws the
oil down through the center area of the tank. The oil
then moves outward through the heater grid and up
outside the cylindrical baffle. The horizontal ring,
at the level of the blades, was added to prevent the
oil from re-circulating without passing through the
heater. Another problem was a volume of still oil at
the center of the tank below the stirrer that caused
some uncontrolled recirculation. This was solved by
the addition of vertical blades (not visible in photo)
welded to the underside of two horizontal blades to
thrust the oil outward after it had been drawn
through the horizontal blades. The stirrer is rotated
at 50 rpm by a shaft inserted through the bottom of
the tank. Leakage through the shaft hole is prevented
by a mercury seal.
The temperature gradients were determined by
establishing the minimum distance between two
points in the oil, which, when probed with a platinum
resistance thermometer, exhibited a temperature dif
ference of 0.001 C. This distance was found to exceed
10 inches in all horizontal directions throughout the
area in which the cells are contained.
EMF Connections
Electrical connection to the cells presented another
problem: the system had to be flexible enough to
accommodate any possible terminal configuration. To
accomplish this, the various connectors are equipped
with leads terminated in polarized copper plugs
which are inserted in copper sockets provided at
strategic points within the bath. The most common
terminal with which cell racks are fitted is a %-inch
diameter copper post to which each platinum lead is
soldered. Since these posts occasionally carry beads
of solder, and since the diameter of the posts as
well as their spacing may differ from rack to rack, a
unique connector had to be designed. This contact
clamp (Figure 6) accommodates any copper post
configuration, clean or solder-covered, that has been
encountered to date.

Figure 6. Specially designed contact clamp for making

electrical connection to cells.
All leads that come in contact with the oil are
insulated with Teflon, since cloth insulation acts as
a wick, and vinyl becomes stiff after prolonged
contact with the oil.
1

Protective Devices
If a standard cell of the type commonly in use today
is allowed to reach a temperature of about 43.6C, it
may become unstable because of a transition of the
cadmium sulfate crystals at this temperature.6 There
fore a protective circuit must be included in the bath
to prevent possible circuit malfunctions from causing
the temperature to increase above, say, 40C. Also, it
was felt that the cells should be protected against an
excessive decrease in temperature, since a tempera
ture shock in either direction can cause the cells to
become unstable for an extended period. The guard
circuit in each bath consists of a bimetal regulator
that is set to take control if the temperature of the oil
either increases or decreases by about ^4C. In the
event of a line power failure, the power to the baths is
supplied through a transfer panel that automatically
connects the baths to a small engine-generator fueled
by natural gas.
Conclusion
The baths have been in use for more than two
years. The temperature at the peak of the heating cycle
does not exceed 0.001 C above the mean temperature,
and that at the lowest point in the cooling cycle is
less than 0.001 C below the mean. While the tem
perature of the room in which the baths are located
may vary by as much as 2C during the day, the
mean temperature of the oil will not change more
than 0.002C per day. The mean temperature in
creases somewhat from month to month due to the
contamination of the platinum wire electrode by the
oxidized mercury. When the temperature has in
creased by 0.01 C, it is readjusted to within about
0.005C of nominal. Such readjustment is required
three or four times a year, during which times the
platinum wire and capillary tube are cleaned.
Because of recent advances in solid-state and mag
netic devices, equally close and reliable control may
be achieved by completely electronic means. These
new devices should be explored with the goal of re
placing the fragile mercury-toluene regulatora de
vice that requires a steady hand to prevent breakage.
References
1. "The Construction and Characteristics of Standard
Cells," by G. D. Vincent, IRE Transactions on Instru
mentation, Vol. 1-7, Nos. 3 and 4, Dec. 1958, pp. 221234
2. "Clark and Weston Standard Cells," by D. A. Wolff
and C. E. Waters, Bulletin oj the Bureau of Standards,
Vol. 4, No. 1, 1907-08, pp. 33-39.
3. "The Construction of Standard Cells and a Constant
Temperature Bath," by G. A. Hulett, The Physical
Review, Vol. 32, No. 3, Mar. 1911, p. 257.
4. "Primary Batteries," by G. W. Vinal, John Wiley and
Sons, Inc., New York, 1950, p. 193.
5. "A Precision Thermostat for the Temperature Regula
tion of a Room," by T. Deighton, Journal of Scientific
Instruments, Vol. 13, Sept. 1936, pp. 298-300.
6. "Metastability of Cadmium Sulfate and Its Effects on
Electromotive Force of Saturated Standard Cells,"
by G. W. Vinal and L. H. Brickwedde, Journal of
Research, NBS, Vol. 26, May 1941, pp. 455-465.
Acknowledgments
The author is indebted to Mr. Frank D. Weaver for
his inspiration and guidance throughout the period of
design and experimentation with the baths, and also to
Dr. Walter J. Hamer and Miss Catherine Law of the
Electrochemistry Section at NBS, Washington, for assist
ance and suggestions that contributed materially to the
success of the project. The preheater resulted from a
suggestion by Mr. David Ramaley and Mr. John F.
Shafer of NBS. Boulder.
17-4

The Operating

Characteristics

Reference

Their

of Zener

Diodes and

Measurements*

WOODWARD G. EICKE.JR.
Electrochemistry Section
National Bureau ofStandards
Washington, D. C.

^ Described is a measurement technique with an accuracy in terms of the saturated

standard cell of 20 ppm and capable of 2 to 4 ppm accuracy for making stability studies.
Typical stability data are presented on several diodes. Results of a study of the effect
of current and temperature on the output voltage are given, including an equation
relating the three variables. The effect of AC components on the DC output is dis
cussed and a cause for this effect is suggested. Also discussed is the electrical noise
generated by the diode.

INTRODUCTION
RECENT TECHNOLOGICAL ADVANCES in the field of
semiconductors have resulted in significant im
provements in Zener reference diodes. As a result
they are finding wider application as a precise standard
of electromotive force. It has been reported that tem
perature-compensated reference diodes having a longterm stability of 20 ppm are now commercially avail
able. (1) At this accuracy level the operating char
acteristics of the diodes become increasingly important
since they can have a significant effect on the observed
stability. Although there are data on some characteristics
for single-junction diodes,12 J) very little has been pub
lished on the temperature-compensated ones. In the
subsequent sections the electrical characteristics of
temperature-compensated diodes will be discussed in
terms of their use as a standard of electromotive force.
Methods for measuring operating parameters, typical
data, and equations relating the variables will be given.
Because a detailed discussion would make this paper
extremely lengthy, only the major aspects of each area
Presented at the 18th Annual ISA Conference & Exhibit,
Chicago, Illinois, September 9-12, 1963.

covered will be presented, and the reader should consult

the appropriate references cited for more specific
information.

STABILITY AND STABILITY MEASUREMENT

The stability of a Zener diode may be defined as the
constancy of its output voltage with time when all
other operating parameters are constant. Unfortunately,
there has been a tendency to confuse stability with
regulation. The latter refers to the ability of a diode
or complete reference source to maintain a constant
output voltage under varying operating conditions.
The stability is the most important factor in the use of
Zener diodes as a voltage reference source, and at
present it must be determined experimentally for each
diode. This determination involves not only the diode
but the measuring system and current source as well.
Therefore, as stability requirements become more
stringent, greater attention must also be given to the
measurement system and power supply in order to
minimize the errors which they contribute to the final
results. There are a number of methods that can be
ISA Transactions April 1964

1 18-93

employed, the basic principles of which have been pre

viously described by the author. ,5) However, the one
to be described in some detail is to be recommended
since it combines both simplicity and potentially high
accuracy.
Diode Measurement
The technique employed to measure Zener diodes is
based on the opposition principle and has been used by
other investigators'2'6-7' in a variety of ways. The
method consists of opposing the unknown voltage to
be determined with a known voltage of approximately
the same magnitude and measuring the resulting small
difference.
In Figure 1 is shown a complete circuit for investigat
ing Zener diodes. This circuit is designed to measure,
one at a time, five or less diodes connected in series. A
detailed description and analysis of the circuit will be
given elsewhere.18' The circuit consists of three major
sections, the diode, the power or current source, and
the measuring circuit.
The diodes under test are mounted in a stirred oil
bath. During the voltage measurements, the oil tempera
ture is controlled to within 0.01 C. Potential leads
are soldered to each device in order to minimize the
errors arising from connecting the diode to the measur
ing circuit.

315-411 O - 68 - 9

119-94

The power supply is an 80-V lead-acid battery

isolated from ground. During use the battery current
is kept constant to improve the regulation of the supply.
The current through the diodes is adjusted by varying
Ri and it can be set to within +0.3 /xa of the desired
value. The current is determined by measuring the
voltage drop across a calibrated 10-il NBS type resis
tor.
The measuring circuit consists of a bank of unsatura
ted standard cells, mounted in a stirred oil bath main
tained at 28.00 0.01 C, and a potentiometer. The
cells are compared to a group of saturated standard
cells just prior to and immediately after measuring the
Zener voltage. The cells are measured singly, and the
sum of the individual emf's is taken as the value of
The standard deviation of a single determination of E,
is approximately 0.4 ppm. The potentiometer is a
0.01% instrument and can be used to determine the
voltage of a Zener diode with an accuracy ranging from
6 to 20 ppm, depending on the voltage being measured,
whether or not the instrument has been calibrated, and
other factors. Independent tests indicate that stability
measurements can be made to about 3 to 4 ppm.,8)
The actual measurement of the Zener voltage is
quite simple. The current is first set to the desired value
by means of Ri and the magnitude determined by con
necting the potentiometer to the standard resistor as

shown by the dashed lines in Figure 1. The voltage of

the diode is then measured by connecting E,, the
potentiometer, and the diode as shown by the dotted
lines. Finally the current is measured again after the
voltage measurement. The mean of the two current
readings is taken as the current at the time of the voltage
measurement. The voltage is given simply by the equa
tion,
Vt = E, + AE

(1)

where Ais the voltage measured by the potentiometer.

Diode Stability
The stability of a number of different diodes was
studied using the measuring technique just described
and an auxiliary power supply to maintain a current
through the diodes at all times. This stand-by supply
current (not shown in Figure 1) was adjusted to keep
the current within 3% of the nominal diode current.
The regulation was approximately 1.5%. Except for
the period of the temperature-current-voltage study,
the diode temperature was 25.2 0.2 C. Diode voltages
were measured at intervals and the data corrected to
give the diode voltage at nominal operating current
and 25.2 C. The stability curves were prepared from
these data (expressed in ppm) by plotting the reading
at a particular time minus the initial reading, as a func
tion of time.
Typical stability curves are shown in Figure 2 for

100 200 300 400 500 600 700 800

Time ( doys)
Figure 2. Stability curves for three typical temperaturecompensated Zener reference diodes.

various diodes. None of these diodes had been pre

conditioned or aged, as is now frequently done with
diodes for reference use. There are large differences
between the types of curves obtained. These curves have
two sections, the first in which the slope (rate of change
of voltage with time) is varying, and the second in which
the slope is approximately constant. This first part will
be referred to as the stabilization time, the second part
the useful period. The first part of the curve represents
the time required for the diode to stabilize or come to a
steady-state condition. Diode 1 stabilized in about 1
week and then began to show a nearly linear drift with
time for approximately 400 days. From 400 to 600 days
it remained quite steady. The rate of drift was some
what less than 75 ppm year. Diode 2 stabilized in about
5 days and showed no tendency to drift with time. It
did, however, show a sensitivity to changes in operating
condition. Diode 3 required about 100 days to stabilize
after which it drifted linearly with time. For the diodes
studied, stabilization periods, ranged from a few days
to as long as 12 months, with the mean of 25 diodes
being 3+ months. The stability of the diodes for the
balance of the time (periods ranging from 6 to 20
months) averaged 40 ppm. For simplicity, the stability
is simply taken as the maximum voltage less the mini
mum voltage during the period in question. Their
distribution was approximately normal and the standard
deviation was approximately 20 ppm. It should be
emphasized that these data are based on the maximum
changes in voltage during the period in question and
therefore include (1) random normal variations, (2)
drifts with time, (3) the effect of temperature cycling
(shaded area of Figure 2), and (4) the effect of not being
under electrical power for a short period (two points
connected by the line). Note that if the last two effects
were eliminated there would be a substantial improve
ment in the stability. These effects have been included
in the overall picture because a reference unit could be
subjected to such conditions in normal operation.
Diode 2 in Figure 2 shows a remarkable stability except
for the effects of temperature and power. When main
tained at nearly a constant operating condition this
diode showed a total variation in voltage of 7 ppm.
These data indicate that it should be possible to obtain
a usable yield of diodes having stability of better than
10 ppm.
There is no ready explanation for the effect of
temperature cycling or the effect of removing the power
from the diode for a substantial time interval. Analysis
of the data indicated that some diodes were affected by
only one condition or the other, some by both, and
others by neither condition. It was noted, however,
that all diodes affected by temperature cycling showed
a drop in voltage, and that when the power was removed
some time later, most diodes showed an increase in
voltage when put back on power. These data are still
being analyzed in an effort to ascertain the causes for
the observed phenomena, if possible. It is also worthy
of note that those diodes showing the poorest stability
and requiring the longest stabilization times were
usually the most adversely affected by the changes in

120-95

operating conditions. These effects of temperature and

power change, may well arise from causes suggested
by Banga.<9> He proposed that drifts and other phenom
ena associated with stability might be due to surface
effects on the silicon die and to AI2O3 that might be
present if aluminum is used as a doping agent. In light
of these data and recently published information*1' it
seems reasonable to anticipate highly stable diodes
that may be more or less immune to operational changes.

VOLTAGE-TEMPERATURE-CURRENT
RELATIONSHIP
In order to make a meaningful statement about the
voltage of a Zener diode, it is necessary to specify the
current and temperature at which the measurement is
made. Therefore, it is necessary to know the manner
in which the voltage is affected by these parameters.
Generally, only the approximate dynamic resistance at
the nominal bias current and the maximum tempera
ture coefficient are given for a diode. However, for
precise applications these data must be known to higher
accuracy than given; hence it is frequently necessary to
measure these parameters for each diode. Temperaturecoefficient measurements are usually made by measur
ing the voltage at a particular current (usually the
nominal bias current) at several temperatures. The
temperature coefficient is assumed to be linear between
the intervals used.*10' Two methods may be employed
to measure the voltage-current relationship. In the first
and most widely used method the dynamic impedance
is actually determined. A small alternating current is
superimposed on the direct-current bias and the alter
nating voltage drop measured; the resistance is calcu
lated from Ohm's law. The other method involves the
measurement of the voltage at various currents. This
method is a direct one and can be combined with
measurements of the effect of temperature variations
on diodes to give a complete picture of diode char
acteristics. The resistance can then be calculated from
the voltage-current data. Where precise information is
desired, the latter method should be employed. How
ever, the first or AC method is extremely useful when
only an estimate of Rz is required. There is no assurance,
however, that this method will give the same value for
the resistance as would be obtained by the DC tech
nique. Of a small group of diodes investigated by both
methods at their normal bias current, it was found that
the average ratio of RacIRdc was 0.88 where Rac
and Rdc are the dynamic resistances measured by the
AC and DC methods.
To make a detailed study of the effect of current and
temperature on diodes, a group of diodes was studied
over a temperature range of 10 to 60 C and at bias
currents of 30 to +50% of the nominal. Typical sets
of curves obtained in this investigation are shown in
Figures 3 and 4. These curves are of two general types:
one in which the voltage varies linearly with tempera
ture over the entire temperature range, the other in
which the data are not linear. The nonlinear case can

30

40
50
Tamperaturt("C)
Figure 3. Variation of voltage with temperature of a typical
Zener reference diode (linear case).
be represented approximately by two straight lines. It
was found that for all diodes showing this nonlinear
behavior, the point of intersection of the two lines was
approximately 30 C. The data of Figure 4 were further
analyzed by making a least-squares fit to the equation
V = V2b + a(r - 25) + P(T - 25)2

(2)

where V is the voltage at a current /, and T is the tem

perature. It was found that, although K25 and a were
functions of the current as expected; ,8 was a constant
over the current range studied. For the linear case the
simple relationship V =
+ a(T 25) can be used.
When a is plotted as a function of the log of the current,

SO
40
SO
Temper olure PC)
Figure 4. Variation of voltage with temperature of a typical
Zener reference diode (nonlinear case).

121-96

a linear relationship results and may be expressed as

a = log (///,,)
(3)

rm

rrr

where A> is the current required to make a = 0. This is

true whether a is found from either the linear or the
quadratic case. The variation of voltage with current
was also studied and the relationship
V = A + B log / + CI
(4)
was found to be satisfactory when T was a constant.
Equations (2), (3), and (4) can be combined to give a
single equation relating voltage, temperature, and
current as follows:
V = y0 + [B + k(T- 25)] log (///0)
+ C(/ - A) + B(T - 25)*
(5)
where K0 is now the voltage at /o. For the case where
the voltages of the diodes are linear over the entire
temperature range, 8 = 0. The quality of the fit of the
above equations depends, as expected, on the span of
the data being fitted. Equation (2) gave a good fit overthe temperature intervals studied; equation (3) shows
evidence that a second-order term is necessary. Equa
tion (4) was not fitted by itself; however, equation (5)
was; a study of the deviation showed that other terms
would be required to improve the fit. The standard
deviation of the fit of equation (5) was approximately
100 ppm over the whole range covered. There is evi
dence that k, B, B, and C may be functions of the
diode voltage and the diode construction.

ALTERNATING-CURRENT EFFECTS
In most applications of Zener reference diodes an
AC component will exist through the diode in addition
to the normal DC bias, unless special precautions are
taken. Because of the relative complexity of the device,
it is reasonable to expect that the alternating current
will have some effect on the measured output voltage
of the diode. To ascertain the magnitude of rectifica
tion and other effects, if any, an alternating current

Figure 6. The DC voltage shift due to an AC component

superimposed on the DC bias.
was introduced into the current circuit of the diode.
The AC component was determined by measuring the
voltage drop across a known resistance, and the shift in
DC was determined using the measuring circuit of Fig
ure I and a permanent-magnet-moving-coil galvanom
eter because it is an excellent filter. Tests were con
ducted in the current range of 100 ^a to 3 ma AC and
in the frequency range of 20 cps to 100 Kcps. The effects
of varying these parameters are shown in Figures 5 and
6.
The effect of frequency was determined at an alter
nating current of 2 ma. Below 1 Kcps varying the fre
quency had virtually no effect on the voltage shift,
while at the higher frequencies there was a marked con
tinual decrease in the effect of alternating current.
When the change in DC is plotted as a function of
current as shown in Figure 6, a straight line having a
slope of very nearly 2 results. The center curve is made
up from data obtained at various frequencies, while the
other two curves were obtained from data obtained at
1 Kcps. When the abscissa is converted from /ac to
/acVdc, all three curves coincide. Hence, the shift in
DC AK, owing to the AC component, can be repre
sented by the equation
- AP= AVac/dc)2

(6)

where A' is a constant characteristic of the diode.

F(*quocy ( cps )
Figure S. Variation of the DC shift in V2 due to AC as a func
tion of frequency.

NOISE
Baker and Nagy,6> reported a relationship between
the measured noise and the stability of single-junction
alloy-type Zener diodes. Their findings indicated that
diodes having a noise voltage, as measured by the pro
cedure they described, of more than 5 ixV would not be
stable and that diodes with a noise voltage of less than

122-97

120
i

iooHType 0
80
60o 40

J
L
_1_
6
8
10
11
12
Bias 'Current (ma)
Figure 7. Noise voltage of two alloy-type Zener reference
0

5
may be stable. Measurements by the author
qualitatively support their findings for alloy-type
temperature-compensated diodes. However, no such
correlation was found for diffuscd-junction diodes.
In Figure 7 are shown typical plots of noise voltage
as a function of bias current for two 1 N430 reference
diodes. These measurements were made using the same
circuit as used by Baker and Nagy except the amplifier
pass band was 80 cps to 40 Kcps instead of 80 cps to
10 Kcps. The upper curve is for a diode that was not
stable; the lower curve for a stable diode. These curves
are typical of those of this type measured. It was found
that the noise is accentuated at the lower currents and
that at currents greater than 5 ma there was little change
in noise voltage as the current was increased. This
suggests that lower currents be used if noise measure
ments are to be used to select diodes. In Figure 8 are

shown typical curves obtained for diffused-junction

temperature-compensated reference diodes. The sepa
rate curves are for diodes of three different manu
facturers. The three different diodes were relatively
stable; however, their noise-current curves were quite
different. Diodes A and B showed a certain amount of
fluctuation whereas diode C did not. It is thought that
these fluctuations arise from microplasms in the diode
structure'1" which result from nonuniformity within
the diode.
Table I gives the average noise voltage for the diode
types shown in the two figures. These values are for
the normal bias current recommended by the manu
facturer. There is wide variation in noise voltage be
tween the different types of diffused-junction diodes as
to both magnitude and standard deviation. No analysis
of the noise spectrum was made; however, such an
analysis would be of interest as it might give a clue as
to the origin of the noise, which may reside in bulk or
surface effects.
In addition to the noise just discussed, noise in the
very low-frequency region is also of considerable
interest since it affects the DC voltage measurement. If
this type of noise were present, it may make the voltage
measurement quite difficult. Unlike the higher-fre
quency noise which arises primarily from the electrical
properties of the diode, this low pass-band noise arises
from both electrical and thermal effects. The latter come
about because of the variation in velocity of the oil
over the surface of the diode package. As a result there
is a change in the temperature of the junctions in the
package. If the temperature changes are not the same
for all junctions, a relatively large variation in output
voltage will result. This low pass-band component was
estimated by observing the variation in voltage on an
oscilloscope. The oscilloscope was driven by the output
of the electronic galvanometer (see Figure 1). In Figure
9 are shown photographs of this type of noise. The
total time for each trace is 50 sec; the voltage scales are
about 10 u.V cm for all but the photograph in the upper
left, which is about 50 ^V/cm. Diode A shows a very
large sensitivity to stirring while B does not. Note,
however, that A shows less noise than B when un
stirred. Although these observations indicate that
thermal effects can be minimized by operating the

TABLE I
Average Noise Voltage of Various
Types of Diodes
Type Av. noise jiV*

6
8
Bias Current (mo)
Figure 8. Noise voltage of representative diffused-junction
Zener reference diodes.
123-98

A
B
C
D

33
H.
4
5.5

Standard
deviation ^V*

Type of
junction

8.4
1 1.4
1 (appro*.)
0.7

Diffused
Diffused
Diffused
Alloyed

*As measured with an averaging type voltmeter.

NO STIRRING

STIRRING

Figure 9. Low pass-band noise of two Zener reference diodes. Time scale: I cm (large div.) = 5 sec. Voltage scale I cm approx. =
10 /iV except upper left photograph which is approximately SO pV cm.
diode in an unstirred oil bath, it is not a complete solu
tion to the problem. What is not shown in the photo
graph is the difference in voltage between the stirred
and unstirred conditions. For diode A it is about 800 /xV
and for B less than 10 /iV. Therefore, very careful con
sideration must be given each diode, if thermal effects
are to be evaluated and minimized. The low pass-band
noise can also be estimated by making successive
voltage measurements at intervals of 10 to 20 sec and
comparing the results between diodes.

5.

6.
7.
8.

REFERENCES
1. Anonymous, "Ultrastable Reference Elements," Electronic
Ind. 22, 84-88, Feb. 1963.
2. Enslein, K., "Characteristics of Silicon Junction Diodes as
Precision Reference Voltage Devices," IRE Trans, on
Instrumentation 1-6, 106-118, June 1957.
3. Chandler, J. A., "The Characteristics and Application of
Zener (Voltage Reference) Diodes," Electronic Eng. 32,
384, 78-86, Feb. 1960.

10
1 1.

124-99

Stohr, H. J., "Bemerkungen zum Stabilisierungsverhalten

von Zenerdioden," Electron. Rundschau 16, 7, 297-301,
July 1962.
Eicke, W. G., "Making Precision Measurements of Zener
Diode Voltages," Conference Paper CP63-4I6, Presented at
the Winter Meeting of the IEEE, Jan. 28.-Feb. 1, 1963,
NYC.
Baker, R. P., and J. Nagy, Jr., "An Investigation of LongTerm Stability of Zener Voltage References," IRE Trans,
on Instrumentation 1-9, 2, 226-231, Sept. 1960.
Arnett, W., "Automatic DC Data Logging System," IRE
Trans, on Instrumentation I11, 3 and 4, 148-152, Dec. 1962.
Eicke, W. G., and H. H. Ellis, "A Study of Zener Diodes
as a Standard of Electromotive Force," manuscript in
preparation.
Banga, J., "Zener Diodes and their Application in Reference
Units," Brit. Communications and Electronics 8, 10, 760-764,
Oct. 1961.
Morganslcrn, L. I.. "Temperature-Compensated Zener
Diodes," Semiconductor Products 5, 4, 25-29, April 1962.
Baldorf, R. L., A. G. Chynoweth, G. C. Dacy, and P. W.
Foy, "Uniform Silicon p-n Junctions. I. Broad Area Break
down," J. Appl. Phys. 31, 7, 1153-1 161, July 1960.

Maki ng Precision Measurements of

Zener Diode Voltages
W. G. EICKE, JR.
MEMBER IEEE
ZENER DIODES have come into wide
use in the last few years as working
standards of electromotive force (emf) for
many applications. They have been used
to replace standard cells in recording
and controlling equipment and are, in
some cases, being used as a standard at the
0.01% level. Interest is also being shown
in their use as a reference standard of emf
at the 0.001 and 0.0001% levels.
Unlike the standard cell, which has an
emf in the range of approximately 1.018
to 1.019 volts, zener diodes presently
being considered as standards have op
erating voltages that range from 5 to
12 volts. At these voltages simple and
accurate comparison with existing stand
ards is somewhat difficult because, in
general, no current should be drawn from
the diode circuit by the measurement
circuit.
In the subsequent paragraphs the re
quirements for making precise measure
ments will be considered, methods cur
rently in use will be described, and several
proposed methods will be presented. The
methods to be described, in so far as possi
ble, are based on equipment that should
be available in standards laboratories.
Since stability studies are by far the
most important aspect of any study of
zener diodes, most experimental methods
are designed primarily for this purpose.
To assess the stability of zener diodes in a
reasonable period of time the method
must be far more precise than the desired
final accuracy of the standard. Some
feel that in order to evaluate diodes to be
used as a 0.01% standard, the measure
ment process must be capable of resolving
precisely a few parts per million (ppm) if
meaningful stability data are to be ob
tained in a period of 3 to 4 months. The
method must also be stable with respect
to time.
In the discussion to follow, equations for
computing the zener voltage Vz and
Paper 63-M45, recommended by the IEEE Re
cording and Controlling Instrumentation Com
mittee and approved by the IEEE Technical Opera
tions Committee for presentation at the IEEEIllinois Institute of Technology-Northwestern
University-University of Illinois National Elec
tronics Conference. Chicago. III.. October 28-30,
1063. Manuscript submitted April 25, 1963; made
available for printing August 7, 1963.
W. G. Eickb, jB.,is with the National Bureau of
Standards, Washington, D. C.

small variations in the zener voltage hVz

of a single measurement will be given
where applicable. The variation equa
tions are useful in studying the effect of
small variations in circuit elements and
operating parameters on Vz. The equa
tions for J Vz have been derived assuming
that the variations are small; therefore
certain approximations can be made. Al
though the variation equations apply
specifically to a single measurement they
can be extended to estimating errors in the
measurement process. However, it is
necessary to know whether the variation
in question is caused by random errors,
systematicerrors.oracombination of both.
Once this determination has been made,
appropriate statistical techniques may be
applied to estimate the error of a group
of measurements. In some studies, such
as stability investigations, systematic
errors, as long as they are constant with
time, do not affect the results since only
changes in voltage are of importance. If,
on the other hand, one is calibrating zener
diodes, both random and systematic
errors, are important. It will be assumed
that in making measurements on zener
diodes the reader will apply statistical
technique where necessary and is familiar
with making precision electrical measure
ments. For further information in the
latter area it is suggested that texts on
this subject be consulted.1-'

General Considerations
Current Source
Zener diodes are passive circuit ele
ments requiring a source of operating cur
rent. Most reference diodes in use at the
present time require a current ranging
from 5 to 15 ma (milliamperes). Be
cause the current source may interact with
both the zener and the measuring circuit
it must be considered in any discussion of
the measurement process. This inter
action may take several forms: (1) the
effect of random supply fluctuations on
the zener output voltage, (2) the effect
of a-c components on the diode voltage,
and (3) the effect of leakage between the
supply and other parts of the circuit.
Current for the diodes may be supplied by
125-433

a regulated a-c operated d-c supply or by a

battery. If the former is used, it may be
either of the constant-voltage or constantcurrent type.
In all power supplied there will be some
fluctuations of output with time and this
variation must be considered as it will
affect the voltage measurements. For a
constant voltage supply having a small
variation of KY% the expected variation
of the zener voltage 6 Vz due to the supply
voltage is given by equation 1 .
KylRz
(1)
iVz100(1- W Vs)
where I is the current, Rz the dynamic
resistance of the zener, Va the supply
voltage, and Vz the zener voltage. For
a particular diode at a specified current,
IRZ will be constant, hence the error
caused by the power supply depends on
K and Vz/Vz.
For a constant-current power supply
the variation in the zener voltage as a
result of supply fluctuations of K\% is
KjIRz
(2)
100
Note that equation 2 is the limiting case
for equation 1 where (Vz/Ve)=0. In
practice Vz/V3 can usually be made small
so that the constant voltage supply acts as
though it were a constant current supply.
In addition to regulation requirements,
the power supplies must have low leakage
to ground (power lines) and low ripple (a-c
supplies). The first requirement is essen
tial in ungrounded systems or those in
which the ground is at some point other
than the power supply.
For the most accurate work a battery
supply is recommended because it can be
easily isolated from ground and affords
excellent regulation. If the battery is
partially discharged and the load kept
constant, and a correction is made for the
slight downward drift in current during
voltage measurements on the zener, the
resultant small variation in the current
will have a negligible effect on the volt
age measurement.
The effect of supply variations can be
further reduced by using a bridge circuit
in which the diode under study is one arm
of the bridge. This circuit is discussed
in detail elsewhere.4 If stable resistors
are used and care is exercised in con
structing the bridge, the effect of small
variations in the supply voltage on the
output voltage of the circuit will be almost
nonexistent.

87*- =

Current Measurements
At some time during the measurement
of a zener voltage the diode current must

be determined. Two general techniques

may be employed:
1. The use of a stable power supply and a
stable current-limiting resistor.
2. The measurement of the current at the
time of the voltage measurement by deter
mining the voltage drop across a standard
resistor.
In the first method a highly regulated
power supply is used to power a group of
diodes. The voltage and regulation
of the supply must be selected in light of
equation 1. From a knowledge of the
supply voltage, diode voltage, and opti
mum diode current, the current-limiting
resistor Rl is selected. Assuming that
the supply voltage, the zener voltage, and
the limiting resistor are accurately
known, the diode current is equal to ( Vs
Vz)/Rl- The limiting resistor is the
total resistance of the circuit, excluding
the diode, between the terminals of the
power supply. Thus not only must the
basic resistor be stable, but contact and
lead resistances must also be constant.
The variation in zener voltage caused by
variations in the measurement of the
supply voltage &Vstt and the limiting
resistor &R,, is
IV,

(3)
Vslil-Vz/Vs)

where hVSM is the variation of measured

value of Vs, and can be thought of as the
measured value minus the true value.
This variation in Vz differs from that of
the previous two equations in that it refers
to the variation associated with the meas
urement process alone and it must be
added to the former in computing the
total variation of a single measurement.
The second method requires a variable
resistance for current adjustment and a
calibrated standard resistor across which
a voltage drop V, is measured. By meas
uring Vt, knowing the resistance of the
standard resistor Rs, and applying Ohm's
Law, the current is determined. The
variation in Vz due to the variations in
V, and R3 is
Ks

(4)

Using a high-grade calibrated 4-terminal standard resistor maintained at a

constant temperature and a calibrated
potentiometer for measuring Vh the
total error as a result of the current meas
urement can be made small. If the stand
ard resistor is 10 ohms or less, precautions
must be taken to minimize thermal emf's.
The precision of this method can be fur
ther improved by measuring V, imme
diately before and after the diode voltage
measurement. The mean is assumed to

be the current at the time of the voltage

measurement. When used in conjunc
tion with the appropriate statistical tech
niques, this method gives an experimental
estimation of the error calculated by
equations 1 or 2.
Grounding
It has been found that grounding is
frequently required to minimize electro
static effects and pick-up. However, the
point at which the ground should be made
must be determined experimentally.
When grounds are used precautions must
be taken to avoid "ground loops."
The actual leakage resistance between
any part of the circuit and ground must
be determined from a careful analysts of
the whole circuit in light of the desired
precision. Where a-c operated equip
ment is to be used extreme care must be
exercised particularly when more than one
instrument is being used in the same cir
cuit. Such an analysis must take into
account the potential differences that exist
between circuit elements, and the effect
of leakage currents through or around cer
tain elements as they affect the voltage
measurements. Furthermore, the effect
of humidity on the leakage resistance must
also be ascertained.
When grounding is used leakage resist
ance of sensitive parts of the circuit to
ground must be checked. Leakage resist
ance can be measured by several methods.
One that is quite simple and effective is
based on the use of an electrostatic volt
meter. In Fig. 1 is shown the equivalent
circuit that exists between the various
parts of the circuit and ground.
To measure resistance using this ar
rangement, the total capacitance (CT)
between the circuit and ground is meas
ured. A potential Vt is then applied
between ground and the circuit and the
time (/) required for the voltage to fall
from Vi to Kj is measured.
The total leakage resistance is given
by
R = -r\MVl/V,))-<

(5)

where CT = C+ Cv.
For a good electrostatic voltmeter the
leakage resistance will be high and R can
be assumed to be the leakage resistance
RL of the circuit to ground. When low
resistances are encountered or the time
to fall from Vi to V, is too small, the over
all circuit capacitance may be increased
by placing a high-quality-capacitor in
parallel with the electrostatic voltmeter.
The initial voltage V\ should be approxi
mately 150 volts.
126-434

CIRCUIT

ts VOLTMETER

Fig. 1 . Equivalent circuit for measuring leak

age resistance to ground
This method is primarily for the current
source and diode; when applied to the
measuring circuit extreme care is neces
sary since instruments such as galvanom
eters may be damaged by the applied volt
age. For this reason it is well to consider
grounding at some point in the measuring
circuit to avoid excessive potentials
between the measuring circuit and ground .
Shielding
Very little is known about the effect of
alternating current on the d-c output
voltage of zener diodes. However, it is
reasonable to expect alternating current
to cause a shift in the d-c zener voltage
in multijunction units because of the
two diodes used in the forward direction
for temperature compensation. There
fore, pick-up by leads should be mini
mized by shielding as a precautionarymeasure. If an a-c operated power sup
ply is used checks should be made to
determine the effect of ripple on the zener
voltage.
The Zener Diode
Although the diodes used as reference
voltage sources are either temperaturecompensated or have been selected to have
a low temperature coefficient, stabiliza
tion of the junction temperature is re
quired. The degree of temperature con
trol required for a particular diode de
pends on its construction and temperature
coefficient. Both oil and air baths are
used to study individual diodes.
In a private communication K. E.
Miller of the Fluke Company told of find
ing some diodes extremely sensitive
to stirring when operated in an oil
bath. He found that the problem
could be eliminated by using an un
stirred air bath. Robinson* reported
that some diodes were position-sensitive.
He attributed this to the effect of diode
orientation on the rate of heat transfer
from the junction to the environment.
The author has also observed the sensi
tivity of certain diodes to stirring. Al
though air baths or unstirred oil baths and
careful mounting overcome these prob

lems it is difficult to be able to define pre

cisely the temperature of the diode be
cause of its own heat generation and the
heat generated by adjacent diodes. For
extremely precise work a well-stirred oil
bath is recommended and the diodes that
are sensitive to stirring should be avoided
unless they can be mounted in such a
manner as to eliminate this problem.
In addition to temperature control the
connection of the diode to the measuring
circuit is of importance. Leads used in
manufacturing diodes are usually of ironnickel or iron-nickel-cobalt alloys. These
alloys have a high resistivity compared
with copper [40X10* ohm-cm (ohmcentimeters) to 1.7X10* ohm-cm]. Fur
thermore, the lead wires are quite small;
hence there is an appreciable resistance
associated with the leads. For example, a
No. 24 Awg(American wire gage) lead wire
of one of the mentioned materials would
have a voltage gradient of about 200 jiv/
cm (microvolts/cm) at 10 ma. Even
copper leads would have a gradient of
of about 20 (iv/cm.
This means, therefore that the points
on the leads at which the potential
is measured must be permanent for very
precise studies. This can best be ac
complished by attaching potential leads
to the diode leads, thereby making it a
4-terminal device. Some zener reference
diodes such as the 1N430 do not have
leads, but terminals. Even so, it is
recommended that the device be made
into a 4-terminal one. For less accurate
work voltage measurements can be made
using a jig that has a fixed distance be
tween clips. Such a device will introduce
a lead error into the measurement of 20
to 50 uv from one time to the next (for
noncopper leads).
Measuring Systems
Numerous systems may be used to
measure the voltage of a zener diode.
They may be classified under one of the
following types of method.
1 . Opposition.
2. Direct comparison.
3. Precision voltage divider.
Although the last two are both of the
ratio type they differ sufficiently, when
applied to the problem at hand, to war
rant separate consideration.
The opposition method is probably
the most widely used and it has been pre
viously reported in the literature.*7
Essentially it consists of measuring the
small difference between the zener voltage
and a known constant voltage of approxi
mately the same magnitude. The direct
comparison method is the direct measure

ment of the zener voltage in terms of a

suitable standard, while no current is be
ing drawn by the measuring circuit.
The precision divider method consists
of placing either a fixed or adjustable
divider across the diode and measuring
an accurately known fraction of the zener
voltage. This latter method differs from
the other two because the measuring cir
cuit draws some current from the diode
circuit.
In addition to the variation in zener
voltage to be discussed in this section the
variations resulting from the current
measurement and power supply must also
be taken into account when considering
the over-all measurement.
Opposition Methods
As was stated before, this method re
quires a known voltage Es to oppose most
of the zener voltage. Since Ea represents
most of the zener voltage being measured,
it will, to a great extent, determine the
precision and/or accuracy of the over-all
measurement. There are several ways in
which s can be constructed ; however, the
one using a group of unsaturated stand
ard cells connected in series is recom
mended. If the cells are carefully
thermostated and calibrated frequently
against calibrated saturated standard
cells, it is possible to set up a 5- to 10- volt
standard with an accuracy of 1 to 2 ppm.
At the National Bureau of Standards two
such working standards have been set up
using 500-ohm unmounted unsaturated
cells. Taps are provided so that the
emf of -ndividual cells may be measured.
It has been found that by measuring the
emf's twice daily, prior to and after
measuring the zener diodes, the differ
ences between the two measurements are
less than 1 to 2 ppm. Where less ac
curacy is required, less frequent calibra
tions are needed. A 1 0-ppm standard can
readily be made using encased cells in a
lag box and calibrating them at 1- to 2week intervals. In this case it will be
necessary to make a small correction for
the cell temperature. [For small changes
in temperature, the temperature co
efficient can be taken as 5 /iv/C (degree
centigrade).] It will be assumed for
the balance of the discussion that a
group of unsaturated cells having the
appropriate accuracy is being used for
Es.
In Figs. 2, 3, and 4 are shown three
circuits making use of the opposition
principle. These vary in equipment and
serve to illustrate ways in which the
measurements can be made.
The first and most precise of the three
is shown in Fig. 2. It is essentially
127-435

POTENTIOMETER
OOP

R SUPPLY
Fig. S. Opposition method using potenti
ometer

Fig. 3. Opposition method using e differen

tial*

SCANNER

TO RECORDING
EQUIPMENT
Fig. 4. Opposition method for automatic
data logging (programming equipment
omitted)

the same as the circuits used by Enslein*

and Baker and Nagy.' Only the essen
tial parts of the circuit are shown in
the figure. The voltage of the zener is
given by
Vz = Es + &E
(6)
and the effect of variation of Ea and A
on the measured value of Vt is
iVt-Es+t(&E)
(7)
Equations 6 and 7 apply to all three
circuits that will be discussed in this
section. Equation 7 shows that there is
a simple dependence of &Vt on the
variations of Es and A. In general

<(A) will depend primarily on the

potentiometer. Since Es is picked to
be near the voltage of the zener being
measured the resulting difference will
not usually exceed 0.5 volt. Hence the
accuracy of the potentiometer need not be
as great as the accuracy requirement for
knowing Es. For example, using 0.01%
class potentiometer, measurements ac
curate to better than 10 ppm can be made.
If the potentiometer is calibrated, its
standard cell calibrated at frequent inter
vals, and the instrument is in a reasonably
constant temperature environment, then
its accuracy will be somewhat better
than 0.01%.
The preceding statements apply to
accuracy in terms of the saturated stand
ard cells used as the laboratory standard.
If only the stability of diodes or relative
measurements are to be made, the same
equipment can be used and the relative
accuracy of a group of measurements will
be greatly improved. In an experiment
conducted by the author it was found that
a 0.01% potentiometer, when used in the
circuit of Fig. 2, would repeat within 15 mv
over a 1-month span. This means that
under these conditions the relative value
of Vz can be determined to 1 to 3 ppm
depending on the nominal voltage of the
diode being studied.
The use of standard cells brings up two
problems not previously mentioned. The
first is the danger of excessive currents
affecting Es. This is solved by placing
two low-leakage diodes back to back as
shown in Fig. 2 to limit the current to
less than 10 microamperes. The use
of these diodes is also recommended in
any circuit in which it is necessary to
limit the current. They will not be shown
in subsequent figures but should be used to
protect galvanometers and standard
cells. For the final balance the diodes
must be short-circuited. The second
problem is caused by the relatively high
resistance of the standard cells. Because
the resistance of the total circuit ranges

from 2,500 to 5,000 ohms, the galvanom

eter must have sufficient sensitivity so
that the null can readily be detected.
The particular null detector that must
be used should be determined from an
analysis of the circuit. It has been found
that a galvanometer sensitivity somewhat
greater than 10"' amperes per millimeter
is required to detect 1 nv under the
conditions mentioned previously.
In Fig. 3 is shown the circuit used by
Miller for measuring the voltage of zener
diodes. This circuit is interesting because
it makes use of two electronic instru
ments: a regulated power supply (PS)
and a differential voltmeter (DIFF VM).
Extreme care must be taken to see that
there are no leakage paths between the
two instruments. One lead of the power
supply is grounded and the differential
voltmeter is isolated from ground. In
such a circuit various circuit configura
tions must be thoroughly investigated to
insure that leakage currents will have a
negligible effect on the measurement.
Miller states that by frequently checking
the differential voltmeter, a precision of
5 to 7 ppm can be attained. This circuit
has two advantages: first, it is very useful
where the voltage of a large number of
diodes must be measured by relatively
unskilled personnel ; and second, the differ
ential voltmeter can be used to drive a
recorder for making continuous records.
In Fig. 4 is shown a circuit for auto
matically recording zener voltages. This
is the circuit described by Arnett The
source of known voltage has been modi
fied to include a 0.1 -volt step in addition
to the 1.019-volt steps afforded by stand
ard cells. In this case the current is
determined by a constant voltage supply
and a known fixed resistor. With this
type of equipment the data are punched
on cards and the system is capable of
measuring 100 diodes in 6 minutes. In
addition to recording 10A the equip
ment also records other data essential to
the measurement process. As described,

Fig. 5 (left).
Direct compariing a voltage
LTAGE DIVIDER

E,
i POT
o-jk>f=5'

Fig. 6 (right).
Direct compari
son method us
ing two standard
resistors Ri and R:
128-436

the voltage of a group of 100 diodes is

automatically recorded once an hour so
that data on both short- and long-term
stability are obtained. Arnett also dis
cussed the method of calibrating the over
all systems. As a result of careful studies
he estimated the precision of a single meas
urement to be about 6 ppm. The use of
active elements in the measuring circuit is
unique. Two operational amplifiers are
used. The XI issimply an isolation ampli
fier and serves to protect the standard
cell from damage. The other is an X10
amplifier and it is used to increase the
voltage A to a level that can be readily
measured on the digital voltmeter. Fur
ther study of the use of such active ele
ments should show other areas in which
they might be useful. A discussion of
this subject is given by Hall and Fulks.'
Direct Comparison Method
To the author's knowledge there are
no direct comparison methods in use for
measuring the voltage of zener diodes
although such a technique is quite feasi
ble. Recently McKnight10 described a
direct-reading voltage divider which
could be adapted to measure zener volt
ages directly with an over-all accuracy of
100 ppm. In addition to this, two other
methods, using conventional equipment
have been devised by the author.
Neither has been tried, but both show
promise of yielding high-quality results.
The first is shown in Fig. 5. It makes
use of a precision voltage divider (volt
box) and a potentiometer. In this cir
cuit the voltage drop IRt across the
divider is adjusted to give a null on
gals-anometer G. The drop o is then
measured by means of the potentiometer.
Two methods are given: in one a po
tentiometer is used alone; in the second,
Ea is opposed by a standard cell and the
difference measured by using the po
tentiometer. The latter gives better
resolution. Defining S=(Rt/Ri), the

voltage of the zener in terms of the meas

urement variables is
Vx = 5o
for A (8)
and
V* = 5(a+A)
for 5 (9)
for the two circuit configurations shown
in Fig. 5. The variation of Vt with
respect to the variables in equation 9 is
SVz = EJS+SSE,
for A (10)
and
iKx = 5(a+A)-r(s + A)i.S for B
(11)
The ratio (Ri/Rt) is obtained by calibra
ting the voltage divider against appro
priate standards.* The only difficulty
is the problem of two balances. How
ever, this can be readily overcome by the
proper layout of the equipment.
Fig. 6 shows the same basic circuit ex
cept that it is somewhat more sophisti
cated. Instead of a volt box this circuit
uses standard resistors. They can be
calibrated more readily than a volt box
and a ratio more nearly equal to Vz/Es
can be established. Since the four
terminal values of both Ri and Ri can be
readily determined, the problem of error
arising from interconnecting the two is
eliminated by use of the arrangement as
shown. Care must be taken, however,
to be certain that the point P, where the
interconnection is made, is isolated so
that no leakage occurs between this point
and other parts of the circuit. The zener
voltage and its variation are given by
equations 9 and 11, respectively. In this
case, however, Ri and Ri are determined
separately by calibrating the individual
resistors. This circuit is, in theory, ca
pable of both high accuracy and precision.
This circuit and that of Fig. 5 can both
be of relatively low resistance as far as
the potentiometer measurement is con
cerned.
A method using a precision variable
voltage divider instead of a fixed ratio
divider is shown in Fig. 7. In this tech
nique two balances are made, one with
the switch in the 5-position, the other with
it in the Jf-position. Although desirable,
the power supply for the divider need not
be adjustable, but it must have excellent
short-term stability. Referring to Fig.
7, the voltage of the zener is given by the
following :

V*-^ Bs

(12)

where Sx and 5a are the ratios (Ri/Rt)

with the switch in the X- and 5-positions
respectively. The variation of Vt is
given approximately by
"-i7[t-t-]
'>
Since the error of the adjustable dividers
is frequently a function of the setting,
tSx and 55s will not necessarily be the
same. As in the previous situation care
must be taken to maintain proper isola
tion between the various parts of the
circuit.
Precision Voltage Divider Methods
In this method the measuring circuit
draws current from the diode. This
method in its simplest form is shown in
Fig. 8. In this arrangement a stable
calibrated voltage divider has been per
manently placed across the diode and the
output is measured using a potentiometer.
The divider must be stable both with
respect to its ratio and its total resistance.
This technique is suitable only for relative
measurements such as stability studies
because of the difficulty of constructing a
divider with a precisely known ratio
(across the diode).
An improvement can be obtained by
using a precision voltage divider as shown
in Fig. 9. In this circuit the voltage
divider must be of high resistance. Fur
thermore, the total resistance (Ri+Ri),
where R\ is the total resistance of the
voltage divider and Ri is the total lead
resistance, must be known with sufficient
accuracy so that a correction for the meas
uring-circuit current can be made. The
voltage of the zener is given by the
equation
V*-SE(l+j[J

where (Ri/Ri)=S is determined by the

setting of the divider. In addition, a
correction is required for the current
because the current determined using Rs
is (Iz+ID) where Iz is the current through
the zener and ID the current through the
divider. If the effects of lead resistance
Rt are small compared to /?, then SVZ
is given approximately by
tVz = EsSS+StEs+SEst(^j

* Voltage dividers, including adjustable ones, are

calibrated by measuring* the ratio Ri/Rt and not A'i
and Ri separately. To take into account the possi
bility of leakage between portions of the dividers
when in use, voltages approximating those found
under operating conditions are usually employed
during calibration (see reference 3, pp. 211-20).

(14)

(IS)

Thus, if appropriate corrections are made

for the current lead, resistance, and
(Rt/Ri), Vt will depend only on the qual
ity of the divider and Es. The lead resist
ance Rt can be estimated by determining
129-437

Fig. 7. Direct comparison using a precision

divider

Fig. 8. Precifion voltage divider method I

Fig. 9. Precision voltage divider method II

the total resistance between points a and

b excluding the divider. To minimize
the effects of R> and of loading of the
diode circuit, Ri should be as large as
possible. However, there is a limit to this
because of difficulties in making, calibra
ting, and using high-resistance dividers.
At present the highest resistance unit
normally has an over-all resistance of
100,000 ohms.
An alternate method is shown in Fig.
10. In this arrangement the voltage
divider measures the complete drop from
a to b, thereby eliminating the error re
sulting from the current drawn by the
divider. This technique, however, intro
duces a new correction, the correction for
Ra which is the resistance of the circuit
from a to i excluding the diode. The
zener voltage for this circuit is
Vz-EzS^+^-IzR*

(16)

The three circuits discussed are more

useful in making stability studies because
of the difficulty of assessing sources of
error. Nevertheless, great care must be
exercised to see that critical resistances
remain constant with time.

b
Fig. 10. Precision voltage divider method III
As before, the variation in V, is given by
S Vt = EatS+SlEs +ESSS

R*tI* ~
/*!<* (17)

References
1. Basic Electrical Mbasubbmsnts (book), M.
B. Stout. Prentice-Hall, Englewood Cliffs, N. J ,
I860.
2. Electrical Measueskents (book), F. K.
Harris. John Wiley & Sons, New York. N. Y ,
1852.
3. Pebcision Measurement and Calibration,
ELECTRICITY AND ElBCTEONICS HANDBOOK 77
Vol. 1. National Bureau of Standards, Washington
D. C, 1861.
4. Silicon Zbnbe Diode Handbook. Motorola,
Inc., 18S8.

5. A Precision. Continuous Voltaob Referbnck foe Industrial Recordbbs, P. B. Robinson.

AIEE Transactions, pt. I {Communication and
Electronics), vol. 78, July 1860, pp. 187-200.
6. Characteristics of Silicon Junction Diodes
as a Precision Voltaob Reference Device, K.
BnsUin. Transactions on Instrumentation, Institute
of Rsdio Engineers, New York, N. Y., vol.
June 1857, pp. 105-18.
7. An Investigation of Lono-Teem Stability of
Zenbb Voltaob References, R. P. Baker, J.
Nagy, Jr. Ibid., vol. 1-8, no. 2, Sept. I860, pp.
226-31.
8. Automatic D-C Data Logging System, W.
Arnett. Ibid., vol. I-ll, nos. 3 and 4, Dec. 1862, pp.
148-52.
8. Tub Use of Active Devices in Precision
Bridges. H. P. Hall. R. G. Folks. Electrical Enginterint, vol. 81. May 1862, pp. 362-72.
10. A Direct Rbadino Voltaob Divxdbb with
Standard Cell Reference, R. P. McKnight.
Transactions on Instrumentation. Institute of Radio
Engineers, vol. I-ll, nos. 3 and 4. Dec. 1862. pp.
128-32

Discussion
John C. Garrigus (The Bristol Company.
Waterbury, Conn.): Mr. Eicke has de
scribed several useful methods for measur
ing the voltage and long-term stability of the
zener diode element.
It would seem that these techniques can
also be used to check the complete reference
package provided some precautions are
observed.
A complete reference source is normally
powered from the a-c power line and consists
of an isolation transformer, a demodulation
and preregulation stage which produces
relatively high output impedance (a current
source), final zener reference stage, and out
put circuitry which may be simply a voltage
divider or which may employ feedback
techniques to obtain the desired output
reference voltage.
The following points should be noted
when testing this type of reference source.

1. A specified stable load resistance should

be connected across the output terminals.
2. Care must be taken to minimize the
effects of electrical coupling or leakage which
is inherent in line-powered sources. If one
side of the output is grounded for initial
testing the common mode voltage is reduced
practically to zero; however, when this unit
is then used as a reference in some applica
tion, such as recording and controlling in
strumentation, the output normally cannot
be grounded. Then, leakage from the a-c line
through the reference and measuring cir
cuitry can cause errors in the reading.
3. Not only must the leakage from the
supply be low but the impedance between
the output terminals and ground must be
very high. As mentioned previously, some
circuit applications require the output to be
floated, as in null-balance potentiometric
recorders, and any ground loops through
the reference can cause voltage drops which
then, show up as errors in the measurement.

W. G. Eicke: The points made by Mr.

Garrigus are extremely important and must
be given careful consideration when measur
ing a complete reference package. In my
paper it should be noted that when measur
ing single diodes it is possible to select a
power supply that will be satisfactory. On
the other hand, when measuring a complete
zener reference unit the power supply and
the diode now form an integral unit, hence
the measuring procedure employed must be
compatible with the unit under study. For
a particular technique for measuring a
zener reference unit, either the opposition
or the direct comparison method may be
employed. Unless the precision divider is
built into the reference unit the voltage
divider method would not be particularly
satisfactory. If there is one generalization
that can be made it is that the voltage
as measured under test conditions must be
identical to that which the unit will be
able to deliver under normal operating
conditions.

A reprint from COMMUNICATION AND ELECTRONICS, published by

The Institute of Electrical and Electronics Engineers, Inc.
Copyright 1964, and reprinted by permission of Ihc copyright owner
The Institute assumes no responsibility for statements and opinions made by
contributors.
Printed in the United States of America
September 1964 ii

130-438

NBS Technical Note 430

October 1967

Designs for Surveillance of the Volt

Maintained by a Small Group of Saturated Standard Cells
W. G. Eicke
Electricity Division
Electrochemistry Section
and
J. M. Cameron
Applied Mathematics Division
Statistical Engineering Laboratory
This technical note describes a procedure for maintaining surveillance over a
small group of saturated standard cells. The measurement process is briefly
discussed and the principle of left-right balance as a means of eliminating certain
systematic errors is developed. Specific designs and their analysis for intercomparing 3,
5 and 6 cells in a single temperature controlled environment are
given. Procedures for setting up control charts on the appropriate parameters are
given, and a technique is described for detecting certain types of systematic errors.
Key words:

Control charts, experiment design, saturated standard cells,

standard cells calibration, statistics, voltage standard.

131-1

Resistors

Resistance

and

Apparatus

Papers
Stability of double-walled manganin resistors, J. L. Thomas
Errors in the series-parallel buildup of four-terminal resistors, C. H.
Page
Precision resistors and their measurement, J. L. Thomas
Calibration of potentiometers by resistance bridge methods, D.
Ramaley
Notes on the calibration of the.direct reading ratio set, P. P. B. Brooks. .
Direct ratio readings from a URS, D. Ramaley and J. F. Shafer
A method of controlling the effect of resistance in the link circuit of the
Thomson or Kelvin double bridge, D. Ramaley
Method for calibrating a standard volt box, B. L. Dunfee
Human engineering a console for the comparison of volt boxes, P. H.
Lowrie, Jr
A method for calibrating volt boxes, with analysis of volt-box selfheating characteristics, R. F. Dziuba and T. M. Souders

135
140
149
184
187
198
200
204
217
222

Abstracts
Methods, apparatus, and procedures for the comparison of precision
standard resistors, F. Wenner
Notes on the design of 4-terminal resistance standards for alternating
currents, F. B. Silsbee
Measurement of multimegohm resistors, A. H. Scott
Calibration procedures for direct-current resistance apparatus, P. P. B.
Brooks
Practical methods for calibration of potentiometers, D. Ramaley
Some modifications in methods of calibration of universal ratio sets,
D. Ramaley
A versatile ratio instrument for the high ratio comparison of voltage
or resistance, A. E. Hess
See also, "A system for accurate direct and alternating voltage measure
ments," by F. L. Hermach, J. E. Griffin, and E. S. Williams
133

231
231
232
232
233
233
234
290

U. S. Department of Commerce

National Bureau of Standardo

RESEARCH PAPER RP1692

Part of Journal of Research of the Rational Bureau of Standards, Volume 36,
January 1946
STABILITY OF DOUBLE-WALLED MANGANIN RESISTORS
By James L. Thomas
abstract
The international ohm is now maintained at the National Bureau of Standards
by means of double-walled manganin resistors. It is assumed that the average
resistance of a group of 10 standards, selected from a group of 24 constructed in
1933, remains constant with time. No one of the 10 standards has changed in
resistance with reference to their average by more than 1 part in a million in 12
years.
CONTENTS
I. Introduction...
II. Description of new resistors
III. Stability of resistance

Page
107
108
109

I. INTRODUCTION
Although the international ohm is denned in terms of the resistance
of a column of mercury, wire-wound standards are used at national
standardizing laboratories for maintenance of the .unit of resistance.
At the National Bureau of Standards the international ohm is main
tained by means of a group of 1-ohm resistors. From this group 10
standards have been selected, and the assumption is made that the
average of their resistances remains constant with time. If there is
a relatively large change in the resistance of one standard with refer
ence to the average, it is replaced with one that has remained constant.
In this way standards are eliminated from the reference group when
they become defective.
Values of resistance were originally assigned to standards of the
reference group in terms of mercury-ohm determinations made in
England and in Germany before 1910. Mercury-ohm determinations
were made at this Bureau in 1911 and 1912, and the results were con
sistent with the English- and German results to 15 or 20 parts in a
million.1 In 1910 the reference group consisted of 1-ohm standards
of the type developed by Rosa,2 and resistors of that group were used
for over 25 years. Starting in 1931 the standards of the Rosa type
were gradually replaced by standards of the double-walled type.'
Since 1939, maintenance of the international ohm has been entirely
by means of a new group of double-walled resistors constructed in
1933. Resistors of the later group have been far . superior to any
others made at this Bureau, and, judged on the basis of relative
stability, are about one order better than well-aged standards of the
Rosa type. It is the purpose of this paper to describe the new stand
ards and give a report of their performance to date.
> F. A. Wolff, M. P. Shoemaker, and C. A. Briggs, Construction of primary mercurial resistance stand
ards, BS Set. Pap. 13, 37S (1916) 8266.
> E. B. Rosa, A new form of standard resistance, BS 8ci. Pap. 0, 413 (1908-8) 8107.
J. L. Thomas, A new design of precision resistance standard, BS J. Research 8, 298 (1930) RP201.
315-411 O - 68 - 10
I

135- 107

II. DESCRIPTION OF NEW RESISTORS

The standards constructed in 1933 consisted of a group of 24 of the
double-walled type, and the same technic was used for annealing and
mounting as that described in Research Paper RP201. The con
tainers, however, were larger than those of the original double-walled
group, and were similar in size and general appearance to O. Wolff's
model IV.
The increase in the size of the containers made possible the use of
larger resistance wire, thus decreasing the ratio of area to volume and
thereby reducing possible changes in resistance from surface changes.
The resistance coils were made of 28 turns of No. 12 AWG manganin
wire (2.05 mm=0.081 in.) from two lots of specially selected wire,*
wound on 8-cm mandrils and spaced with copper wire. Before the
wire was annealed, it had a slightly negative temperature coefficient
of resistance at 25 C, and after being annealed, the coefficients were
very small, averaging 1 part in a million per degree centigrade for the
24 standards, the maximum coefficient being 3.05 ppm per degree
centigrade.
For annealing the resistance coils, a special furnace was constructed,
using a fuzed-quartz tube about 10 cm in diameter and 75 cm in
length. This tube was sealed at one end, and the other end was
covered with a ground-glass plate through which were sealed thermo
couple leads and an exhaust tube. Several parallel aluminum plates
were placed in front of the glass cover plate to protect it from radiant
heat from inside the furnace. The heating cou around the tube was
nowhere nearer than 6 inches to the cover plate, and a good seal was
maintained by the use of ordinary stopcock grease.
The resistance coils mounted on the steel mandrils were annealed
in the quartz furnace. An oil-sealed pump was used to evacuate the
furnace, maintaining a pressure of the order of 0.001 mm of mercury.
About an hour was required to bring the furnace to the annealing
temperature of 550 C, at which it was maintained for half an hour,
after which the power supply was disconnected and the furnace
allowed to cool fairly rapidly.
After the annealed resistance coils had been adjusted to the correct
lengths, manganin terminal plates were silver-soldered to the ends.
Manganin rather than copper plates were used in order to keep the
temperature coefficient of resistance as low as possible.
The mounted resistance coils were spaced with many turns of linen
thread wound between the turns of wire. All but two of the coils
were then impregnated with shellac varnish and baked at 120 C
before sealing them in dry air in the containers. In the 2 months
immediatelv after sealing, each resistor, except the two which were
not shellacked, decreased in resistance by from 1 to 5 parts in a
million, after which the resistance appeared to be stabilized. No
change was obtained even during the initial period for the coils that
were not shellacked, which suggests that the initial change was from
dimensional changes in the shellac.
One of the completed standards is shown in figure 1. The outside
diameter of the container is 9 cm and its length 13 cm. The series of
holes near the top are just above the double-walled part and are
intended to increase the facilities for cooling, and the containers are
1 Chemical analysis of one lot: Cu 82.2, Mn 13.2. Ni 4.2, Fe 0.3; of second lot, Co 82.7, Mn 12.1, Ni 4.0,
Fe0.5.
136-108

left open at the bottom for the same purpose. The terminal posts are
entirely of nickel-plated copper, mounted on a hard-rubber top; the
distance between the amalgamated current terminals being 16 cm.
III. STABILITY OF RESISTANCE
In the fall of 1933, after all standards had been on hand for at least
2 months, the group of 24 was intercompared, resistance measurements
being made to parts in 10 million. Three of the standards were thfn
sent to the International Bureau of Weights and Measures, and since
that time four more have been sent to other national laboratories.
The remaining 17 standards have been intercompared to parts in 10
million nearly every year since their completion. Two of the stand
ards have changed in resistance by 4 and 8 parts in a million, in the
12 years since their construction and are considered to be of inferior
quality. The remaining 15 resistors comprise the group now used
for maintenance of the international ohm. The average value of 10
standards is assumed to remain constant with time, and 5 are kept in
reserve to replace standards in the group of 10 if any of them should
become defective and change in resistance by an unusual amount.
Table 1.Stability of double-walled resistors in maintenance group
Change in resistance, In microhms, from 1933 until
No.
1934

1933
0
0
0
0
0
0
0
0
0
0

08
70
74
77
78
81
82
84
89
90

0.1
.0
-.2
-.8
.0
.0
-. 1
.5
.0
-.2

1936
0.8
-.1
-. 1
-.8
.2
-.4
-.1
.0
.5
-.3

1937

1938

1939

0.6
-.7
.0
-1.2
0. 1
-.9
.0
.3
.0
1.3

0.7
-.2
-.2
-.8
.4
-1.0
-0. 1
.0
.1
.8

0.6
-.2
-.4
-1.0
0.6
-1.2
-0.2
.8
.1
.8

1940
0.1
-.4
-.6
-.9
.9
-.9
-.1
.8
.0
.6

1941

1943

1944

0.2
-.4
-.6
-1.0
0.8
-.9
+.0
.8
.0
.7

0.0
-.6
.0
-.7
.9
-.9
-.2
.8
.1
.7

0.0
+.2
-.4
-1.1
1.0
-1.0
-0.2
.8
-.3
.6

Average of numerical values without regard to sign..

1948
0.1
.3
-.7
-1.0
1.0
-0.9
-.4
.<
+.1
.8
0. S6

Table 1 shows how exactly standards in the maintenance group of

10 have kept their relative values during the last 12 years. This
table shows the total change, in parts per million, since 1933, calcu
lated on the assumption that their average value has remained con
stant. Table 2 gives similar data for the 5 reserve standards, meas
ured in terms of the average of the group of 10. Data for the 5 are
somewhat incomplete.
Table 2.Stability of double-walled resistors in reserve group
Change in resistance, in microhms, from 1933 until
No.
1933
87
69
72
J3
83

_..

1934
0
0
0
0
0

1.1
-0.1
.2
.0
.2

1936
0.6
.1
.3
.4

1937

1938

1939

1940

1941

-0.1
.1
.0
.2

-0.2
.7

-0.4
1.0

-as
1.2

-0.6
1.8

1.9

-1.0

-1.8

-.2
137-109

1943

1944
-0.9
2.0
0.6
-.8
-1.9

1948
-1.8
1.9
0.4
-.9
-2.1

For the 10 resistors now being used in maintaining the international

ohm, the maximum net change in 12 years has been 1.0 part per
million. The average net change without regard to sign has been
0.56 part per million in 12 years, or at the average yearly rate of
about 5 parts in 100 million.
The bridge which has been used for the intercomparison of these
resistors has dials that can be set to parts in a million, the next place
being obtained from galvanometer deflections. The method of
measurement is such that the bridge is relied upon to measure differ
ences in resistance only. The facilities for cooling and the lowtemperature coefficients of resistance make possible measurements
to 1 or 2 parts in 10 million, when the standards are measured in a
well-stirred, thermostatically controlled, oil bath.
With double-walled containers having the resistance coils mounted
tightly in contact with one wall, variations in atmospheric pressure
tend to expand or compress the walls and thus apply stresses to the
coils. To test the change in resistance so produced, three of the
double-walled resistors were placed in a closed chamber and measured
at about normal and half-normal pressures. The decrease in resist
ance as the pressure changed from norma] to half-normal was found
to be from 2 to 4 parts in a million. Assuming the change to be
linear, changes in barometric pressure can therefore be expected to
produce uncertainties in the resistances of the standards of only 1 or 2
parts in 10 million, which is about the limit of measurement.
On the basis of relative values, standards of this group are about
one order better than well-aged standards of the Rosa type. It does
not necessarily follow that their average value is more stable. Un
fortunately, there has been no satisfactory method of judging the
quality of resistors used in maintaining the unit except their ability
to maintain relative values. A group that keeps the same relative
values could all be changing, but at the same rate.
Before the war, some preliminary results had been obtained from a
commutating method of measuring resistance in terms of inductance,
which in turn could be calculated from the measured dimensions of a
mutual inductor. By this method it was found possible to duplicate
measurements of a 1-ohm resistor to 1 part in a million in terms of
the calculated mutual inductance. It is expected that such measure
ment will in the future give data on the stability of groups of resistors
over long periods of time. The stability of the inductor can be
checked from time to time from its measured dimensions, and it is
believed that this can be done to 1 part in a million. Measurements
of the resistors in terms of the inductance would then disclose changes
as large as 1 or 2 parts in a million in the resistance of the group
being used to maintain the unit. It is not expected that resistance
can be measured in absolute units to 1 part in a million by this method.
It does appear, however, that systematic errors will be sufficiently
constant to permit checks of the stability of the unit of resistance
to 1 or 2 parts in a million.
Washington, September 24, 1945.

138-110

Journal of Research of the National Bureau of Standards

Research Paper 1692

Figure 1.Double-walled standard resistor.

Since the original publication of this paper

these resistors have been measured at approximately
yearly intervals. The changes in resistance between
the yearly measurements are about the same as the
changes shown in tables I and II, indicating that
there has been no deterioration in the stability of
the resistors.
Thomas E. Wells
August 17, 1967

139-111

JOURNAL OF RESEARCH of the National Bureau of StandardsC. Engineering and Instrumentation

Vol. 69C, No. 3, July-September 1965

Errors

in

the

Series-Parallel

Buildup

of

Four-Terminal

Resistors
Chester H. Page
(April 20, 1965)
The use of n equal resistors
in scries and (b) in parallel provides an ns:l ratio of
potentially high accuracy. Such devices arc important for extending the use of the national
one-ohm standard to the hundred-ohm, and thence to the It) 000-ohm, level.
Formulas are derived for the error in the ratio, expressed (a) in terms of design tolerance,
and (b) in terms of first-order residual misadjustments which combine to yield the secondorder error of the ratio.
It seems feasible to construct a 1 :10() ohm buildup device with a ratio uncertainty of
less than 1 in 10".
The use of n equal resistors (a) in series and (b) in
parallel provides an /r:l ratio of potentially high
accuracy. Such devices are important for extending
the use of the national
standard to the 100-12, and
thence to the 10 000-fi level.
The first stage of this step-up involves the use of
ten 10-ohni resistors, which must he four-terminal
resistors to avoid large errors due to the connecting
networks.
Haniou 1 has described an arrangement of fourterminal resistors permanently connected in series,
and convertible to a parallel connection by adding
jumpers. In that paper, he shows the use of com
pensating resistors in the potential leads to eliminate
errors introduced by the added connections.
Compensated lead "fans" may be used for the
potential terminals of the paralleled resistors, for the
current terminals, or for both. The aim of the
present paper is to present a complete analysis of the
general case, and formulas for t he errors introduced
by imperfectly compensated fans.
i It. V. ll:iiiioii. A lOilSf t)tlild-U[i rrsislor Tor the nilil 'ration of standard resistors,
J. Set. Instr. 31, 450-453 (Dir. 1954).

We consider four-terminal resistors connected in

series by means of "tetrahedral" junctions, having
the equivalent circuit shown in figure 1. Each junc
tion supplies current and potential leads. This array
can be converted to a parallel connection by adding
four "terminal fans," as in figure 2. For analysis,
the junction resistances can be considered as ab
sorbed in the fan-conductor resistances. The prob
lem is to make the four-terminal resistance of the
combination precisely equal to Rjn. If the resist
ances of the various arms of the current fan are
adjusted to make each main resistor
carry iden
tical current, the corresponding voltage drops will
be identical and there will be no circulating current
in the potential fans, and the potential across Vt V2
will be independent of the resistances of the potential
fans, and equal to IH/n. This requires that each
arm of the current fan on the left have the same
resistance, say r, except for the top and bottom arms,
each of which feeds only one main resistor. These
end arms must have the resistance 2r. For the
right-hand current fan, each arm must have the
same resistance, say /'. The reciprocity theorem
leads to the conclusion that if we use these com
pensated fans as potential fans, the four-terminal

Figure 1
140-181

Figure 2

resistance would be R/n, independent of the currentfan resistances. Since the potential fans can tolerate
larger resistance than can the current fans, the
compensation is usually made in the potential fans;
resistances of several tenths of an ohm can be added
to the arms, allowing adjustment to reasonable
accuracy.
Making either set of fans perfect yields zero error.
This suggests that the overall error is in the nature
of a product of fan errors, and suggests the possibility
of reducing the effect of residual potential fan errors
by making at least a rough adjustment of the current
fan. For analysis, we consider four perfectly com
pensated fans as a nominal condition, with arbitrary
maladjustments allowed in each arm. In addition
each main resistor is allowed a departure from
nominal.
The complete circuit to be analyzed is shown in
figure 3, where the labels indicate conductance rather
than resistance. Capital letters indicate "average"
values, i.e.,

Considering terminal G as the "ground" terminal,

we have a network possessing three external nodes,
and
internal nodes. A complete description
of the network requires 3 + (n+l) simultaneous
equations, and we must eliminate the last n + 1.
141

If we impress currents Ilt Iu, Im into the external

nodes, and I\t 1% . . . In+i into the internal nodes,
the voltages on these nodes are implied by the
simultaneous equations:
ii=Yllvl+Yinvu+

+ry1+

+YIn+lvn+1

In = Y111Vl+

+YUn+lVn+1

In+l YH+ijVi-\-

-\-Yn+in + iVn+i
(1)

These equations are conveniently symbolized by the

matric equation
I=YV
(2)
where J and Vare column vectors (each having n+4
components), and Y is an (n+4) X(n+4) square
matrix.
Consider Jlt In, and Im as components of an
"external" current vector, Ie, and Ix . . . In+i as
components of an "internal" current vector, If,
similarly consider "external" and "internaP'unknown
voltage vectors, V, and V(. Grouping the terms
appropriately:

I<=PVt+yV(
182

(3)

FlGDKE 3

where a is a 3X3 matrix, /Sisa3X(rc+l) matrix, 0

is its transpose, and 7 is an (w+l)X(?i + l) matrix.
In other words, the matrix is partitioned:
r G

y)V'

fl=(8-3ll) = (0, l,-l)z

(S)

(4)

Since we are interested in knowing the external

voltage with no impressed internal current, we must
eliminate Vt from
i,=av.+pv,
(5)

0=pV.+yVt

The problem is to carry out all the indicated

algebraic manipulations, with all resistors subject
to arbitrary tolerances. This will be done in an
appendix, using the following strategy.
First, we carry out the manipulations required
for (a jiy'1
assuming the nominal values of the
network elements; then, assuming that the de
partures from nominal are small, find the resulting
correction to the nominal R,

finding

(6)
This last equation can be written as

Zu

>
Zn

rir

Zll
v = Z21

222

Z23

In

^32

Z33^

as a series m
powers of the perturbations. This yields a formula
for the fractional error as a sum of terms of the
type gtgu gta(, g,bt, etc.; a(6(, c,di. The terms in
volving the gt (tolerances in the resistors of the
series-parallel set) are ordinarily negligible com
pared with the pure measurement-network errors.
These latter give (from eq (A38)):
Error=

(7)

G
nA+B VAB+AB
+2
AB

The required four-terminal resistance is the transfer

resistance (VnVIU)/Ii under the condition 7U=
-^111=0, hence
142-183

B J

G /Cj^.Cirf,,
, C, d_n\
+.n C+D \CD~rCD~r " ' - ^CDJ

,Qv
w

Since -a( = 0, etc., this error arises from accidental

correlation between errors of fan connections to
common points. The worst case is where half the
a, have maximum positive error, the other half
having maximum negative errorand where these
pair up with the same situation among the 6(.
Normal design would have DA, B=C, and
*4>>Z}. Now let the maximum fractional error
in the voltage fans be ; in the current fans,
Q
The resulting extreme error is 4 -^ e,t,. Reasonable
estimates are ^=10 fi, -^=1 mfi, t| = 10_1, =10~3,
yielding a maximum error in the series-parallel
ratio of 4X10"8. Fractional errors of 10~5 in the
main resistors make the gt contributions completely
negligible.
The above computation is for maximum error.
Random pairing of the fan errors gives a much
smaller expected error.
The next problem is to adjust the device after it
has been constructed, and to evaluate the residual
error. We note that if the current fans are perfect,
there will be no voltage across any arm of the
potential fans, hence any voltage across an arm of a
potential fan is a measure of current-fan errors.
The potential fans have relatively high resistance,
so are conveniently "trimmed" on a resistance
bridge. To adjust the current fans, we can connect
a microvoltmeter across each potential fan resistor
in turn. This essentially measures the deviation
of the potentials on the internal nodes, from their
average. The current-fan arms can be filed to
reduce these voltages to approximately zero.
In the analysis, we have treated the tetrahedral
junctions as though their centers were available
terminals. Any adjustment procedure must take
cognizance of the fact that these terminals are
fictitious; any physical connection is separated from
the tetrahedron center by a low resistance. This
could cause errors in direct measurement of currentfan resistance. The proposed scheme of measuring
potential-fan resistances and voltage drops avoids
this difficulty, since the potential fans are loaded
with additional resistance.
In appendix A, we evaluate these residual adjust
ment errors quantitatively. We find to first order
that the voltage across the arm B-\-b3 is given by

Error=nG J^(A+B)

(-^+A'3H'3+A'5ir5+ . . .

+(C+D)(XtWi+XiWi+ . . . +Xn\\\) V
(12)
For a test current of 1 A, the fan tolerances
assumed before yield voltages of the order
/A'~2X10-5V
ZH~2X10-7V
with
nGA~ 103.
Note that the larger voltage is generated by the
current-fan error (under the assumed construction
tolerances); this should make it feasible to adjust
the current fan to the next order of magnitude.
The most sensitive set of measurements for evalu
ating the residual error is probably the set of residual
adjustment errors; i.e., the residual potential-fan
resistance differences and voltage drops. The ap
propriate formula is (from eq (A34J):
Residual error=2<?{ XJJB+XJJB .
+ Xn+1bH+JB- X&/C- XtcJC-

-XncJC}
(13)

where positive Xt is associated with node i positive

with respect to the appropriate potential terminal.
Note that the measured error quantities are firstorder effects in the network tolerances; the computed
residual error is a second-order product of these
terms, so can be reasonably accurately evaluated by
this procedure.
There are, of course, additional errors not attrib
utable to the current and potential fans. The first
of these arises from the fact that the series resistance
is not n/G, but is

(10)

JX3
~n(A+B) A

and similarly the voltage across A + a3l for current /

into terminal (II) and out of terminal (III), is
;,r=
2
f>3
3 n(A+B) B
Comparison with eq (9) yields

(11)

yielding a fractional error of only 10"10 for g,/G^- 10~5.

Another source of error lies in the imperfections of
the tetrahedral junctions. Under conditions of
suitable design and proper adjustment procedure,
these junctions can be balanced to transfer resist
ances of less than 10"s 0. In a string of 10-Q
resistors, this is then a potential source of error of
the order of 10"9.
It seems feasible, therefore, to construct a 1:100 S]
buildup device with a ratio uncertainty of les> than
1 in 10".

143-184

Appendix A
The submatrices of equation (4) are
rf -hh+A+a,+
0
0

0
C+et+C+ct+

Cn4

0^
'A 0 0^
(Al)

o 5? o
0

Lo

(\

-(A+aJ .
(A2)

-(<7+c2)

-r'+c<)
0
o

-(Q+9i)
(G+gt)

2G+Ul+g2+C+c2+D+d2

-(G+gt)

-(G+gt) . . .
(A3)

7=
2G+g2+gi+A+a3+B+bi

It is convenient to decompose these matrices into

sums of nominal values and perturbations:
Let
7 = T+7o+7i+72

(A4)

(A+B
2

C+D

A+B

0
(A7)

r=

where

A+B
-A/2

-A

-A

-B/2

-B

-B

-c

-c

Yo = GS

(A5)
-1
2

0
-1

0
0

-1
(A8)

a,

a3

01= -6,

-63

Ci

Ct

. 0

(A6)

144-185

-1

-1

'a.+ft,
0

c2+d2

This yields
^2
2
+1 4-
A^G^D

12
?+D

2 "
D

0
(A9)

7: =

2+1+1
B^G^D

-+G^D
2
D

2
D

2
D

(A14)

2 2
C+Dj

1+2

-+i7i

-Ox

0J

</2

(J2

z,=i i+n

(J2 + U3 -03

(A 10)

7i=

9n

9n j

We write the admittance matrix as the sum of a

nominal value and a perturbation:

\A>

r+70/

j2
D

2^
Z?

i+-

2
D

(A15)

These matrices are readily verified by inspection of

the circuit when the elements have their nominal
values.
The matrix Z2 involves the term (r+70)_I standing
by itself; we shall eventually need this term and
shall use a series expansion for it.
The four-terminal resistance of interest is the
transfer resistance expressed as Vn Vm for 7i=l,
all other input currents being: zero:

\/3 71+72/

/1\
0
w

(All)
Let
R=[(0

(A12)

1-1) (0

0...0]F-

"0"
0

(A16)

where
Zo=(a-/3o(r+7o)-/3o)-

(A 12a)

Z^-ZACr+Yo)-'

(Al2b)

ZJ=(r+7o)-1-(r+7o)-1/S0Z1.

(Al2c)

In (A16), we have partitioned the vectors into

"external" and "internal" vectors, corresponding to
the partitioning of Y. For the nominal case we find

Xow (r+7o)-1 if not readily found, but the

combination /30(r+70)_1 is, from appendix B:
&(r+7o)-,=-(i+2<?CE+F)]-1
\ 2GEF[B
I
\c
IAE
+(#'
\ 0

0
0
<?F

B
C

fl0=(0

l-l)Z0^)j=-^-

(A17)

We are also interested in the transfer resistances that

are measured as voltage drops across the potentialfan arms:
B ...)
C..J

ylE 0 .
BE 0 .
0 CF.

(A13)

^,= [(0-1

0)(1

0 0 . . . ojjr-

where
1
A+B

1
~C+D

LloJJ
145 -186

All A',=0 for the nominal case.

The perturbation from nominal is found bv ex
panding Y~' in a series (as shown in appendix 1$ for
(r+7o)-):
A%=[(0 0-1) (0

0. . .0)}Yr-=r.-i-r0- Q AW0-

LLoJJ
.
0 l"
w
-Y=[(0-1

0)(0 0

to second order in the perturbation.

The algebra will be alleviated by noting some
special properties of the matrices ft, yu and y2:

0 . . . 0)]Y-

(A1S)

(1

0 . . 07i=(i+&i

(0

1 . . 0?i=

(1

0..072=

(0

1 . . .)72= (-ffi

(A19)

(1

O...)0i=(O

0 0)

(0

1 . .

= (0

(A20)

since Sa,=26t=2c,=0

03+63

(0

o2+(/2

Ct+d*

(t7i

-(7i+^)

fa+fe)

(ffi+fl,)

Cj7i+j7a)

..).

Note that premultiplication of either 7, or72 bv either

(1 0 1 0 . . .) or (0 1 0 1 . . .), followed by
postmultiplication by either

0)

(A21)

These simplify the multiplications required by eq

(A19):
/0 0i\/^o ZA/0

T
TP
1
0
1
0
results in a null vector.
or
0
1

0 \k /

\z.
=\z*fc

ft\

zJU */

/ 0

z0^+zl\

\z2h

za+ZhJ

z^+z2t)

a"'"

Note that
A/Z^Zo^Z.+Z.tZ^O,

Z,M=0

(A24)

so that
Thus from (Al2b) and (Al3) in conjunction with
(A20) and (A21), we have the very useful relations:

MZ2

(A25)
and finally
zmz,=o
I
ZiY2Z, =0.

(A22)
146-187

Neglecting g/O relative to 6/B and c/C, we have

_/

MZ2M

Mz2(0iZl+tz2)\

\(z2( + zMz2M

etc.

2 Xtt+xbu+xlB~ S A-2(c2(/e

(A34)

Using these expressions in (Al9) yields

for evaluation of error in terms of measured residuals.
R=-^+(0
nG

1-1)MZ2M(0
Voy

(A27)

To express R in terms of circuit element tolerances,

we need to express Ar:

with no first order correction.

The A", have first order terms:

A=-(l
= -(1

A, = -(l

0...)Z2A>/o]
w

0){Z0^1 + Z1(71+72)}(r+7o)-'

which becomes
1 /2a,

A2=-(0

0 0)M(r+7o)-'

(/,

0...)Z2M(0
Voy

2rf2

0, + fir,

2a3
A

+ g3
G '

)(r+7o)-.

(A35)

We need (r+70)_I; we have

X3=-(0

0...)Z2M 0
Voy

(A28)

etc.
These can be interpreted as the components of a
vector, A':
(A2C)

(r+7o)-1 = (r+GS,)-1=r-,-GT-1,sr-1
+G2r-lsr~lST-'Now
2E
0
0
0
0
.F 0
0
0 E

(A36)

We can now write

r-'=
R=^-(0

1-1) MX

(A30)

as an alternate expression.
From (A 12c) and (A24) we have

2/<;

Z2M = (r+7o)-'M

(A31)

so that
^=^+(0

i-i)A/(r+7o)-'M(o

and the successive terms of (A36) diminish as powers

of GE and GF, which are small compared to unity.
Hence the first term of (A36) is sufficient for our
purposes, for use in (A35):
i

(A32)
n \A

2rf2
-^Yo ^
n \ ' D

Now
(0

2a3
2f/,
a'0' A

l-l)A/=(0

l-l)Z0/S, + (0

1/1 2,1

1-1)Z,

gi+fo n Mi
G
U' D

g3+g<
G
u ..)

(A37)

and

2\
*

nG+n*V\^
/2a3

71 Or
1 /gi 26,
-n\G~ B

(h+fjz
>0
G

GJ\B

ffz+ffA /263

G)
ff2+g3\

0...)(7i+7j)

g2! 4-??,
G
(h+!h_l^, . . \

}
//2c2_1+2\ /2rf2 _L-tfif2\
"V\\C
G )\D
G )

(A33)
It7 g~7 V^d g~)+
147-188

(A38)

Appendix B
then

A power series expansion yields

C2E 0 0 . . . 0 ^1
(r+7o)-, = (r+GS)-1 = r-,(/+GSr-1)-1
= T-l(I-GSr-' + G2ST-'ST-1- . . .)
I

o^o
I 0 0 E

:
(B5)

=r-1 GT-'ST-'+^r-'iST-'Sr-1- .
(Bl)

v. 0

The series converges if the norm of Sr 1 is less than

l/G. A sufficient condition is
G<mm(^(A+B),?^y

2E_

f-AE

-AE 0

-5

-BE 0

r 2E

(132)

...

-2
0

-CF

-F

(B6)

-rF

. .

2F

-E
2.E

-E

2F

(B7)

sr-'=

Premultiplication of the series by /30 yields:

^o(r+7o)-1=0or-,-/3or-,(Sr-1)
v.
+<32/90r-1(sr-')2-

(B3)
We compute
r-2A"2

2AEF

A,r-1(Sr-') = -2BE2

For convenience, let

-2AE2

. .

2BEF -2BE2

. .

L 2CEF -2CF1

2CEF . . J
(B.s)

(B4)

E
and
-4AE2(E+F)

4AEF(E+F)

-AAE*(E+F)

. .

ftr-'(Sr-')2= -4BE2(E+F)

ABEF{E+F)

-lBE2{E+F)

. .

L 4CEF(E+F)

-4CF>{E+F)

. . J

continuation) without restriction on the size of G.

Substituting (BO) and (B8) into (Bl2) yields

Comparison of (B9) with (B8) shows that

Al^-,(ST-,)2=2('+/')A>^-,(S^-,)

4CEF(E+F)

(B9)

(Bio)
p0(r+y0)-' = -\\+2G(E+F)]->

so that
'A
t30r-'(sr-1)k+l=2*(E+F)%T-,sr-1

2GEF B
c

allowing (B3) to be evaluated as

/3o(r+7o)-1=/3or-'-^/3r-'Sr-1
{l-2G(E+F)+4G2(E+F)2- . . .}
-^-i^E+F)^-^-1

A A..'

(Bii)

(B12>

We note that the result (Bl2) is valid (by analytic

148-189

AE

B
c
AE

+ BE 0 BE
^0
CF 0

B. . .
r.
0 . .
y
0 . .
CF...JJ

(Bi3)

(Paper 69C3-197)

UNITED STATES DEPARTMENT OF COMMERCE Charles Sawyer, Secretary

NATIONAL BUREAU OF STANDARDS E. U. Condon, Director

Precision

Resistors

and

Their

Measurement

by James L. Thomas

National Bureau of Standards Circular 470

Issued October 8, 1948

149-1

Preface
There are few fields of scientific investigation in which accurate measure
ments of electrical resistance are not required. For this purpose the Wheatstone bridge, in one of its several forms, is almost universally used. This instru
ment is comparatively simple to use and at the same time has a very high sen
sitivity. Unfortunately, however, sensitivity and accuracy are not synonomous, and some knowledge of the practical limitations of resistance-measuring
bridges is needed by all users. This circular is intended to supply such infor
mation. Although the presentation is essentially nontechnical, it is believed
that the subject matter will be of value to any one interested in the accurate
measurement of resistance
In addition to information about the use of resistance bridges, this circular
presents methods for their calibration. The subject matter is limited to directcurrent calibrations, and the methods discussed are those regularly used at this
Bureau when an accuracy of 0.01 percent or better is required. No attempt
is made to present a complete discussion of methods of resistance measurement
or to consider the relative merits of various methods. Those presented are
comparatively simple, being based largely on substitution procedures, yet they
are capable of yielding results of high accuracy.
E. U. Condon, Director.

15 0-11

Contents
Preface
I. Introduction
1. Definition of resistance
2. Importance of resistance measurements
3. Types of resistors
II. Resistance Materials and Construction Methods
1. Resistance alloys
2. Spools, winding and adjustment
3. Sheet-metal resistors
4. Accelerated aging
5. Annealed resistors
6. Effects of humidity
7. Load coefficients
8. Stability of resistors with time
III. Methods of Comparison of Resistors
1. Ammeter-voltmeter methods
2. Ohmmeters
3. Potentiometer method
4. Differential-galvanometer method
5. Bridge methods
IV. Special Apparatus for Precision Measurements
1. Direct-reading ratio set
2. Universal ratio set
V. Calibration of Precision Bridges
1 . Calibration of Wheatstone bridges
2. Calibration of thermometer bridges
VI. Resistivity of Solid Conductors
1. Resistivity, definition and units
2. Measurement of resistivity
VII. References

315-4U O - 68 - 11

151-III

.-.

-----

"
1
1
1
2
2
2
4
5
6
6
7
9
9
10
10
11
11
12
13
19
19
26
28
28
29
31
31
31
32

Precision Resistors and Their Measurement

By James L. Thomas
Abstract
This circular contains information on the construction and characteristics of wire-wound
resistors of the precision type. There are also included descriptions of the methods used at
this Bureau for the test of precision resistors and the calibration of precision resistance
measuring apparatus. Although the presentation is nontechnical, there is a considerable
amount of information on the characteristics and limitations of apparatus of this type that
should be of interest to any one making accurate measurements of electrical resistance.
I. Introduction
1. Definition of Resistance
The modern concept of electrical resistance is
based largely on the work of G. S. Ohm, who in
1826 published an equation that was formulated
on the basis of his experiments with direct-current
circuits. In modern terminology this equation is
l=yfe,

(i)

where E is the potential difference across a con

ductor of length I and cross-sectional area A when
a current, /, flows. The factor 7 depends upon the
material of the conductor and is now called con
ductivity (see sec. VI). The above equation was
abbreviated by Ohm as

where J/' is the length of a hypothetical wire

having unit conductivity and cross-sectional area.
y was then the length of a wire across which unit
current would produce a unit of potential difference,
that is, If was Ohm's unit of resistance and his
experimental equation can be abbreviated as
E=RI.

(3)

This equation is almost universally referred to as

Ohm's Law, although some writers contend that
the longer form should be so designated. Although
experimentally determined for individual con
ductors, Ohm's Law was soon applied to the entire
circuit if E designates the net electromotive force
in the circuit.
2. Importance of Resistance
Measurements
Although Ohm found that the ratio of potential
difference across a conductor to the current flow

ing in it is dependent on the material and dimen

sions of the conductor, more precise experiments
showed it to depend upon temperature and even
upon the presence of stress in the conductor.
Instead of incorporating such quantities in our
equation for the current, we say that the resistance
is a function of temperature and stress, and for a
given conductor we must state the values of
temperature and stress for which a value of
resistance is given.
The fact that resistance is a function of tem
perature is made use of in temperature measure
ments, the resistance of a wire being measured at
known temperatures or fixed points and then at
the unknown temperatures. Also, changes in
resistance with dimensions are utilized for measur
ing small displacements, and change in resistance
with stress is utilized for the measurement of
liquid pressures. In addition to such phenomena,
a large number of physical and chemical phenom
ena are investigated by means of measurements of
electromotive forces, and the measurement of
electromotive force is customarily carried out by
measurements of resistance ratios. Electric cur
rent is readily measured in terms of the potential
difference across a known resistor. In fact, a
large proportion of electrical quantities is meas
ured by methods that involve the measurement
of resistance.
In some cases the resistance of a conductor
depends upon the magnitude of the current flowing
through it. That is to say, Ohm's law is not
applicable, and the resistance must be determined
under the conditions of use. Also, the resistance
of all conductors is to some extent a function of
frequency, and the resistance must often be deter
mined in such a way as to allow for the effect of
frequency. However, the content of this circular
is limited largely to direct-current measurements
of resistors which follow Ohm's law, with special
emphasis on precision measurements, i. e., meas
urements to 0.01 percent or better.

152-1

3. Types of Resistors
Resistors are used for many purposes with a
correspondingly large range of types and accuracies
of adjustment. They are used as electric heaters,
as current-limiting devices such as motor starters,
for component parts of radio, telephonic and
similar equipment, and in electrical instruments of
greater or less precision. For some applications
adjustments must be made to perhaps 10 to 20
percent, whereas in others the resistors must be
within 0.01 percent or less of their nominal values.
For use as heaters, resistors are usually made of
special alloy wire that will withstand high temper
atures for long periods of heating. The most used
alloy for this purpose is of nickel and chromium,
with or without the addition of a considerable
amount of iron. These alloys can be kept at a
"red heat" in air for long periods of time without
damage from oxidation. High-resistance units for
radio circuits are often made from a nonconducting
binder, such as clay, with which is mixed sufficient
powdered graphite to make the material slightly
conducting; or from a nonconducting rod on the
surface of which is deposited a conducting film.
For resistors to be used in precision instruments
the important quality desired is stability with
time and temperature, and special alloys have been
developed, which are discussed in section II.

Stability with time and temperature are also the

important characteristics required for resistors
that are to be used as reference standards for the
calibration of other resistors.
Standard resistors are usually commercially
available only in decimal multiples or fractions of
the unit, usually from 0.0001 ohm to 10,000 ohms.
These are of two general types of construction,
either two-terminal or four-terminal types. For
values of resistance where the resistance of the
contacts, made in connecting to the resistor, is not
negligible, it is customary to use four-terminal
resistors. Whether or not the contact resistances
are negligible depends upon the accuracy desired,
but in general standard resistors of 1-ohm and less
are of the four-terminal type, higher valued re
sistors needing only two terminals. Special stand
ard resistors are required if an accuracy of better
than 0.01 percent is required. These are usually
sealed to protect the resistors from the effects of
oxygen and moisture in the atmosphere, whereas
resistors for an accuracy of 0.1 percent do not re
quire such protection. Although precision resistors
are ordinarily adjusted to their nominal values to
0.01 percent or better, they gradually change in
resistance with time. For work to this accuracy it
is desirable to have standards recalibrated occa
sionally, applying if necessary corrections corre
sponding to their departure from nominal values.

II. Resistance Materials and Construction Methods

1. Resistance Alloys
Manganin. Since their introduction in about
1890, alloys of copper, manganese, and nickel
have come into almost universal use as resistance
materials for precision resistors and for resistance
measuring apparatus. The most common of these
alloys is "manganin" which has the nominal
proportions of 84 percent of copper, 12 percent of
manganese and 4 percent of nickel. This material
has a resistivity of 45 to 50 microhm-cm, a thermoelectromotive force against copper of 2 or 3
/iv/ C, and, when properly heat treated, is very
stable in resistance with time.
The electrical properties of alloys of copper,
manganese, and nickel over a large range of pro
portions were investigated in 1925 by Pilling [1] 1
and later for alloys made of high-purity ingredi
ents by Dean and Anderson [2]. Both of these
investigations showed that an alloy having a
temperature coefficient that averaged zero over
the interval 0 to 100 C. would be obtained with
approximately 10 percent each of manganese and
nickel, the remainder being copper. A series of
alloys also having zero temperature coefficients
could be obtained by increasing separately either
the manganese or the nickel up to as much as
1 Figures in brackets indicate the literature references at the end of this
sper.

20 or 30 percent, with a corresponding reduction

in copper.
On the basis of small temperature coefficients
there would appear to be a wide choice of com
positions for alloys of the manganin type. How
ever, in order to keep the thermoelectric power
against copper as low as possible it is necessary
to keep the nickel content low, as the thermo
electric power increases rapidly in proportion to
the amount of nickel above 2 or 3 percent. Also,
if the nickel is kept constant and the percentage
of manganese is increased the curvature of the
resistance-temperature curve increases.
This
means that although the resistance might have
the same value at 0 and 100 C, the departure
from this value at intermediate temperatures
increases with the manganese content. It is
therefore desirable for general use to have an alloy
that is as low as possible in both nickel and
manganese.
The published data on copper-manganesenickel alloys give average temperature coefficients
over a temperature interval of some 80 to 100 C.
These data do not show the best compositions for
use at ordinary laboratory temperatures. If it
is desired to keep the thermoelectric power against
copper as small as possible, the alloy should have
a content of about 2 percent of nickel and 14
percent of manganese in order to obtain at the

153-2

same time a small temperature coefficient of

resistance at ordinary room temperatures. On
the other hand, if the thermoelectric power is of
no importance the resistance of an alloy of about
20 percent of nickel and 10 percent of manganese
would be most constant with temperature in the
ordinary range of laboratory temperatures. The
accepted composition of manganin, 84 percent of
copper. 12 percent of manganese and 4 percent of
nickel is reasonably close to the optimum for a
general purpose resistance alloy. Fortunately
the proportions need not be exact, as the melting
losses make the composition somewhat difficult
to control.
If the resistance of a sample of manganin is
plotted against temperature, the curve will be found
to be of the general shape of that shown in figure 1 .
;.oi

-100

Figure 1.

100

200
300
400
MO
TEMPERATURE *C
Resistance-temperature curve for manganin.

The maximum, M, of this curve is in the neighbor

hood of room temperature, being generally be
tween 20 and 50 C, while the minimum, JV, is
at about 350 C. Somerville [31 found the differ
ence in resistance between the maximum and the
minimum to be about \>A percent of the value at
the maximum.
For an interval at least 15 to'20 C on each side
of the maximum, the resistance-temperature
curve for manganin is symmetrical with respect
to a vertical axis. Its equation can be accurately
represented by
R,=R25[l + a(t-25) + P(t-25y),

(4)

where R, is the resistance at <C and R2S is the

resistance at 25 C. The coefficient a is the slope
of the resistance-temperature curve at 25 C, and
for manganin of good quality a has a value of
10X10"" or less. The value of 0, which deter
mines the curvature at any point in the interval,
is usually between 3X10"7 and 6X10"7.
This means that 10 C either side of the maximum
the resistance is less than at the maximum by
from 30 to 60 parts per million (ppm).
The temperature at which the maximum of
resistance occurs is a function of the thermal and
mechanical treatment of the manganin as well
as of its composition. Different size wires drawn
from the same melt will have maxima at different
temperatures, that is, at any given temperature
their coefficients of resistance will not be the

same. From the ingot, manganin is usually

worked cold with an occasional softening by
heating to a "red heat." If this annealing is done
in air there is a selective oxidation of the surface
that breaks down the alloy and leaves a coating
with a relatively high conductivity. This surface
layer may be removed by "pickling" in an acid,
but again the action of the acid is somewhat
selective, and the surface layer is left with a
slightly different composition from that of the
interior. These complications add to the diffi
culty of the control of the quality of manganin
wire. In addition, the completed resistance coils
are usually baked at about 150 C for 24 to 48
hours in order to stabilize the resistance. This
baking also affects the temperature coefficient of
the wire by amounts depending upon the size
of the wire and the length of time of the baking.
All of these factors make it difficult to produce
resistors with the maximum at the best tempera
ture, usually between 20 and 25 C. However,
precision resistors of good quality usually can
be obtained with temperature coefficients in that
interval of not more than 10 ppm/ C.
Temperature coefficients of resistance change
little with time, and if changes do occur in their
values there is very probably an accompanying
large change in resistance. Standards that are
stable in resistance need not have redetermina
tions of their temperature coefficients. Although
the data are meager and only the order of magni
tude is known, it appears that for manganin
resistors a change in resistance of one part in a
thousand will be required to change the slope
of the temperature-resistance curve at room
temperature by one or two parts in a million
per degree centigrade.
Constantan.A series of alloys of nickel and
copper containing 40 to 60 percent nickel, with
a small amount of manganese to improve their
mechanical properties, all have practically the
same electrical properties. These alloys are
sold as "constantan," or under various trade
names, for use as thermocouple materials, and
have thermoelectric powers against copper of
about 40 aW C. However, except for their
large thermoelectric powers the electrical prop
erties of these alloys are remarkably similar to
those of manganin.
The resistance-temperature curve for constantan
is similar to that for manganin shown in figure 1 .
Its maximum is at or near room temperature with
the minimum around 500 C. The difference in
resistance between the maximum and minimum is
somewhat less than for manganin, and the curva
ture in the neighborhood of the maximum is also
less. As a consequence, constantan changes some
what less in resistance over the ordinary range of
atmospheric temperatures than does manganin.
Its stability with time is about the same as that of
manganin. At room temperature the resistivity

154-3

of constantan is 45 to 50 microhm-cm as it is for

manganin.
Resistance coils of constantan are sometimes
used instead of manganin in values of 1 ,000 ohms
and above. They may also be used in smaller
denominations in cases where no difficulty will
arise from the large thermal emf's, as for example
in alternating current circuits.
Therlo. When manganin was first developed a
small amount of nickel was added to the copper
and manganese in an attempt to reduce the ther
moelectric power of the alloy against copper.
With the proportion of nickel now used the ther
moelectric power against copper at room tempera
ture is almost the same, although of opposite sign,
as when the nickel is omitted. However, the
nickel improves the mechanical properties of the
alloy and probably reduces the surface action
during forging and annealing.
The development of therlo was another attempt
to reduce the thermoelectric power of coppermanganese alloys against copper. Instead of
nickel, an equal percentage by weight of aluminum
was added to the copper and manganese. As its
name implies, the resulting alloy had a very small
thermoelectric power against copper at room tem
peratures, less than ljtv/C. Its other electrical
properties are almost identical with those of man
ganin. However, as the thermoelectric power ol
manganin against copper is only of the order of 2
or 3jiv/C, the improvement was of little signifi
cance.
At the National Bureau of Standards [4] an
investigation has been made of copper manganesealuminum alloys of the therlo type. There it was
found that for a resistance alloy the best composi
tion is 85 percent of copper, 9.5 percent of manSanese, and 5.5 percent of aluminum. This alloy
as nearly the same resistivity as manganin, its
temperature coefficient at 25 C can be brought to
zero by a suitable heat treatment, and the change
in the temperature coefficient with temperature is
about half that of manganin, at least in the ordi
nary range of room temperatures. Its thermo
electric power against copper at 25 C is only
about 10 percent of that of manganin, and this
thermoelectric power may be further reduced by
the addition of a very small percentage of iron,
without materially affecting the other properties
of the alloy. The stability of such alloys with
time was found to be equal to that of manganin.
Except for unusual applications, the difference
between manganin and therlo is of little importance
and the alloys may be used interchangeably.
Gold-chromium.An alloy of recent introduc
tion, which appears to be very promising for some
applications, is gold with slightly over 2 percent of
chromium [5]. This alloy has a resistivity at
room temperatures of about 20 times that of cop
per. By baking at fairly low temperatures the
temperature coefficient can be made extremely

small. Resistors of this material have been

produced such that the total change in resistance
in the interval 20 to 30 C did not exceed a few
parts in ten million. The thermoelectric power of
this gold-chromium alloy against copper is several
times that of manganin, being 7 or 8 juv/ C at
25 C. The stability of this alloy with time has
not been thoroughly tested but preliminary results
were promising.
For many applications the extremely small
temperature coefficient of gold-chromium alloy
makes its use desirable. However, the tempera
ture coefficient must be adjusted for each coil by
baking, and the cost of this adjustment limits the
use of the material. Although the temperature
coefficient may be made small at room tempera
ture, the interval over which the coefficient is
small is not more than 20 or 30 C. The tem
perature-resistance curve is similar to that for
manganin, as shown in figure 1 , but with M and N
much closer together in temperature and in re
sistance.
Other Alloys.-It is probable that all resistance
alloys that have small temperature coefficients at
room temperature have temperature-resistance
curves that are cubics, similar to that of manganin
shown in figure 1. These curves are nearly
straight in the neighborhood of the inflection
points between the maxima and minima. The
ideal resistance alloy for use in instruments would
have this inflection point at room temperature
with a zero slope. Moreover, the maximum and
minimum should be at widely separated tempera
tures so that the zero slope would be obtained over
the usual range of atmospheric temperatures. Of
the alloys already discussed, only gold-chromium
has the inflection point in the neighborhood of
ordinary room temperatures, the others having
small slopes because of use near the point of
maximum resistance. None of these have small
coefficients over a very large temperature interval.
Alloys of nickel and chromium arc commercially
available that have practically linear temperatureresistance curves over an interval of several
hundred degrees centigrade, which interval in
cludes ordinary atmospheric temperatures. Al
though the temperature coefficient is constant, it
is too large for use in apparatus where the highest
accuracy is required. Recent attempts to reduce
the coefficients of these alloys by the addition of
comparatively small amounts of other materials,
such as copper and aluminum., appear very promis
ing. It is quite possible that an alloy and heat
treatment will be developed such that no correction
for temperature will need be made, at least
throughout the range of laboratory temperatures.
2. Spools,

Winding and

Adjustment

In the beginning of the electrical instrument

industry, wire coils were wound on wooden spools

155-4

like those that arc still used for thread. How

ever, because of the demand for increased accu
racy, these wooden spools have been entirely re
placed by metal spools for resistors of high quality.
The reason for the change to metal has been two
fold. In the first place, the wooden spools absorb
moisture in amounts dependent upon the humidity
in the air, and expand or contract therefrom. This
results in varying stresses applied to the wire, with
accompanying changes in resistance.
A more important reason for the use of metal
spools is the fact that they more readily dissipate
the heat produced in a coil by the passage of the
current. The wire is wound in rather intimate
thermal contact with the metal spools, and the
heat is readily transferred to the spools. The en
tire surface of the spools, both inside and outside,
is effective in dissipating heat to the surrounding
air. However, for wooden spools the area that is
effective in dissipating heat is largely the exposed
outer surface of the resistance wires. When metal
spools are used, the temperature rise for a given
heat dissipation depends primarily upon the size
of the spool and only to a minor extent upon the
size of the wire. However, the wire size should
be selected so as to cover the spool as completely
as possible, and if necessary the turns should be
spaced to prevent bunching of the coil at one end
of the spool.
For many alternating-current applications, the
use of metal spools is undesirable, or even out of
the question. For such applications, when wirewound coils are required, wooden spools may be
used, although for these purposes ceramic spools
have come into rather general use. The objections
to ceramic spools for resistors of high precision
arc their poor heat conductivity and the fact that
their temperature coefficients of linear expansion
are very much smaller than for the resistance wire.
Metal spools are ordinarily of brass, which has
a coefficient of thermal expansion nearly the same
as that of the resistance alloys. This avoids large
changes in stress in the coils because of tempera
ture changes. The spools are ordinarily mounted
with their axes vertical, and both ends should be
left at least partially open in order to allow a
ready flow of convection currents of air through
the spools. Before being wound, the spools are
enameled or covered with a single layer of silk,
which is impregnated with shellac varnish and
allowed to air dry.
The resistance wire is generally double-silk or
silk and cotton covered, and often the wire is
enameled before these coatings are applied. The
correct length of resistance wire is cut, doubled at
its center, and the center is attached by means of
a thread near one end of the insulated metal spool.
The two halves are then wound sidc-by-side (bifilarly) after which the free ends arc tied down
with silk thread.
High-quality resistors are wound with only one

156-5

layer of wire. Although this requires the use of

smaller wire than for multilayer coils, there, are
several advantages. In the first place, the heat
dissipation is more satisfactory for a single-layer
coil, since a considerable temperature rise may be
obtained in the center of a multilayer coil as a
result of the passage of the current through the
coil, and the load coefficients are usually large.
Moreover, multilayer coils are more subject to
change in resistance because of changes in atmos
pheric humidity (sec section II, 6), and are
usually found to be less stable in resistance with
time.
After being wound, the coils are artificially aged
by baking in air at about 150 C for 48 hours,
after which they are kept for a considerable period
of time before final adjustment. Some manu
facturers impregnate coils before baking with a
shellac varnish, while others impregnate them
after baking with special waxes. The final adjust
ment is usually accomplished in two steps. The
excess wire is cut off in order to make the resist
ance just slightly less than the nominal value.
Copper lead-wires, wliich are usually somewhat
larger in diameter than the resistance wire, are
then silver-soldered to the ends of the coil.
Final adjustment is made by filing or scraping the
resistance wire near the end, care being taken to
see that the metal cuttings are not forced into the
insulation. Any filed part is finally painted with
shellac varnish, which is allowed to air dry. For
coils of the highest precision the interval between
baking and final adjustment should be as long as
practicable, an entire year being desirable.
3. Sheet-Metal Resistors
Precision resistors having values of 0.1 ohm or
below are usually made of sheet manganin brazed
or silver-soldered to heavy copper terminal posts.
Potential leads are attached as shown in figure
2. Here C and (J are copper rods, with binding
posts at the top, attached to the sheet of resistance
material, S. The resistance material may be a
single sheet of manganin or several sheets in
parallel. The sheets are often not straight but
are bent in an S-shape in order that greater
lengths can be used. Additional binding posts
P, P arc used as potential terminals and are
connected by copper wires soldered at some point
on the copper terminal bars. For a resistor of
this type the resistance is measured between the
two branch points. This is to say, the resistance
is equal to the ratio of the potential difference
between the potential terminals P, P to the current
flowing in and out the current terminals C, C.
For standards, sheet-metal resistors are mounted
in perforated containers with hard rubber or
bakelite tops. These are often used in oil baths
in order to facilitate the dissipation of the heat

are lowered on the current posts [6]. This effect

is large when the resistance of the copper terminal
posts is high in proportion to that of the resistance
element. For standards of very low resistance,
large temperature coefficients may result unless
the potential branch points are carefully located.
4. Accelerated Aging

Figure 2. Sheet-metal resistor.

caused by the flow of current through the resistance
element.
An accelerated aging of these resistors is ob
tained by heating in air at 150 C for 48 hours, as
is done for wire-wound resistors. Final adjust
ment of the resistance may be accomplished by
changing the points of attachment of the potential
leads, the resistance decreasing as the points of
attachment are lowered. Increases in resistance
may be obtained by scraping or by drilling holes
in the resistance material. Small changes can
be obtained by filing the copper current posts,
but the direction of the change is dependent upon
the location of the potential leads.
Adjustment of sheet metal resistors is often
accomplished by making U-shaped saw cuts in
the sheet, with the open ends of the U's pointing
in the direction of flow of the current. The Ushaped tabs so formed are bent outward and to
them are attached the potential leads. Adjust
ment is made by increasing the length of the U
cut. Lengthening one cut will increase the fourterminal resistance, whereas lengthening the other
cut will decrease the resistance.
Not only may the resistance of a four-terminal
resistor of the type shown be altered by changes
in the points of attachment of the potential leads,
but it is possible to change the temperature co
efficient of resistance in the same way. In fact,
the temperature coefficient is lowered at the same
time as the resistance if the points of attachment

After being wound, resistance coils are impreg

nated with a shellac varnish and given an accel
erated aging by baking. The temperature for
baking is limited by the silk insulation, which
should not be heated above about 150 C, and the
coils are usually baked at this temperature for 48
hours. As a result of the baking the resistance of
a coil may decrease, sometimes as much as 1 or
2 percent, and a sufficient length of wire must be
used to compensate for this change. After being
baked, the resistance of the coils is much more
stable with time than is that of unbaked coils.
This aging process is often called "annealing",
but it is doubtful that the improvement in stability
results from the relieving of internal stresses in
the wire, as happens during true annealing. In
the two or three months immediately following
their baking, resistance coils will ordinarily in
crease in resistance by an amount that is usually
of the order of 0.01 percent. They then are ready
for final adjustment.
Sheet metal resistors are usually painted with
a lacquer or shellac varnish as a protection of
the surface. This coating limits the temperature
to which these resistors can be raised during aging,
and they are usually treated in the same way as
insulated wire-wound coils, being baked at 150 C
for 48 hours. Sheet resistors may be heated at
high temperatures before being lacquered, but
such treatment is apparently no improvement
over baking at 150 C, as far as subsequent
stability is concerned.
5. Annealed Resistors
As has already been stated, the usual baking of
wire-wound and sheet-metal resistors is not done
at a sufficiently high temperature to anneal the
resistance material. As baking at 150 C im
proves the stability with time, it would be logical
to expect greater stability if the heating were
carried on at a sufficiently high temperature to
obtain actual annealing. In the case of manganin
this takes place between 500 and 600 C.
* The annealing of metal is a complicated process,
but the first step is probably a reforming of the
metallic crystals to the shape they had before
being cold-worked. This results in the relieving
of many of the internal stresses that resulted from
the distortions of the crystals. The amount of
this restoration of the metal to its preworked
condition depends upon both the annealing tem

157-6

perature and the time. In general, the higher

the temperature the shorter the time required
for a given annealing. If the annealing is con
tinued after crystals have been restored to their
original condition, there may result an actual
uniting of adjacent crystals, with an accompany
ing decrease in the mechanical strength of the
material.
For base-metal alloys, annealing should take
place in a vacuum or in an inert atmosphere to
avoid, a reaction between the metal and the sur
rounding air. Such reaction might take place
inside the metal at the intercrystafline boundaries
as well as at the surface of the metal. These
reactions do not necessarily decrease the stability
of resistance with time if the products are stable
There may, however, be a selective reaction that
makes the material inhomogeneous, and this might
have a considerable effect on the resistivity and
temperature coefficients.
When a resistor is made by winding wire on a
spool, the wire usually will not straighten if it is
removed from the spool. This means that parts
of the wire have been stressed past their elastic
limit and a permanent deformation has taken
place. For such a bent wire the portions farthest
from the center of the spool have been elongated
past their elastic limit, while the filaments nearest
the center of the spool have been compressed
beyond the elastic limit. Intermediate filaments
are subjected to stresses that depend upon the
changes in their length, which resulted from the
bending. The stress distribution is from a maxi
mum in tension to zero and then to a maximum
in compression. These stresses are superimposed
upon the stresses that were produced in the wire
as it was being fabricated.
These internal stresses in a wire result in a
slight change in the shape of its cross section. In
addition to this change there is a change in re
sistivity, which results from the presence of the
stress [7]. The resistivity is therefore not uni
form across the wire, and the difference between
parts of the wire may amount to nearly 1 percent
in the case of manganin, and perhaps more for
other alloys. Although the resistivity changes
considerably when a wire is bent, there is not
necessarily much change in resistance, as the
change in resistance of the parts under compression
may compensate for the change in the parts under
tension.
The effect of annealing of a coil of wire is to
reduce the internal stresses, thus making the
resistivity more nearly uniform in the wire. It is
probable that a slow annealing takes place at
room temperatures, and the accompanying reduc
tion in the internal stresses may be one reason
for the change with time of the resistance of a coil
of wire. Another cause for change might well
be some reaction between the wire and the sur
rounding atmosphere. Both of these sources of

instability would be avoided or reduced if a coil

were annealed and mounted in a vacuum.
Annealed resistors mounted in vacuo have been
tested at the National Bureau of Standards and
found to be very stable. It is difficult, however,
to seal the coils in suitable containers for high
evacuation. Equally good results have been
obtained with annealed coils mounted in sealed
metal containers filled with dry air. Whatever
the effect of the air, an equilibrium condition is
soon obtained when the supply of air is limited.
Probably the most stable resistors that have been
made are a group of annealed 1-ohm manganin
resistors mounted in double-walled air-filled con
tainers, now being used at the National Bureau of
Standards [8J for maintenance of the unit of re
sistance.
Although good annealing improves the stability
of sealed resistors, it is apparently of no special
value for unsealed coils. When mounted in open
containers, annealed resistors cannot be expected
to be any better than, if as good as, those baked
at 150 C. This is true even if the resistors are
varnished or enameled after the annealing. None
of these coatings is impervious to the atmos
phere, and they merely retard any reaction be
tween the air and the resistance material.
6. Effects of Humidity
It has long been known that wire-wound re
sistors undergo seasonal variations in resistance,
being higher in resistance in summer and lower in
winter. This effect is most noticeable in high
resistance coils of small wire, and even in highgrade resistors may amount to several hundredths
of a percent of the resistance. The effect is to a
large extent a result of changes in average humid
ity, and is greatest in climates where there is a
large difference in humidity between winter and
summer. This seasonal change was first observed
in the case of manganin resistors made with silkcovered wire. The accepted explanation was
that the resistance changes resulted from dimen
sional changes of the shellac with which the coils
were impregnated, as the shellac absorbed or gave
off water vapor.
The effect of moisture on resistors has been
thoroughly investigated by Dike [9], who came
to the conclusion that the effect of changes in
humidity is to change the tension in the silk with
which the wire is customarily insulated. This
change in tension changes the pressure trans
mitted to the wire by the insulation and hence
changes the resistance. He also found that the
effect of humidity on cotton insulation is opposite
to that on silk, and by using a mixture of cotton
and silk fibres for insulating the wire he was able
to eliminate most of the seasonal changes in re
sistance that result from changes in humidity.
For standard resistors the effect of humidity

158-7

Figure 3.

National Bureau of Standards standard resistor-

may be also eliminated by mounting the coils

in sealed containers. This procedure was first
advocated by Rosa [10] who designed the resistor
shown in figure 3, known as the NBS type of
standard resistor. In this figure, R represents
the manganin coil mounted on a silk-insulated
brass spool and baked as described in section
II, 4. This coil is supported from the hardrubber top, T, by means of the thermometer
tube, W, which is so arranged that a thermometer
can be inserted from the outside. The copper
lead wires, which are silver-soldered to the ends
of the resistance oil, are in turn soft-soldered to
the copper binding posts, B. The hard-rubber
top is screwed into the metal container, C, which
is filled with a good quality light mineral oil.
The binding posts, thermometer well and the
threads by which the hard-rubber top is connected
to the container are all sealed with shellac, which
is not soluble in oil.
The purpose of the oil is to give good thermal
contact between the resistor and the case and to
facilitate the dissipation of the heat developed
in the resistor by the current through it. In
addition, the oil in effect increases the heat
capacity of the resistor, thus increasing the cur
rent that it can carry temporarily without over

heating. The objection to the oil is the fact that

it may in time become somewhat acid, and the
acid may corrode the resistance wire or injure
the insulation.
The advantage of ready dissipation of heat
combined with the advantage of hermetic sealing
is found in the double-walled type of standard
resistor [8] developed at the Bureau. In this type
the container is made of coaxial cylinders only
slightly different in diameter witli the space
between the cylinders sealed. The resistance
element is mounted in this sealed space in good
thermal contact with the smaller cylinder, which
serves as the inside wall of the container. One
of these resistors is shown in figure 4. The outside
diameter of the container is 9 cm and its length
13 cm. The series of holes near the top are just
above the double-walled part and are intended to
increase the facilities for cooling, and the con
tainers are left open at the bottom for the same
purpose. These double-walled resistors readily
give up heat to an oil bath yet are not affected
by humidity changes. The sealed space in which
the coil is mounted is filled with dry air, and no oil
comes in contact with the resistance1 material.
The seasonal changes in resistance of standard
resistors that arise from changes in humidity
are readily eliminated, as has just been discussed,
by sealing in metal containers. This arrangement
is not satisfactory for large instruments and meas
uring apparatus, which are not readily sealed.
If the cases of such equipment are reasonably
tight a drier, such as calcium chloride, may be
kept inside the case. This procedure is somewhat
hazardous, since if not replaced with a sufficient
frequency the drier may become dissolved in

159-8

Figure 4.

Double-walled standard resistor.

adsorbed water and be spilled on the coils. The

use of silica gel avoids this danger. A more
satisfactory method is to mount a heater in the
apparatus and, by means of a thermostat, main
tain the temperature of a metal box containing
important resistors at a constant value, above that
of the laboratory. In this way the relative
humidity is kept always low and seasonal varia
tions may be reduced. Additional advantages
arc that no temperature corrections need be
applied when calibration is made under conditions
of use, and that resistors will usually be more
stable in resistance if kept at a constant tempera
ture instead of being allowed to follow variations
in laboratory temperatures.
7. Load Coefficients
The load coefficient of a resistor is defined as
the proportional change in resistance caused by
the production of heat in the resistor at the rate
of one watt. This change, however, is a function
of time. When a current starts flowing in a
wire-wound resistor the wire very quickly takes
up a temperature above that of the spool on which
the wire is mounted. The amount of this change
depends upon the thermal contact between the
wire and the spool. If the current continues to
flow, this initial difference in temperature is
maintained, but the coil and spool continue to
rise in temperature together. The maximum
tempera turc that will then be attained depends
upon the facilities for dissipation of heat by the
spool, and often upon the facilities for cooling
the material to which heat from the spool flows.
Suppose we have a standard resistor of the NBS
type resting on a table and we send through it a
sufficient current to liberate one watt in the coil.
Within about 30 seconds the temperature of the coil
will rise to a steady value about 1 C above that
of its spool. Spool and coil will then continue to
rise in temperature at the rate of about 10 or 15
C per hour, and if allowed to continue will in about
2 hours reach a steady temperature of some 10
to 20 C above that of the room. This may cause
a permanent change in resistance, which, however,
ordinarily amounts to only a few parts in a million.
If, instead of being mounted in an oil-filled con
tainer, the same resistor had been left open in the
air of the room, its final rise in temperature would
probably have been less by some 50 percent and
would have been reached in 15 to 20 minutes.
The rise in temperature of the resistance coil will
be accompanied by a change in resistance, whose
amount depends upon the temperature coefficient
of resistance of the wire and also upon changes in
stress in the wire because of dimensional changes
of the coil and its support. For single-layer
manganin coils mounted on brass spools the relative
changes in dimensions arc small, and the change in
resistance results primarily from temperature

changes of resistivity. For such resistors the

effect of the heating may usually be taken into
account by measuring the temperature change and
calculating the change in resistance from the
temperature coefficients of the resistor. For
sheet-metal resistors the effect of stress changes
may be important and the calculated change cannot
be relied upon.
It should be evident from the above discussion
that the load coefficient of a resistor is a rather
indefinite quantity, varying with time of flow of
the current and with the environment of the resis
tor. To be of value it should be measured under
conditions of use. Measurement is usually made
by passing the desired current through the resistor
and a second resistor connected in scries, measuring
the ratio of resistance both with negligible and
with the required heating. This ratio may be
measured by means of a bridge or a potentiometer,
and the comparison resistor should be one that
is not appreciably affected by the test current.
A resistor of one-tenth or less the resistance of that
under test should be used, and its load will be
one-tenth or less than for the resistor under test.
For the reference lower-valued resistor one that is
known to have a small load coefficient should be
chosen. If such a resistor is not available one
with a low temperature coefficient should be se
lected, on the assumption that its load coefficient
is correspondingly low, and its load coefficient
should be roughly determined to make sure that
it is small. This can be done by balancing the
resistor in any bridge using a small test current.
From an external source a large current is then
sent through the resistor under test for several
minutes, after which the extra circuit is discon
nected and the bridge circuit is again balanced.
This last balance must be made quickly before the
heat from the large current is dissipated.
If a resistor is being used under conditions where
the load changes its resistance, the amount of the
change may often be determined experimentally.
To do this the heating may be doubled, by in
creasing the current by 40 percent, and the result
ing change noted. To a first approximation this
doubling of the heating doubles the change in re
sistance, and twice the change should be subtracted
algebraically from the final value. Such a pro
cedure is especially satisfactory in the case of
Wheatstone bridges, as the procedure will correct
back to zero test current, whichever branch or
branches are being changed by the current.
8. Stability of

Resistors

With Time

In applications where stability with time is of

importance, as for instance for standard resistors or
precision measuring apparatus, manganin is used
almost exclusively. In such applications low tem
perature coefficients of resistance and small ther
mal emf's against copper arc usually also required,

160-9

and few alloys other than manganin are suitable.

Consequently, a discussion of the stability of re
sistors is largely a discussion of the stability of
manganin.
Most national standardizing laboratories keep a
selected group of manganin resistors, which are
regularly intercompared and used for maintenance
of the unit of resistance. The relative values of
such standards are remarkably constant, the in
dividual resistances not changing by more than
one or two parts in a million per year with refer
ence to one another and for some groups very much
less. It is supposed that the average of a group re
mains constant to a high degree, but such stability
can be only assumed. No method of measurement
has been used that would detect with certainty
changes of less than 10 or 20 parts in a million in
the group as a whole. The international ohm was
defined as the resistance of a mercury column of
specified dimensions at the temperature of melting
ice. However, such resistors have not been con
structed with sufficient accuracy to demonstrate
the performance of manganin resistors used to
maintain the unit. Likewise, absolute ohm. deter
minations have not been sufficiently reproducible
to give such information. If the unit as main
tained by means of manganin resistors were tested
every ten years by comparing against mercury
ohm or absolute olim determinations, the apparent
change would probably not exceed 10 or 20 parts
per million. This could be just as well attributed
to errors in realizing the unit experimentally as to

changes in the unit as maintained by the manganin

resistors.
The average user of standard resistors is interest
ed in the stability that may be expected from
standards available commercially. In this con
nection, an analysis made in 1941 at this Bureau is
relevant. Of nearly 600 standard resistors that
had been submitted more than once to this Bu
reau for test, the average yearly change in resist
ance, without regard to sign, was found to be
8 ppm. Of the total only 2 percent averaged
greater than 60 ppm per year, and for nearly 90
percent of all standards tested the annual change
was 10 ppm or less. If signs were neglected there
was no significant difference between the average
yearly change of sealed and unsealed resistors.
This would appear to mean that sealing merely
reduces seasonal variations without improving the
long-time stability. However, if regard is taken
of sign, the performance of sealed and unsealed
resistors was quite different. In this case the
average yearly change was about .3 ppm for
sealed and about + 5 ppm for unsealed standards.
That is to say, sealed standards about as often
decrease as increase in resistance with time,
whereas the change in unsealed standard resistors
is predominantly upward. In connection with
the sealed resistors it is interesting to note that
practically the same result would have been obtained
if the unit of resistance had been maintained by
supposing the average value of the 400 sealed
resistors had remained constant, as was obtained
by assuming the average of a group of 10 selected
resistors of special construction to be constant.

III. Methods of Comparison of Resistors

1. Ammeter-Voltmeter Methods
Precise measurements of electrical resistance are
made with comparative ease with bridge methods,
an accuracy of a few parts in a million being readi
ly obtained in the comparison of nominally equal
resistances of say 10 or 100 ohms. Although the
actual measurements are rather simple, special
apparatus is required. For many types of re
sistors high accuracy is not desired and the meas
urements may be made with deflecting instru
ments, which are usually available in electrical
laboratories. The most common method, where
an accuracy of 1 or 2 percent is sufficient, is the
ammeter- voltmeter method. In this method a
measured current is passed through the resistor
under test, and the potential difference across its
terminals is also measured. The current is meas
ured with an ammeter and the potential differ
ence by means of a voltmeter, the resistance in
ohms being the ratio of the voltmeter reading in
volts to the ammeter reading in amperes, in ac
cordance with Ohm's Law.

The accuracy that may be attained by the

ammeter-voltmeter method depends upon the
accuracy of the two instruments. However,
there are a few precautions that must be observed.
Referring to figure 5, if the voltmeter, V, is con-

iOi

-^-WNAAAAAA^
Figure 5.

Connection for ammcter-vcltmeter measurement

of resistance.

nected as shown across the resistor, R, the current

read by the ammeter, A, is the sum of the current
through R and the current through the voltmeter.
If the resistance of the voltmeter is large as com
pared with R, the current through the voltmeter

161-10

may be neglected and R may be calculated as the

ratio of the instrument readings. If the resist
ance of the voltmeter is not large as compared
with R, the latter may be calculated from the
equation
R==I(l-E/IR,)

(5)

where R, is the resistance of the voltmeter, its

reading being E volts, and the ammeter reading /
amperes. If the resistance of the ammeter is
known or if it is negligible as compared with the
unknown resistance, the voltmeter may be con
nected across both the resistance and the ammeter,
as shown in figure 6. In this case the ratio of

<>

Q-WvWvV
Figure 6.

Alternate connection for ammeter-voltmeter

measurement of resistance.

voltmeter to ammeter readings gives the total

resistance between the points of attachment of
the voltmeter, that is, the resistance of the am
meter and of the connecting leads is included.
That is
R=T Ra,

(6)

where RA is the resistance of the ammeter and of

all lead wires between the points of attachment of
the voltmeter leads.
With suitably calibrated instruments it is possi
ble to measure resistance to 0.1 or 0.2 percent pro
vided the instruments are of such ranges that
large deflections are obtained. For this accuracy
it is usually necessary to calibrate the voltmeter
with the same leads as those that are to be used
in the resistance measurements.
2. Ohmmeters
Ohmmeters are instruments for indicating
directly on a scale, with a minimum of manipula
tion or computation, the resistance of the circuit
connected across their terminals. They are avail
able in a wide range from milliohmmeters reading
to 0.001 ohm to megohmmeters reading to 50,000
megohms. Their accuracy is limited both by the
calibration and reading of the indicating instru
ment and in the simple ohmmeter by their depend
ence upon a fixed value of voltage. They are,
therefore, in general, not suited for applications

requiring high precision such as the determination

of temperature rise or of the conductivity of line
conductors. When used with circuits that are
highly inductive or capacitive, the precautions
appropriate for such resistance measurements
should be observed.
Ohmmeters may be classified according ' to
either their principles, of operation, their source
of energy, or their range.
The principles commonly used are: Simple
ohmmeter, ratio ohmmeter, Wheatstone bridge.
In the simple ohmmeter a source, the voltage
of which is assumed to be definite and to corre
spond to the calibration of the instrument, is
applied to the unknown resistor and the resulting
current causes an indicating instrument to deflect
over a scale. This scale is so graduated that the
pointer indicates directly the resistance in ohms
(or megohms). In many cases provision is made
by a magnetic or electric shunt to adjust the in
strument at one point, usually at zero resistance,
to fit the existing value of the voltage. In some
of these instruments the final indication is by a
vacuum-tube voltmeter, which measures the drop
produced in a very high resistance by the current
through the specimen.
In the ratio meter or "crossed-coil" type of
ohmmeter, the current through the unknown
resistor flows in one of the coils, while the other
carries a current that is proportional to the voltage.
The current is led to the coils by ligaments, which
exert a negligible torque so that the moving system
takes up a position that depends on the relative
magnitude of the currents in the two coils. The
scale can therefore be laid off to indicate resistance
directly, and the indication will be independent of
the voltage used, provided that the resistor under
test obeys Ohm's Law.
The designation "ohmmeter" is also applied
(though perhaps incorrectly) to certain forms of
the Wheatstone bridge in which the dial that
adjusts the balance is calibrated to read directly
the value of the unknown resistance.
3. Potentiometer Method
It was pointed out in section 111,1 that the
ammeter-voltmeter method for measuring resist
ance is complicated by the current drawn by the
voltmeter. Such complications may be avoided
by using a voltmeter of some type that requires
no current from the circuit being measured, i. e.,
electrostatic or vacuum tube voltmeters.
Potentiometers are also suitable for the measure
ment of potential difference when it is desired to
avoid drawing a current from the source of
potential difference. They are especially good in
cases where an accuracy of 0.1 percent or better is
required, as such accuracy is difficult to attain with
deflecting instruments. Where a potentiometer
is available, the ammeter-voltmeter method may

162-11

be modified so as to use the potentiometer to

measure the current through the unknown resistor
as well as the potential difference across it. This
requires the replacement of the ammeter with a
standard resistor, and the measurement of the
potential difference across the standard resistor
yields the value of the current if this potential
difference is divided by the value of the resistance
of the standard.
Actually, if a suitable standard resistor is
available, it is unnecessary to calculate the
current through the unknown resistor. If the
same current flows through the known and the
unknown resistor, the ratio of the potential
differences across the two is the same as the ratio
of the resistances. Hence
X=S~*,

(7)

where X and S are the values of the unknown and

standard resistances, Vx and V, are the measured
potential differences across X and S respectively.
Care must be exercised in using this method to
insure that the current through X and S remains
constant during measurement. This may be
verified by measuring the potential differences
alternately several times.
High accuracy in the measurement of resistance
can be attained with the potentiometer method if
a good potentiometer and good standard resistors
are used. It has an advantage over the ordinary
Wheatstone bridge in that the resistance in terms
of which the unknown is measured may be that
of an actual standard resistor instead of one of the
coils of the bridge. Standard resistors are so
mounted that they are ordinarily more constant
in resistance and hence more accurately known
than are unsealed coils usually used in bridges.
The potentiometer method, however, is more
difficult to use as the currents through the poten
tiometer and in the measuring circuit must both
be kept constant, whereas the balance of a
Wheatstone bridge is independent of the current
flowing through it. In comparing resistors, the
accuracy of a potentiometer is not dependent
upon the accuracy of calibration of the standard
cell used with the instrument.
A type of measurement for which the potenti
ometer method is especially well suited is for the
measurement of four-terminal resistors, which are
parts of complicated networks, where connections
to the resistors must be made through other re
sistors. The resistances in the potential leads,
which connect to the potentiometer, have no
effect upon the balance of the potentiometer.
There uiay be, however, a reduction in sensitivity
unless the damping resistor for the galvanometer
may be changed to allow for these extra resistances.

4. Differential-Galvanometer Method
The differential galvanometer was formerly
used rather extensively for the comparison of
equal resistances. Such a galvanometer has two
separate windings made as nearly the same as pos
sible, so that when equal emf's are applied to the
terminals of the windings, equal and opposite
torques are produced on the deflecting element.
The windings are constructed with two wires
side-by-side, wound at the same time and as nearly
as possible symmetrically with respect to the
magnetic circuit.
For the moving-magnet galvanometer the field
coils are wound in duplicate, and a small movable
coil is usually connected in series with one wind
ing. This moving coil is adjusted in position to
compensate for any lack of equality of the fields
produced by the two windings. Unfortunately
this adjustment is different for different conditions
of use. Moreover the differential galvanometer
of the moving-magnet type has the same handi
caps as others of the moving-magnet type. That
is to say, the damping is difficult to control and
elaborate precautions must be taken to avoid
magnetic disturbances from external sources.
Many of the troubles of the moving-magnet
galvanometer arc avoided in the D'Arson val, or
moving-coil galvanometer, and this coil may be
made in duplicate for differential use. However,
this requires two sets of leads from the moving
element, which are difficult to arrange, and which
also in effect stiffen the suspensions and lower
the sensitivity. With the need for greater and
greater precision, the differential galvanometer
has been gradually discarded, but it still is very
satisfactory for some types of measurements.
In theory, the use of the differential galvanom
eter is very simple. A current, /, is passed through
the standard, S, and the unknown, X, connected
in series, and one winding, G, of the galvanometer
is connected to the terminals of each resistor, as
shown in figure 7. If the galvanometer circuits
were exactly alike in resistance and opposite in
their magnetic effects there would be no deflec
tion if X and S were equal. These conditions on
resistance and magnetic effect are difficult to

is
il
S
l
MMAAAr

1
X
(i
laaa/wv-

Figure 7.

163-12

Connections of differential galvanometer.

I
e
H'M

1
Figure 8.

Kohlrausch's connection of differential

galvanometer.

meet, and it is necessary to devise methods of use.

for which the conditions need be only approxi
mately met, if high accuracy is to be attained.
Such is that due to Kohlrausch [11 J, and known
as his method of overlapping shunts.
For Kohlrausch's method the circuit is as shown
in figure 8. The two galvanometer windings, G,
are connected respectively across X and S, and in
addition each shunt bridges the resistance, L,
which is used to connect X and S in series. The
resistance of L is usually small as compared with
the other resistances. In series with the galvanom
eter windings are connected the resistors, Ri and
R2, one or both of which are adjustable. In addi
tion, a special switch not shown must be used,
which will interchange the battery B and the lead
resistance L. This must be done without materi
ally changing the current furnished by the battery.
The effect of the interchange is to reverse the
current through both galvanometer windings.
The values of the adjustable standard, S, and the
rheostat, Rt or R2, are adjusted until there is no
deflection of the galvanometer for either position
of the battery, or until the deflection is the same
in magnitude and direction for both battery posi
tions. Under either condition of adjustment, S
and A' are equal.
In actual practice, where S and X are standard
resistors under comparison, the larger is made ad
justable by a precision rheostat connected in
parallel, and the amount of the shunt required to
make X and <S equal allows an easy calculation
of the difference in resistance between the two.
The balance is exact even if the two circuits of
the galvanometer are not exactly alike electrically
or magnetically.
5. Bridge Methods
By far the largest proportion of measurements
of electrical resistance are made by means of
bridge methods. For resistors above 1 ohm the
simple Wheatstone bridge is used, whereas for
1-ohm resistors and below the Kelvin double-

bridge is more suitable. The choice between the

simple and the double bridge is usually made on
the basis of the required accuracy.
It is difficult to attach a copper lead wire to a
resistor by means of binding posts or other
clamped connections without introducing an un
known contact resistance of the order of 0.0001
ohm. For resistors above 1 ohm such an uncer
tainty is usually negligible, whereas for a resistor
of say 0.01 ohm the uncertainty is 1 percent.
For standard resistors the contact resistance is
often reduced by amalgamating the contact sur
faces. For clean well-fitted contacts the resistance
then amounts to only a few microhms, but the re
sistance of such contacts will rise with time as
the copper combines with the mercury to form a
granular material. This material should be re
moved every few months by scraping and wiping
the surfaces.
Where accuracy is required, low-valued resistors
are usually of the four-terminal type. For these,
two leads are soldered or brazed to each end of the
resistance material, as shown in figure 9. The
p
p

^/WWWWV
Figure 9.

Four-terminal resistor.

resistance in question is that between the branch

points at the two ends. That is to say, the resist
ance is the ratio of the potential difference between
the terminals P, P to the current flowing in and
out the current terminals, C, C. Methods of
measurement are used sucli that any effect from
the lead resistances is avoided or reduced to a
negligible amount.
For such a purpose the
potentiometer method is suitable as no appreciable
current is drawn through the potential leads, and
the potential drop between branch points is inde
pendent of the magnitude of the lead resistances in
the current circuit. Double-bridge methods bal
ance out the lead resistances or connect them in
high-resistance branches where they are negligible.
Simple Wheatstone bridge.The simple Wheat
stone bridge is primarily a group of four resistors
connected in series-parallel as shown in figure 10.
A current, I, is passed through the two parallel
branches, and G is a detector connected to the
junctions of the resistors as shown. It may be
readily shown that if there is no potential differ
ence across the detector, G, the relation between
the resistances of the four arms is as follows:

or

164-13

X=S~

(9)

Figure 10.

Wheatstone bridge.

The detector, G, is usually a galvanometer, and

the lack of a potential difference across it is evi
denced by a lack of motion of the galvanometer
coil if a switch or key in series with Q is opened or
closed, or better, if the battery circuit is opened
or closed. Opening or closing of the battery cir
cuit should be avoided, however, if there is
inductance or capacitance in the circuit of the
unknown resistance or in any arm of the Wheatstone bridge. In measuring the resistance of a
field winding of an electric motor, for example,
the galvanometer may be damaged if its circuit is
left closed and the battery circuit of the bridge is
opened. In this case the current should be left
on until a steady state is reached before testing the
bridge balance by opening and closing the gal
vanometer circuit.
From the above equation it is evident that the
resistance of any one of the four arms, say X, can
be calculated if the resistances of the other three
are known, or if one of the three and the ratio of
the other two are known. The usual general pur
pose Wheatstone bridge is made in such a way that
any one of several known resistors may be selected
for use as either A or B and hence their ratio may
have any one of a number of values. S is then a
known resistor whose value may be adjusted in
small steps over a wide range of resistance.
In commercial instruments the ratio arms A and
B are usually coils that are connected into the
bridge by inserting suitable plugs. As there are
resistances in these plug contacts, which may be
rather variable, it is desirable to use relatively high
values of resistance for A and B in order to reduce
the uncertainty in the resistance of the ratio arms.
On the other hand, high-resistance coils are more
affected by humidity and are less stable in resist
ance with time and therefore retain their calibra

tion for a shorter time. Other things being equal,

ratio coils of from 10 to 100 ohms arc therefore
about the best compromise.
The choice of resistanccfor the adjustable arm
of a Wheatstone bridge is influenced by the same
factors as the choice of resistances for the ratio
arms. That is to say, the lower the value the
greater the stability in resistance but the more
troublesome the contact resistances become, and
in the adjustable arm several contact resistances
are required. Steps smaller than 0.1 ohm, which
allow readings of the adjustable arm to about 0.01
ohm by interpolation from galvanometer deflec
tions, are seldom used as it is not wise to rely upon
the combined effect of the several contacts of the
arm to be definite to much better than 0.01 ohm.
Wheatstone bridges for the measurement of re
sistance to 0.1 percent are available commercially
at moderate prices. To this accuracy these can
usually be relied upon without the application of
corrections to the readings. However, for meas
urements to 0.01 percent, corrections to the read
ings of the ratio arms and of the rheostat arm must
usually be applied, and the bridge must be main
tained at the temperature of calibration within a
few degrees. Primarily because of the effects of
changes in humidity, calibrations of a Wheatstone
bridge must be made rather frequently if an accu
racy of 0.01 percent is to be attained. This is
especially true where there is a marked change in
ambient conditions. For example, between winter
and summer the ratio coils and the rheostat arm
may each change by 0.01 percent or more, and the
errors may be additive rather than compensating.
The use of air-conditioned laboratories improves
the performance markedly, but even then it is
advisable to make occasional spot checks by
measuring standard resistors. Whenever possible,
ratio arms below 10 ohms or above 1,000 ohms
should be avoided, as should rheostat readings in
excess of 1 ,000 ohms. This means that the Wheat
stone bridge is best suited for the measurement of
resistance in the range 10 to 10,000 ohms.
When an accuracy greater than 0.01 percent is
required, special bridges are required or special
techniques are used. One of the best of the special
methods is that of substitution, in which the un
known resistor is replaced with a standard resistor
or resistors having the same nominal resistance as
that of the unknown. The bridge that is being
used is then relied upon to determine only the
difference between the standard and the unknown,
and this difference need not be accurately meas
ured. If the unknown and standard differ by 0.1
percent, the difference need be determined to only
1 percent to give the unknown to 10 parts in a
million in terms of the standard. Nearly any good
bridge can be so used without calibration, provided
contact resistances are sufficiently constant that
readings can be repeated to the desired precision.
It is evident that the calibration of the measuring

165-14

bridge is of least importance when the standard

and unknown resistances are nearest equal.
Standard resistors are usually available only in
decimal multiples or submultiples of an ohm and
for many resistance measurements the lack of a
suitable standard prevents the use of the substitu
tion method.
Mueller bridge.For the accurate measurement
of odd-sized resistances, one of the best bridges
is that designed by E. F. Mueller [12] of the
National Bureau of Standards for use in resistance
thermometry. This is a special bridge intended
for the accurate measurement of resistances up to
about 110 ohms. The effects of humidity changes
are greatly reduced by mounting the resistors in a
compartment that is electrically heated and whose
temperature is maintained at 35 C by means of
a thermostat. As this is a temperature to which
the laboratory temperature seldom rises, the
relative humidity in the compartment housing
the resistors is kept low and variations in the low
humidities have little effect on the resistances of
the coils.
For the Mueller bridge the ratio arms are equal,
and the arms can be interchanged to test the
equality. A small slide wire is connected between
the ratio coils, and the sliding contact is used as
the branch point for the bridge as shown in figure
1 1 . The operator can set the.ratio arms to equality
by properly proportioning the resistance of the
slide wire between the two ratio arms by changing

Figure 11.

Ratio arms of Mueller bridge.

the setting of the slide contact. This setting is

correct if no change in the bridge results when the
arms A and B are interchanged. The remaining
two arms are the adjustable rheostat and the un
known resistor, which must be of the four-terminal
type. Resistance thermometers are usually of the
four-lead type, and other resistors should be pro
vided with four leads when they are to be measured
on a bridge of tliis type. The method of connec
tion of the unknown resistance is as shown in figure
12. R is the adjustable rheostat arm that will be
discussed in detail later. X is a four-terminal
resistor with lead wires Lx, L^, L3, and Lt connected
to the binding posts 1, 2, 3, and 4, respectively. If
the galvanometer is connected to binding post 2,

Figure 12.

Connection of bridge to 4-lead thermometer.

the resistor X and the lead resistor Z are connected

into the right-hand arm, and the lead resistor Lx is
in the adjustable arm, R. With equal ratio arms,
R and X would be equal for a balance of the
bridge only if Lx and Lt arc equal. Instead of
adjusting Lx and Z4 to exact equality, their con
nections are interchanged and a second bridge
balance is obtained, the average of the two balances
being that which would be obtained with Z, and
Lt equal. When Z, and Z, are interchanged, it is
necessary to shift the branch point from lead 2 to
lead 3 in order to keep X in the right-hand arm.
In actual use, instead of the lead wires being inter
changed, the internal connections of the bridge
arms to the binding posts i and 4 are interchanged
to obtain the same result. This interchange is
effected by means of an amalgamated switch for
which the uncertainty in contact resistance is only
a few microhms.
The. rheostat arm of the Mueller bridge is ad
justable in steps as small as 0.0001 ohm. In order
that such steps should not be masked by changes
in contact resistance, special types [13] of decades
are used. The 0.0001-ohm-per-step decade is
approximately as shown in figure 13. With the

>
> I2A

166-15

^WWV
.110/1
Figure 13. Shunt-type decade.

switch contact on the 0 stud, the resistances of 12

ohms and 0.1 10 ohm are connected in parallel, and
the resistance between the terminals S and T is
about 0.1090 ohm. When the switch contact is
moved to stud 1, the 12-ohm branch is increased
sufficiently to increase the parallel resistance from
0.1090 to "0.1091 ohm, an increase of 0.0001 ohm.
Likewise when the switch is moved to 2 the parallel
resistance increases to 0.1092 ohm, etc. Thus the
resistance between the terminals S and T may be
increased in steps of 0.0001 ohm, although the
value is not zero when the switch is set to read zero.
The 0.001- and 0.01-ohm-per-step decades are
made in the same way as the 0.000 1-ohm decade.
When these three decades are connected in series
and all set to zero, there is a series resistance of
about 1.6 ohms. However, since the bridge has
equal ratio arms an equal resistance of 1.6 ohms
may be included in the A'-arm, and the value of
X is measured by observing the increase in the
resistance of the rheostat arm when A' is inserted
in its arm. A short-circuiting plug is provided
with the bridge that connects together the ter
minals 1 and 4 of figure 12. The bridge is then
balanced with and without this shorting plug,
and the difference in the readings of the rheostat
arm gives the resistance of A', and leads.
The purpose of the shunted type of decade that
was described above is to reduce the effect of
variations in contact resistances and transient
emf's in the switch. Referring again to figure 13,
it is seen that the switch contact is in the high
resistance branch. The ratio of resistance for the
two branches is about 100 to 1, and it may be
readily shown by differentiation that the effect
of a variation of switch contact resistance is
reduced by the square of this ratio. Thus a con
tact variation as high as 0.01 ohm would vary the
resistance from S to T by only 1 microhm, which
is negligible in the rheostat arm where the mini
mum steps are 100 microhms. For the 0.001 and
0.01 decades, the ratio of currents in the branches
is less than 100 to 1, and less variation in the
switch-contact resistance can be tolerated.
The 1-ohm and 0. 1-ohm steps of the rheostat arm
are not made of the shunt type, as the resistance
with the dial set on zero would be rather large.
The contact resistances for the 1-ohm dial are
thrown into the ratio arm A, as shown in figure 14.
The bridge current is introduced through the
10 1-ohm coils of this decade, and the switch
merely changes the point of connection of the
ratio arm, without opening the circuit of the 1-ohm
decade. In this case a variation of the switch
contact resistance changes the resistance of ratio
arm A. This effect is reduced by using high
resistance ratio arms, A and B each being 1,000
ohms. The 0.1-ohm-per-step dial is arranged in
the same way as the 1-ohm-per-step dial, with the
switch contact in series with the B ratio arm.
With the 0. 1-ohm dial set at zero, all 10 of the

315-4U O - 68 - 1J

Figure 14. Switch connections for Mueller bridge.

0.1 -ohm resistors are in series in the A'-arm.
When the dial reading is increased, the resistance
is removed from the A'-arm, the effect being the
same as if equal resistance were added to the
rheostat arm, R, since the bridge has equal arms.
This may be seen from the following consideration.
Assume that the bridge is balanced with the un
known resistor and the 10 steps of the 0.1-ohm
dial all connected in the A'-arm. If now A' is
increased by 0.1 ohm, the A'- and /?-arms may be
brought again to equality by either increasing F
by 0.1 ohm or by decreasing the resistance in
series with A' by 0.1 ohm. The readings of the
dial in the A'-arm are such that increases in
readings of the dial correspond to decreases in the
resistance in the A'-arm.
The 10-ohm-per-step decade of the rheostat,
not shown, is the only one that has contact resist
ances directly in series with the arm. A special
type of dial that has amalgamated contacts is used
for this purpose. Such contacts are uncertain by
only a few microhms, which may be tolerated in
the rheostat arm where the minimum steps are
100 microhms. For Mueller bridges available
commercially, the galvanometer and battery are
interchanged from the positions assumed in the
above discussion. This in no way affects the
validity of the conclusions.
Adjustable-ratio bridges. As has already been
stated, the Wheatstone bridge shown in figure 10
is balanced if
x=s*

(10)

and this balance may be realized by keeping A

and B fixed and adjusting S, or by keeping S fixed
and adjusting the ratio A\B. In most commercial
Wheatstone bridges the balance is obtained with
S. However, for comparing nominally equal
resistances, bridges are sometimes constructed in
which the ratio is adjustable.

167-16

For measurement, a four-terminal resistor is

usually connected in series with a four-terminal
standard. The ratio of the potential differences
across the two resistors is then determined when a
current flows through the two. This ratio may be
determined by means of a potentiometer or in
terms of a resistance ratio by means of the Kelvin
double-bridge.
The circuit of the Kelvin double bridge is
shown in figure 16, where X and S denote the
unknown and standard resistors, each with both
current and potential terminals. The two are
connected by means of a conductor L, preferably
of low resistance as compared with X or S. The
resistors A and B and also a and b provide resist
ance ratios that must be known, as must also be
the value of S. This bridge is balanced if there
is no change in the deflection of the galvanometer
when the current circuit is opened or closed.
When balanced the following relation holds
between the values of the resistance:

Figure 15.

Limit bridge.

An adjustable-ratio bridge is very convenient

for use in sorting large numbers of resistors that
must be equal within given tolerances. Such a
bridge may be constructed as shown in figure 15,
where A and B are the two arms of a resistance
ratio that is adjusted by changing A. A resistor
is connected to the binding posts S, and the resist
ors that are to be compared with it arc connected
in turn to the X terminals. If the resistors are
to be the same to. say, 0.1 percent, the arm A is
made to have either of two values, which give
ratios, A/B, of 0.999 or 1.001. This could be
realized by making B 100 ohms and A either 99.9
ohms or 100.1 ohms, the latter accomplished, for
instance, by making A equal to 100.1 ohms and
having a plug to short-circuit 0.2 ohm of this total.
The procedure is then to observe the direction of
deflection of the galvanometer with either ratio
setting and then change to the other ratio. If
this causes a reversal of the direction of deflection
of the galvanometer, the correct balance of the
bridge is between the two ratio settings. In other
words the ratio of X/S is between the two ratios
0.999 and 1.001, i. e., the two are equal within
0.1 percent.
Besides this use in limit bridges, adjustable ratio
bridges may be used for precise measurements of
resistors in terms of nominally equal standards.
Such bridges when used in conjunction with the
substitution method permit very accurate meas
urements of resistance. Such an adjustable ratio
is the "direct-reading ratio set" described in section
IV, 1, where the appropriate techniques are also
discussed.
Kelvin double-bridge.As has been already
stated, four-terminal resistors are used in order
to avoid uncertainties arising from variations in
contact resistances. For precision work, resistors
of 1 ohm or below are usually of the four-terminal
type.

ii-t)
X-SB+~i+b+L

(ID

This equation would be exactly the same as that

for a simple Wheatstone bridge if the last term
on the right side were zero or negligibly small as
compared with the term S A/B. That last term
can be made zero by making ajb=AjB, irrespective
of the value of the link resistance, L. However,
the smaller the value of L, the less important is
any lack of equality between the ratios A/B and
a/6.
In actual use two methods are employed for
balancing the double-bridge. The resistors A
and B, and also a and 6 may be fixed ratio coils
so chosen that A/B=a/b and the bridge balanced
by varying S, which is an appropriate adjustable
standard; or S may be a fixed standard resistor
and the bridge balanced by adjusting the ratios
A/B and a/6, keeping these ratios at all times
equal.
When a double-bridge is to be balanced by means

168-17

Figure 16.

Kelvin double-bridge.

of an adjustable standard, care should be taken

in connecting up such a standard in order to keep
the resistance small between it and the unknown.
Adjustable standards are usually made such that
the position of one or both of the potential
terminals is adjustable. As a result, varying
amounts of resistance are left at the ends in series
with the used part of the resistor. The adjustable
standard should be connected in the circuit in
the way that connects the larger part of unused
resistance in the external circuit rather than
between resistors so as to form part of the link
resistance, L.
When a fixed standard is used for S, and the
bridge is balanced by adjusting the ratios A/B
and a/6, the most convenient arrangement is to
have the ratios adjustable together. This is
accomplished by having the same dial handle
operate two dials together so that both ratios are
changed simultaneously. This arrangement re
quires special apparatus, but such double-ratio
sets are commercially available.
If not "ganged" together, the two ratios are
changed separately but always by the same
amount. Actually, however, it is not necessary to
be able to adjust the auxiliary ratio, a/6, in known
steps, and any convenient adjustable ratio, such
as a slide wire, may be used. In this case the
balance of the double bridge is bv successive ap
proximation. It is first balanced by adjusting the
main ratio, A/B, after which the circuit is opened
at the link, L, and a balance now obtained by
varying the auxiliary ratio a/6. The link circuit
is again closed and a new balance obtained with
the main ratio. The procedure is repeated until
the same balance is obtained with or without the
link circuit's being open. Ordinarily the final
balance is obtained after only a few sets of ad
justments, and accurate results may be so at
tained. Since only one ratio need be known any
apparatus having such a ratio, as for example a
simple Wheatstone bridge, may be used in con
junction with a slide wire and a fixed standard
resistor to make good measurements by the
double-bridge method.
In work of the highest accuracy, account must
be taken of the resistance of the leads that are
used to connect the ratio arms to the unknown
and standard resistors. There is also lead re
sistance in these resistors, between the branch
points and the terminal binding posts, which is
not necessarily negligible. One way of taking
these lead resistances into account is to use such
high resistances in the ratio arms that the con
necting leads have a negligible effect. Usually,
however, this reduces the sensitivity of balance of
the bridge. A better arrangement is to make the
lead resistances adjustable and select their ratio
in such a way as to balance out their effects. A
convenient method of adjusting the lead re
sistances is as follows: Referring to figure 17, the

Fl(H'RE 17.

Kelvin double-bridge.

ratio A and B are such that their ratio is known

only for the resistance between the two binding
posts s and s'. The lead resistors LA and LB and
the internal leads in X and S add to the resistance
of A and B. This ordinarily prevents the ratio
of the bridge arms from being the same as the
known ratio of A to B. In other words
(12)
b+lTb'
except under the condition that LA and LB are
respectively negligible as compared with A and B,
or under the condition that the ratio of the leads
is the same as the ratio of A to B. That is
A+LA_A
~B~+LB~B'
LA_A
LB~B'

(13)
(14)

If the lead resistances are small as compared with

A and B, the second equation need be only ap
proximately satisfied.
The same considerations apply to the lead re
sistance La and Lb in the auxiliary ratio arms.
Their effect may be made negligible by making
their ratio the same as the ratio a/6.
The following procedure for balancing the bridge
and eliminating the effects of the leads by adjust
ing their ratio may be followed. Arrangement is
made for readily opening and closing the circuit
at L and for short-circuiting the main ratio arms
by connecting together s and s'. Shorting s to s'
is the equivalent of reducing A and B both to zero,
leaving only the lead resistances in the arms. If
the ratios are ganged and nominally equal, the
bridge is balanced in the following steps:
1. With the circuit as shown, the ratios A/B
and a/6 are adjusted until the galvanometer
indicates a balance. This gives
A+LA

X
.
.
g approximately.

(15)

2. The main ratio A/B is now short-circuited

169-18

and the bridge is again balanced, by changing the

ratio of the lead resistances. This may be done
by adjusting the length of one lead wire, or a small
rheostat may be used in series with one of the
leads. This step makes the ratio of the lead
resistances the same as X/S and hence to a first
approximation also to A/B.
3. The short is removed from the main ratio
and the link opened. The bridge then becomes
the simple Wheatstone bridge shown in figure 18.
To a first approximation
X/S=A/B=a/b=LA/LB,

(16)

and the simple bridge is unbalanced only if the

resistances of the leads La and Lh are not in this
same ratio. They are made in this ratio by ad
justing La or Lb until the simple bridge shows a
balance.
4. The link, L, is restored, giving again the
double-bridge, and a second balance is obtained
by varying the ratios A/B and a/b. This balance
is more nearly correct than the first, as the leads
have been approximately adjusted to their proper
ratio.
5. The entire procedure is repeated until no
further change is required in the settings of the
main ratio, under which condition
X/S=A/B=a/b,

(17)

or
X=SA/B=Sa/b.

(18)

Figure 18.

Equivalent Wheatstone bridge.

In case a/b is obtained from a slide wire, or a

ratio that is adjustable but not known, the same
procedure is followed as that outlined above. In
this case, however, the balance for step 3 is obtained
by varying the ratio alb, if necessary making a
fine adjustment of the ratio by varying the lead
resistances La and L.

IV. Special Apparatus for Precision Measurements

1. Direct-Reading Ratio Set
The problem of calibrating precision resistance
apparatus usually involves the comparison of
resistors in the instrument with standard resistors
of the same nominal value. This is most readily
done by some substitution method, which is
usually a method for determining differences
between the resistances of the unknowns and
of the standards. For well-adjusted instruments,
the differences are small and need to be determined
only approximately. For example, if the difference between the standard and unknown is 0.01
percent, the difference need be determined to
onl3r 1 percent to give the value of the unknown
to one part in a million in terms of the standard.
One of the most convenient instruments for

Fici'RE 19.

the measurement of such differences in ratio is

the "direct-reading ratio set." With this comparatively inexpensive instrument and a group
of standard resistors, it is possible to calibrate
accurately most types of resistance apparatus
such as Wheatstone bridges, potentiometers,
resistance boxes, etc.
The direct-reading ratio set is merely an adjustable resistance ratio with which bridges may be
assembled, the remainder of the bridges being
ordinary laboratory equipment. In its simplest
form the ratio set is as shown in figure 19. The
resistor B is a 100-ohm coil, which constitutes
the fixed arm of the ratio. The adjustable arm
consists of a fixed resistor of 99.445 ohms and
three dials, Du D,, and Z?a. The dial Dx consists of
ten steps of 0.1 ohm each, and D2 and D3 have

Direct- reading ratio set.

170-19

ton stops oaoh of 0.01 and 0.001 ohm respectively.

When at their center positions, the resistance of
the three dials is 0.555 ohm, and with the fixed
coil of 99.445 ohms the total of the arm is 100
ohms. Starting from these center positions, a
step on Di will either raise or lower the ^1-arm by
0.1 ohm. The ratio of A to B is then changed
from 100/100 to 100.1/100 or to 99.9/100, that
is to say, it is raised or lowered by 0.1 percent.
Likewise, steps on D-> and D3 change the ratio
by steps of 0.01 and 0.001 percent respectively.
By interpolation of the stops on the lowest dial,
changes of 0.0001 percent may be determined.
Although the ratio sot described above is
correct in theory, variations in the resistances
of the contacts of the dial switches would make
the readings uncertain by several steps on the
lowest dial when ordinary dial switches are
used. The use of mercury switches will reduce
the variations to a few microhms, but such switches
are somewhat difficult to operate and keep in
condition.
To avoid difficulties from variations in switchcontact resistances the design is usually modified
so as to reduce their effect. This is done by plac
ing the switches in high-resistance shunt circuits
that require comparatively large changes in resist
ance to obtain small changes in the parallel re
sistance, as was done in the rheostat of the Mueller
bridge (see section III, 5). The shunt circuits
may be made sufficiently high that switch contact
variations are negligible even when switches of
moderate quality are used. An example of a
shunted dial for obtaining steps of 0.1 ohm is
shown in figure 20. Witli the dial set at 0 the
resistance of the shunt arm totals 133.636 ohms,
and this in parallel with the 30-ohm branch gives
a total resistance of 24.5 ohms. When the dial is
moved to stud 1, the resistance of the parallel
combination increases by 0.1 ohm to 24.6 ohms,
which requires that the shunt arm be 136.666
ohms. That is to say, between studs 0 and 1
there is a resistance of 3.030 ohms. Likewise

Figure 20. Shunl-typc decade.

between studs 1 and 2 there is a resistance of

3.145 ohms so that when the dial is at 2, the shunt
arm totals 139.811 ohms and the parallel combina
tion is 24.7 ohms. The successive steps of the
dial are not of equal magnitude, but the parallel
resistance may be changed by 0.1-ohm stops
from 24.5 to 25.5 ohms. For this the steps on
the dial average about 3.6 ohms each.
Since a change in the high-resistance arm of
about 3.6 ohms is required to change the parallel
resistance by 0.1 ohm, it is obvious that switch
contact variations in the high resistance arm will
have their effect reduced by a factor of about 36 to
1. Their effect could be still further reduced by
increasing the resistance of the shunt, the effect
being reduced as the square of the ratio of the
current in the high resistance branch to the total
current.
In the same way, it is possible to make a decade
for changing in 0.01-ohm or 0.001-ohm steps by
means of a 150-ohm shunt in parallel with a
30-ohm resistor, the 1 50-ohm shunt to be changed
in steps that average about 0.36 or 0.036 ohm re
spectively. Throe decades with appropriate shunts
and a 25-ohm series resistor could be used to ob
tain the equivalent of the A-arai of figure 19, but
with the ratio not appreciably affected by normal
variations in switch contact variations. An
equivalent instrument is available commercially
and, together with a group of standard resistors,
is one of the most useful pieces of apparatus for
the measurement of electrical resistance, especially
for the calibration of resistance apparatus. Someof the procedures will now be described.
Comparison of two-terminal resistance standards,
substitution methods. In nearly all measurements
whore the highest possible accuracy is desired, a
substitution method is used. That is, the change
required to restore balance after replacing a stand
ard with an unknown is measured. In comparing
two-terminal standard resistors by substitution,
the direct-reading ratio set is very rapid and con
venient, and accurate results may be obtained.
For this comparison a Whentstone bridge is set
up as shown in figure 21. A and B are the two
arms of the ratio set, and Y is an auxiliary resis
tance of the same nominal value as the standard
resistoi's under comparison. The two resistors are
in turn placed in the mercury cups Q, and the
bridge balanced by varying the ratio A\B. The
difference in the ratio for the two balances gives
the percentage difference between the two stand
ard resistors. Tims, if the difference in the ratio
for the two balances is one step on the 0.001 dial,
one resistor is 0.001 percent higher than the other.
If we are comparing 1,000-ohm coils, for example,
the difference is 0.001 percent of 1,000 ohms, i. e.,
0.01 ohm. Which coil is the larger is determined
by observing whether the ratio is increased or
decreased when the standard resistor is replaced
by the unknown resistance. Actually the dif

171-20

Ficvbe 21.

Whealstone bridge for substitution method.

ferencc is determined in percentage of the arm Y,

but for small differences any departure of Y from
its nominal value is of no consequence.
Although the dials of a direct-reading ratio set
change the ratios in decimal parts of a percent, it
is desirable to mark the dials in parts per million.
Thus the steps on the 0.001-percent dial would be
marked, 0, 10, 20, etc., and the 0.01-percent dial
would be 0. 100, 200, etc. Readings of the instru
ment and data taken from it are then whole num
bers of moderate sizes rather than decimals. In
the discussion that follows, it is assumed that the
ratio set is marked in parts per million, and that
corrections to resistors are also expressed in parts
per million. Bridge readings and corrections to
standards are then readily added and subtracted
and calculations are greatly simplified.
In comparing nearly equal resistances by sub
stitution, the lack of correct adjustment of the
coils of a well-made ratio set will cause errors of
much less than 1 part in a million. The variation
in the contact resistance of the dial switches may
be appreciable unless they are kept clean. The
switches should be frequently cleaned and lubri
cated with a little good-quality light oil. When
kept in good condition, the resistance of the three
switch contacts should not vary as much as a
thousandth of an ohm. From the values of the
coils on the 0.001 percent dial, we see that a vari
ation of a thousandth of an ohm will cause vari
ations in the ratio of 1/3.6 of a part in a million.
In comparing resistances by the substitution
method, the two balances of the ratio set should
be made quickly so that temperature changes will
not cause variations in the ratio set between read
ings. It is best to have all coils of the ratio set
made from the same lot of resistance material.
If this is done and manganin coils are used, a

change in temperature of 1 deg C should not

change the ratio more than 2 or 3 parts in a million.
In spite of these sources of error, it is probably
possible to compare standard resistors ranging
from about 10 ohms to 1,000 ohms to within
1 or 2 parts in a million by the use of this ratio
set. This, of course, requires that the set be well
constructed, that a fairly sensitive galvanometer
be used and that the resistances differ by not more
than about a tenth of a percent. With resistances
less than 10 ohms, the resistance of the mercury
cup contacts may cause trouble. For this reason
standard resistors smaller than 10 ohms are now
almost always four-terminal resistors. Methods
of comparing such resistors will be discussed later.
Another method for comparing nominally equal
two-terminal standard resistors, which is practi
cally equivalent to the preceding substitution
method, is what might be called a double substi
tution method. In this latter arrangement, the
two resistors under comparison are used to form
two arms of a Wheatstone bridge, the ratio set
forming the other two arms. The resistors are
both mounted in mercury cups so that they may
be interchanged without affecting lead resistances
in series with them. After a balance is obtained
by adjusting the ratio set, the resistors are inter
changed, and a new balance is obtained. The
{>ercentage difference between the two resistors is
lalf the difference between the two readings of
the ratio set. This method is used for the same
range of resistances as those measured by the
simple substitution method, and about the same
accuracy may be attained.
Comparison oj Jour-terminal resistances with ttooterminal resistance.The direct-reading ratio set is
convenient for the comparison of two-terminal with
four-terminal resistances. Although the occasion
seldom arises for the comparison of two-terminal
with four-terminal standard resistors, it is often
necessary to compare a standard resistor of one
type with a resistance coil of the other type. Thus,
in measuring the coils of many pieces of electrical
apparatus it is impossible to make connection with
the coils except through comparatively large con
necting resistances. However, it is generally pos
sible to make potential connections to the two
ends of the coils and measure them as four-terminal
conductors. In doing so, it is often convenient to
compare them with two-terminal standard re
sistors.
In figure 22, X i? a four-terminal resistor to be
measured, having the current terminals T\ and T3,
and potential terminals Pi and P2. The potential
leads may contain considerable resistance in addi
tion to that of the leads. A and B are the two
arms of the direct-reading ratio set, and Mi and
M2 are mercury cups into which either a twoterminal standard resistor or a short circuiting
link may be placed.
Suppose we start with the galvanometer con

172-21

nected at Pu a standard resistor nominally equal

to X inserted in Mu and with M2 shorted. We
then have a simple Wheatstone bridge, which is
balanced by varying the setting of the ratio set.
After this balance is obtained, the galvanometer
connection is shifted to P2, and the standard re
sistor is placed in M2, M, being shorted with the
link. We again have a Wheatstone bridge but
with the standard resistor and unknown inter
changed. This interchange has been obtained
without making any change in the resistance of the
leads of the measuring circuit, except for possible
variations in the resistances of the mercury cup
contacts, which will be small if the mercury con
tacts are clean. The bridge is now again balanced
by means of the ratio set. The percentage differ
ence between the unknown and the standard re
sistor is half the difference in reading of the ratio
set for the two balances. Unless the resistances
under comparison are fairly large, it will be neces
sary to take into account the resistance of the
short-circuiting link. This is done by subtracting
the link resistance from the resistance of the
standard resistor and considering that the resistor
has this new value and is being interchanged with
a link of zero resistance.
The resistance of the link can be measured as
follows: Connect the link between two 1-ohm
resistors to form two arms of a Wheatstone bridge,
using the direct-reading ratio set for the other
two arms, as shown in figure 23. L is the link,
and A and B are the arms of the ratio set. Two
balance readings are taken, first with the gal
vanometer connected at one end of L and then
at the other. Half the difference in the readings
is the value of the link resistance in percentage
of the 1-ohm arms. If the link resistance is large,
it may be necessary to use larger resistances in
place of the 1-ohm coils. This method is very
convenient for the measurement of small resist
ances such as links, connecting wires, switch
contact resistances, etc. It is not a precision
method but usually is sufficiently accurate for the
measurement of resistances such as those just
mentioned, which are to be used in series with
larger resistances.
Substitution method for decades.In the calibra
tion of precision rheostats the occasion often
arises for the measurement of a series of resistors
of the same nominal value. This is readily done
by the substitution method, using a standard
resistor of the same value as the steps of the
rheostat, reading differences on a direct-reading
ratio set. The procedure is illustrated in figure
24. In this figure PR is the precision rheostat
to be calibrated, and let us assume that the 10ohm-per-step dial is to be checked. PR is then
a plug box or any decade with 10-ohm steps, and
M is a pair of mercury cups in which is placed a
standard 10-ohm resistor. The arm Y is a 100ohm resistor whose value need not be accurately

173-22

VWWWWWV

Figure 22.

Measurement of 4-terminal resistor.

L

FmrRE 23.

Measurement of lead resistance.

FinuRB 24.

Test of decades by substitution.

known, and A and B are the two arms of a direct

reading ratio set. The reading of the rheostat
under calibration is set at zero, the plug box is
set at 90 and the 10-ohm resistor is placed in
M, making the total nominal resistance of the
arm 100 ohms.
After the bridge has been balanced by changing
the setting of A, the standard resistor is removed
and the mercury cups shorted with an amal
gamated copper link, and also the reading of
PR is changed from 0 to 10 ohms. This has in
effect substituted a 10-ohm step of PR for the
10-ohm standard resistor. The change in the
reading of A required to again obtain a balance
of the bridge is a measure of the difference between
the step on the rheostat and the standard resistor,
such difference being a percentage of 100 olims.
The second 10-olun step on the rheostat is obtain
ed by leaving PR at its 10-olun reading, again
placing the standard resistor in M and reducing
the PB resistance to 80 ohms. These three
resistors still total 100 ohms, nominally. The
standard is now again replaced by a short-circuit
ing link, and PR is set to read 20 ohms, the result
ing change being read from A. The change was
produced by the substitution of the second 10-ohm
step of the precision rheostat for the 10-ohm
standard.
The procedure is continued, the steps of PR
being successively substituted for the standard
resistor. It should be noted that the steps of
the auxiliary plug box, PB, are also being replaced
by the standard resistor as PR is being increased
and PB decreased in reading, the standard being
cut in and out of the circuit. Data are tberefore
obtained for calibration of both PR and PB in
terms of the standard resistor. Hence by this
method, two precision rheostats may be calibrated
simultaneously.
When 10-ohm steps are calibrated in a 100-ohm
arm, the accuracy is reduced by one order. That
is to say, differences in readings of the ratio set
must be obtained to 0.0001 percent if the 10-ohm
steps arc to be, determined to 0.001 percent. The
method has the advantage that changes in the
over-all resistance of the rheostat are determined
under the conditions of use, which is often not the

Figure 25. Dial connections.

Ficure 26.

Lead resistances in Wheatstone bridge.

case when the individual resistors are measured

directly. This may be seen by reference to figure
25, which represents the connections to one dial
of the rheostat. The switch contact may be set
on the contact studs marked 0, 1,2, 3, etc., thus
connecting 0, 1, 2 or more resistors between the
dial terminals S and T. It should be noted that
the resistors are usually connected in series and
lead wires are connected from the junctions of the
resistors to the switch contact studs. If the resis
tors are calibrated by measuring between studs 0
and 1 ,1 and 2, 2 and 3, etc., two of these leads are in
cluded with each resistor. In actual use all the
leads to the coils in use are not in the circuit, but
only one lead is used for any setting of the dial.
The above method of calibration determines the
step as in actual use, the only additional data
required being the resistance between terminals
with the dial set at zero. This is readily measured
by the method just described for the determination
of link resistances. For any setting, this "zero
resistance" must be added to the sum of the steps
as determined by substitutions.
Comparison of four-terminal resistance stand
ards.Before taking up the question of the
comparison of four-terminal standard resistors,
let us consider briefly the effect of the lead wires
of a simple Wheatstone bridge. In figure 26, A'
and Y are nominally equal resistances, and A and
B are the two arms of a direct-reading ratio set.
The conductors x, y, x', and y' are used to connect
up the bridge, and we will also denote their resist
ances by x, y, x' , and y' , respectively. The ratio
of the resistances of the two arms containing A'
and Y is not in general the same as the ratio of
A' to Y because of the resistance of these con
necting leads. If we could select leads such that
the ratios xjy and x'/y' were the same as A"/F, the
ratio of the two arms would be independent of the
actual values of the lead resistances. That is, the
balance would be the same as if the resistances of

174-23

the leads were negligibly small. It is possible to

make the leads adjustable and make the ratios
x/y and x'jy' the same as X/Y. In fact, such an
arrangement is used at the National Bureau of
Standards to reduce the effect of the leads when
using the Kelvin double bridge. Instead of adjust
ing the leads, it is possible to balance the bridge
with fixed leads, and then find what this balance
would have been with the proper ratio of lead re
sistances, or with negligibly small lead resistances.
X and Y were assumed to have practically equal
resistances. Instead of connecting the galvanom
eter as shown, suppose we balance the bridge with
the upper galvanometer connection first at a and
then at b. The average of the two readings is
the value that would have been obtained had the
conductors x and y been equal or negligibly small
in resistance. Thus, by taking two readings we
can take into account the effect of these two
connecting resistances. This average balance
reading is not the correct reading, however, unless
also x'=y', which is probably not true. We must
now find what this balance would have been with
x'=y'. To do this suppose we shift the current
connections from the points shown to the points
a' and b'. In doing so we remove the lead resis
tances x' and y' from the X and Y arms and
connect them in series with the ratio arms A and B.
If x' and y' are not equal, we will change the
balance of the bridge by removing them from the
arms X and Y. We will have still further changed
this balance by adding them to the ratio arms A
and B. Which of these changes is the greater
depends upon the relative sizes of the two pairs
of arms. If A and B are equal to X and Y, the
changes will be equal. That is, with all arms
nominally equal the balance would be changed a
certain amount if the battery connections were
shifted to a' and b' . This change is twice that
which would have been obtained had we only
removed x' and y' from the X and Y arms. The
average of the readings before and after changing
the battery connections is then the reading that
would have been obtained with x'=y' or both
negligibly small.
Suppose, however, that A and B are ten times
as large as X and Y. Then connecting %' and y'
in series with A and B produces only a tenth as
large a change as is caused by their removal from
X and Y. Then ten-elevenths of the change in
balance when the battery connections are shifted
is due to the removal of x' and y' from the X and Y
arms, and the remaining one-eleventh is due to
the connection of the leads in the ratio arms.
This enables us to calculate what the balance
would be with x' and y' equal or negligibly small.
As an example, suppose X and Y are each 10
ohms, and the arms of the direct-reading ratio set
are 100 ohms. Assume that with the battery
connected as shown in figure 26 and the galva
nometer at o, the reading of the ratio set with the

bridge balanced is 5497 millionth^, and with the

galvanometer connection changed, to b, the reading
is 5613. Let us further assume' that when we
change the battery leads to a' b', leaving the
galvanometer connected at b, the balance reading
changes to 5835. Then with leads of the proper
ratio, or leads with negligibly small resistances,
the balance would have been at (5,613+ 5,497)/2+
10/11(5,835-5,613); i. e., at 5,757.
As a matter of fact, this scheme for taking intp
account the connecting leads is practically never
used in comparing two-terminal resistors. As the
substitution method requires no consideration of
the lead resistances, except to see that they are
reasonably small, it is generally used. However,
in comparing four-terminal resistances with a
simple Wheatstone bridge we follow exactly the
steps outlined above. Figure 27 shows a bridge
set up for this purpose. A and B are the two
arms of the direct-reading ratio set. X and Y are
the four-terminal resistors under comparison,
with current terminals Tu T2, 77, and 77, and
potential terminals Pu P2, TV, and P2 . The
bridge is balanced when connected as shown, and
a second balance is obtained after shifting the
galvanometer connection to P2'. The third bal
ance is obtained after now changing the current
connections from Tt and 77 to Pi and P,', and
calculations are made as above. The two arms
X and Y are now interchanged and the three
readings again obtained. From these two sets of
readings, we get two balance points on the ratio
set. Half the difference between these two balance
readings is the percentage difference between the
two four-terminal resistors.
Although the direct-reading ratio set was
developed for use in the comparison of nominally
equal resistances, it can be readily adapted for the
comparison of resistances of any ratio provided
some independent means is available for accurately

175-24

Figure 27.

Compariton of 4-ttTminal resistors.

realizing the same resistance ratio. As an example

of such a use let us consider the calibration of a
25-ohm resistor by comparison with a 100-ohm
standard.
As has been described above, the direct-reading
ratio set consists of the arm A, which is adjustable
in small steps from values slightly below to slightly
above 100 ohms. The B-arm is ordinarily a fixed
100-ohm arm. Suppose we change the i?-arm
by connecting an additional 300 ohms in series.
The ratio A/B then becomes 100:400, which is
adjustable in the same percentage steps as was
the 100:100 ratio. Using two sets of mercury
cups, let us set up the bridge shown in figure 28.
The 25-ohm resistor is placed at X and the 100ohm standard at S, and the bridge is balanced
by adjusting A, and let us call this reading A.
Also let A0 be the reading A would have had if the
X and S arms had been exactly in the ratio 25 :100.
If cz and c, designate the proportional corrections
of X and S, the actual balance A, of the bridge,
will be
A=A0+cI-c,.
(19)
Since A is obtained experimentally and c, is
known, this equation could be solved for cx if A0
were also known. To obtain A0, the arms X and
S must be replaced by a resistance ratio of exactly
1:4.
To realize a 1:4 ratio it is necessary merely to
have five resistors that are reasonably nearly equal.
If one of these five is connected in place o? A' and
the other four in series in place of S, the ratio will
be only approximately 1:4. However, if the five
resistors are placed one after the other in X, the
remaining four each time being connected in S,
the average of the five ratios will be 1 : 4 to a very
high accuracy. In other words, the average of the
five readings of A with the five resistors in turn
at X, the remaining four in series in the S-arm,
will be A0, the reading of A for an exact 1 :4 ratio.
This value of Ao can be substituted in eq 19 to
allow the calculation of cz, the proportional cor
rection to the unknown 25-ohm resistor.
Another method of obtaining A0 would be to
balance the ratio set with one resistor at X and
the other four at S, thus obtaining A of eq 19.
The five resistors could then be substituted in any
bridge and their differences determined. The
terms cz and c, are then the amounts in propor
tional parts that the resistor in X and those in S
differ respectively from the average value of the
five.
To get an accurate value of the 1:4 ratio, the
five resistors should have large enough resistances
that the lead resistances of the X and S arms are
negligible, or the lead resistances should be in the
ratio 1:4. They may usually be so set with a
sufficient accuracy by shorting X and S and adj usting the bridge circuit so obtained to a balance by
changing the length of the lead wires. The five

Figurk 28.

Bridge with 1:4 ratio.

resistors do not need to be equal to a very high

precision. However, they must not differ by more
than 0.1 percent if the average of the five ratios is
to be correct to one part in a million.
A procedure analogous to this method of obtain
ing a ratio of 1 :4 may be followed to determine any
ratio
where n is an integer, by taking the
average of n+1 ratios. It may also be used to
determine any ratio r:n, where both r and n are
integers. Starting with r+n equal resistors, r in
one arm and n in the other, the resistors are
rotated in a cyclic order until each resistor occupies
each position once. The average of the r+n
readings is the reading that would be obtained if
all resistors were equal. Moreover, the resistors
in either arm may be in parallel rather than in
series.
In the calibration of high resistances, use may
be made of the fact that the proportional correc
tion to a group of nominally equal resistors is the
same when they are connected in parallel as when
connected in series. For example the ten 100,000ohm sections of a megohm box may be connected
in parallel and measured against a 10,000-ohm
standard resistor. If the parallel group is high
in resistance by 0.01 percent, the series resistance
of 1 megohm will also be high by 0.01 percent.
Here it is assumed that the ten sections are suffi
ciently near to equality that in the expansion
T^r = l-c + c2-c3, etc.

(20)

the second and higher powers of c are negligible,

where c is the proportional amount by which the
resistance of any section differs from the average
of all. If this condition i3 satisfied, the ratio of

176-25

the resistance of n resistors in parallel to their

resistance in series is exactly 1 :n2.
The methods that have just been presented
assume the use of a direct-reading ratio set. These
methods, however, are entirely satisfactory when
use is made of any ratio set that has small and
definite steps. Such a set may be assembled from
ordinary laboratory apparatus. For example, the
adjustable 100-ohm arm might be made of a
105-ohm coil with a parallel decade box reading
about 2,500 ohms. A change in reading of the
decade box by 0.1 ohm would change the 100-ohm
arm by about 1 ppm. The changes in the parallel
resistance are not directly proportional to the
change in the high resistance arm, but they may
be readily calculated. This adjustable 100-ohm
arm, together with a fixed 100-ohm resistor,
constitutes an adjustable ratio set.
2. Universal Ratio Set
Precision standard resistors are usually made
only in integral multiples or submultiples of an
ohm. Consequently odd-valued resistors usually
cannot be measured by a substitution method,
except in a few cases where standard resistors can
be combined to give a resistance nearly that of the
unknown. The comparison of odd-valued resis
tors with standards is then not possible with a
direct-reading ratio set, but a ratio set is required
that is accurately adjustable over a wide range,
at least from a 1:1 to a 5:1, or preferably to a
10:1 ratio.
A very convenient wide-range ratio set is one
used at the National Bureau of Standards and
called a "universal ratio set" [14]. This instru
ment is one having a constant resistance, between
two external terminals, of about 2,111 ohms. An
arrangement of dials is such that in effect a poten
tial connection may be made at any point of the
2,111 ohms to the nearest 0.01 ohm. The ratio
of the resistance between the potential point and
one terminal of the set to the resistance between
the potential point and the other terminal is
therefore adjustable in small steps over a very
large range. The device is the equivalent of a
long slide wire with a movable contact, and its
uses are analagous.
Suppose it were desired to measure a resistance
of say 6.8 ohms by comparing it with a 10-ohm
standard resistor. The two resistors could be
connected in series and the combination connected
across a slide wire as shown in figure 29, X being
the unknown, S the standard, and W a slide wire.
If a galvanometer, 0, is connected to terminal a
of the unknown and to the slide wire, a Wheatstone bridge is obtained that will balance with
the slide wire at some point, 1, near the end. If
now the galvanometer connection is changed suc
cessively to 6, c, and d, successive balances will be
obtained with the slide wire at 2, 3, and 4, respec
tively. The ratio X/S is then the same as

Figure 29.

Z
3
Comparison of resistors with slide wire.

Hij/Rt.t where i?i,2 and iZj,4 are the resistances of

the slide wire between the points 1 and 2, and
3 and 4, respectively.
It is very difficult to make a wire of sufficient
length and uniformity that the ratio RuiiRi,* can
be accurately determined. Instead of a slide
wire, the universal ratio set makes use of a group
of wire wound resistors so that the resistances
between the balance points 1, 2, 3, and 4 can be
accurately known. The arrangement of its dials
is as follows. The highest dial consists of twenty
100-ohm resistors in series, with the dial contact
acting as the potential connection to the instru
ment, as shown in figure 30. As the dial is
rotated in a clockwise direction, the 100-ohm
resistors are successively changed from the right
to the left side of the contact. To change resist
ance from the right to left side in 10-ohm steps,
two more dials are used, each having ten 10-ohm
steps as seen in figure 31. These two dials are
operated by the same handle but with one dial
increasing as the other decreases in resistance.
The total resistance between S and T remains
constant for any setting of the 10-ohm decades,

177-26

Figure 30.

100-ohm dial of universal ratio set.

IOOASTEPS

IOASTEPS

Figure 31.

100- and 10-ohm dials of universal ratio set.

and of course it is unaffected by the position of

the 100-ohm dial switch that is merely a potential
contact.
Steps of 1, 0.1, and 0.01 ohm-per-step are
obtained in the same way as the 10-ohm steps,
that is with two decades of each denomination
operating together, one on each side of the 100-ohm
dial. This scheme for the change of the potential
point along a fixed resistance is essentially the
same as that of the Feussner type potentiometer.
The large number of contacts in series limits its
use for precision work to circuits having a rather
high resistance.
The universal ratio set is used only for the
determination of resistance ratios and hence may
be calibrated in terms of any unit. It is most
conveniently calibrated in terms of an average
step on the highest or lowest dial. This is done
by comparing each step of a decade with the ten
steps of the preceding dial starting with the
smallest dials. Only the steps on one side of the
100-ohm dial need be tested, as the function of
the other group is to keep the total resistance
constant. A check of the constancy of the over-all
resistance needs to be made for all readings of the
double dials.
The use of a universal ratio set is the same as
that of a slide wire as described in conjunction
with figure 29, for the measurement of odd-sized
resistances. It may be used in the measurement
of four-terminal resistors and is especially con
venient for the tests of potentiometers. The test
of a potentiometer consists in the measurement of
the ratio of the emf-dial resistance to the stand
ard-cell resistance for all settings of the emf and
standard-cell dials. These resistors are of the
four-terminal type with potential connections
brought out to emf and standard-cell binding
posts. The method is satisfactory even when

some resistance is common to both the standardcell and main-dial resistance. The arrangement
for such a test is shown in figure 32. The ratio
set, URS, which is shown as a slide wire, is con
nected in parallel with the potentiometer, the
connection to the latter being made to the battery
binding posts, BA-\- and BA-. Readings on
URS are made with the galvanometer coimected
successively to the SC and emf binding posts, for
all settings of the emf dials. It should be noted
that changes in the emf dials are merely changes
in potential points and do not affect the readings
obtained for the SC dial.
Potentiometers are provided with rheostats in
the battery circuit for adjusting the potentiometer
current. A change in this battery rheostat will
change the differences in readings obtained on the
universal ratio set but not the ratio of the differ
ences. It is possible to make the ratio set directreading by adjusting the battery rheostat until the
difference on the ratio set for the standard-cell
posts is a decimal multiple, preferably 1,000, of
the reading of the standard-cell dial. Corrections
for the emf dials may then be. read directly from
differences across the emf terminals.
POTENTIOMETER

Figure 32.

178-27

URS
Check of potentioni&er with universal ratio set.

V. Calibration of Precision Bridges

1. Calibration of Wheatstone Bridges
The circuit of a commercial precision Wheatstone bridge is usually as shown in figure 33. The
two ratio arms A and B may have any one of
several values, the choice being made by inserting
a plug in series with the desired resistor. The
rheostat arm, R, consists of four to six decades of
not less than 0.1 ohm nor more than 10,000-ohm
steps. The unknown resistor is connected to
heavy binding posts, X, and a battery and gal
vanometer arc connected to the external binding
posts BA and OA. The switches SB and Sa open
and close the battery and galvanometer circuits.
In calibrating a bridge of this type, it is neces
sary to determine the resistance of the ratio arms
between the branch points 2 and 3 or 3 and 4
rather than between the external binding posts,
as the resistances from the branch points to the
binding posts arc usually not negligible. Also it
is necessary to find the resistance between 1 and
4 with the rheostat dial, R, set at zero, as well as
the corrections to the readings of the rheostat
dials themselves. It is also necessary to measure
the lead resistances between the branch points 1
and 2, as these are in series w itli the unknown
resistance connected at X. The bridge balance
determines the entire resistance of the AT-arm,
and the leads must be subtracted in order to
obtain X itself. These four types of resistance
measurements are made by application of some of
the general principles previously discussed.
To measure the ratio arm B, for example, it
is necessary to determine the four-terminal
resistor having branch points 3 and 4. This is
easily done by comparing it with a two-terminal
standard of the same nominal value, making
application of the method outlined in section
IV, 1 and using the circuit shown in figure 34.
The resistors a and b are the arms of a directreading ratio set, or any two ratio arms that may
be adjusted in small known steps. As shown,
one side of the galvanometer is in effect connected
to the branch point 4, through the rheostat arm,
R, which may be set at zero. The standard
resistor is placed in the mercury cups, Mu and a
short-circuiting link is placed across the other
mercury cups, il/2. After a balance is obtained
by adjusting arm a, the standard and shorting
link are interchanged and the galvanometer
connection is shifted in effect to point 3 by con
necting to the other X binding post. Half the
change in the ratio that results gives the difference
between the #-arm and the standard resistor.
The resistance of the A ratio arm may be
measured in exactly the same way as for the
B arm by making connection to the other OA
terminal instead of the one shown. Some diffi
culty may arise from variations of the contact

Figure 34.

Check of ratio arm of Wheatstone bridge.

resistance of the galvanometer key, which will

now be in the measuring circuit. This may be
avoided by interchanging the adjacent X terminal
and the GA terminal. The galvanometer would
then be shifted from one X oinding post to the
GA binding post instead of from one X binding

179-28

post to the other. That is to say, the current and

potential connections at point 2 may be inter
changed. This will place the variable galvanometer
switch resistance in series with the galvanometer
where it will not affect the bridge balance.
The rheostat arm, R, may be calibrated like
any other precision rheostat as described in
section IV, 1.
Connection should be made
through the appropriate A and GA terminals.
The resistance of the rheostat arm with all
dials set at zero may be measured by connecting
the arm between two equal resistors, using a
direct-reading ratio set to connect up a Wheatstone bridge. The change in the reading of the
ratio set is determined when the galvanometer
connection is changed from one branch point of
the rheostat arm to the other. This is the same
as. the method given in section IV, 1 for the
measurement of lead resistances. In a similar
way, the lead resistances in the X arm may be
determined, taking readings with the galvanom
eter connection to branch points 1 and 2 and to
the external A' binding posts, with the latter
connected by a shorting wire. The calculation
must be made in two parts in order to exclude the
resistance of the shorting wire.
For calculating X from the calibration data, it
is convenient to express the corrections to the
readings of the rheostat arm in ohms and to ex
press the corrections to the ratio- arms in pro
portional parts. The calculation in proportional
parts involves merely a division of the correction
in ohms by the nominal value in ohms. The
value of A' for a given balance is calculated from
the equation
X=*(\+a-b){R+r+r0)-X0-

(21)

In this equation a and b are the proportional cor

rections to the ratio arms A and J) respectively,
R+r is the sum of the dial readings and correc
tions, and r0 is the resistance of the rheostat arm
with all dials at zero. The term A0 is the resist
ance in ohms of the lead wires in the A-arm of
the bridge. The factor (1-fa b) is an approxi
mation for (1 +o)/(l +b) and is accurate if a and 6
are small as compared with unity. If neither a
nor b exceeds 0.001, the error from neglecting the
second-order terms docs not exceed two parts in
a million.
2. Calibration of Thermometer Bridges
Because of the space limitations and structural
difficulties, platinum resistance thermometers usu
ally have resistances of less than 100 ohms. The
common values are about 2.5 or 25 ohms, the
actual values being so chosen that the change in
resistance is very close to 0.01 or 0.1 ohm per
degree centigrade change in temperature. In

order to read to 0.001 C, it is necessary to read

these thermometers respectively to 0.00001 or
0.0001 ohm. For such measurements, bridges of
special design are usually used.
In this country thermometer bridges for preci
sion work are usually made with equal ratio arms.
These arms are interchangeable, so that the use
of an average value eliminates errors from lack of
equality of the ratio arms, or the ratio arms may
be interchangeable and adjustable so that they
may be made equal at any time. With such
interchangeable ratio arms, actual calibrations of
them are unnecessary. The calibration of a ther
mometer bridge requires only a calibration of the
rheostat arm.
In order to measure platinum resistance ther
mometers, which as stated above are usually not
very high in resistance, with a bridge having equal
ratio arms, it is necessary to have the rheostat arm
adjustable in small steps. In fact, steps as low as
0.0001 or 0.00001 ohm are needed for work of the
highest precision. In order to obtain such small
steps, contact resistances cannot be used directly
in the rheostat arm. Hence recourse is had to
decades of the Waidner-Wolff type described above
in section III, 5, in which the changes in resistance
result from changes in the values ofhigh-resistance
shunts on comparatively small resistances.
Switches are placed in the nigh-resistance shunts,
where variations in their resistance will have a
negligible effect. Decades of the Waidncr-Wolff
type cannot be set to have zero resistance, but
the small changes start from an appreciable mini
mum value. However, for equal-arm bridges,
compensating resistance may be placed in the
A-arm. The value of the unknown is then meas
ured by the increase in reading of the rheostat arm
when the unknown is connected into the circuit.
The use of a resistance thermometer to measure
temperature involves merely the determination of
resistance ratios. Hence in calibrating thermom
eter bridges it is necessary to determine only rela
tive values of resistance, which does not require
the use of standard resistors. This calibration is
readily made by the user, especially for bridges
having equal ratio arms.
The calibration of such thermometer bridges
requires as auxiliary equipment only an adjustable
resistor that has the same range as that of the
bridge rheostat. This resistor needs to be accu
rately adjustable, although the resistance need
not be known for any position. Such an adjust
able resistor may be assembled, for example, from
a decade box with minimum steps of 0.1 ohm in
series with an 0.1- or 1-ohm resistor, which is in
turn shunted by a slide wire or a rheostat of fairly
high resistance. This adjustable resistor is con
nected across the A-terminals of the bridge and is
used to balance the bridge after the rheostat arm
is set to certain required readings. The shunted
0.1- or 1-ohm should be attached to the A-termi

180-29

rials so as to be adjacent to the rheostat arm. If

a slide wire is used for the shunt, the sliding con
nection should be used as the galvanometer branch
point so that its variable resistance will be in the
galvanometer branch where it will not affect the
balance. This, in effect, throws the shunted re
sistance partly in the X and partly in the rheostat
arm. We then have the equivalent of the bridge
shown in figure 35, the ratio arms A and B being

0.1 A
Figure 35.

Calibration of rheostat arm of thermometer

bridge.

accurately adjusted to equality. For a balance

the rheostat arm, rh, and the X-&rm must be also
equal. If then X is changed by any amount, as
by the insertion of a resistance thermometer in
series, the rheostat would have to be changed by
an equal amount to again balance the bridge.
Thus, actual values of the rheostat arm need not
be known if changes in the arm are accurately
determined.
The procedure for checking the rheostat arm is
usually as follows. The resistances of all steps in
the arm are determined in terms of an average
stop on the smallest dial. The steps of the small
est dial are first intercompared to see that they
are equal within 0.1 step, which is as accurately
as the steps may be readily read by interpolation
from galvanometer deflections. An easy way to
intercompare the steps on the lowest dial is to set
it at 0, the other dials being at any convenient

setting, and balance the bridge by adjusting X.

The galvanometer deflection is now read when the
lowest dial reading is changed from 0 to 1 . Leav
ing this dial on 1, the bridge is again balanced
with the ^-rheostat, and the galvanometer deflec
tion then again read for a shift of the lowest rheo
stat dial from 1 to 2. These alternate balances
with rh and X are continued until the value of
all the steps on the smallest dial of rh are deter
mined in terms of galvanometer deflections. Un
less one of the steps is defective, these should be
the same to the nearest 0.1 step.
The steps on the second dial of rh should each
equal the 10 steps of the lowest dial. They are
measured in terms of the lowest dial by setting
the lowest dial at 10 and the second dial at 0, the
other dials being at any convenient value, usually
small in order to obtain good sensitivity. With
the lowest dial at 10 and the next larger dial at 0,
the bridge is balanced by adjusting the value of X,
after which the smallest dial is turned to 0 and
the other dial shifted from 0 to 1 . If there is now
any galvanometer deflection, it is because the first
step on the second dial is not the same as 1 0 steps
on the smallest. The amount that they differ, in
terms of steps of the lowest dial, is determined by
reading the galvanometer deflection and evaluat
ing this deflection by reading the additional change
in deflection resulting from a change of the smallest
dial setting from 0 to 1 .
To measure the second step on the second dial
its reading is left at 1 and the lowest dial now set
at 10, a halance being obtained with X. The
lowest dial is now set back to 0 and the other
increased to 2, thus substituting the second step
of the higher dial for 10 steps on the lower. The
lack of balance is again translated into fractions
of a step on the lowest dial by interpolation, using
galvanometer deflections. This procedure is con
tinued, each step of the second dial being compared
in turn with the 10 steps of the lowest dial. A
table is now made showing the value, in terms of
steps of the lowest dial, of the first step of the
second dial, the sum of the first two steps, sum of
the first three steps, etc. This table will give the
resistance corresponding to any reading of this dial
in terms of steps on the smallest dial, which we
might call "bridge units".
The values of the steps on the next higher dial
are now determined in the same way but in terms
of the 10 steps on the second dial, the unit again
being a step on the smallest dial. The continua
tion of this process gives the resistance of each step
of each dial in terms of the 10 steps on the preced
ing dial, from which finally is calculated the re
sistance for each setting of each dial in terms of
bridge units.
The resistances as determined may be used
with any resistance thermometer without con
verting their values to ohms. It is sufficient to
standardize the resistance thermometer on this

181-30

same bridge, by measuring its resistance at known

temperatures. However, it is often desirable to
convert the values to ohms in order that the bridge
may be used for the measurement of resistance
other than that of thermometers. This is done by
measuring the resistance of a standard by balanc
ing the bridge with the standard resistor connected
to the ^-terminals. All readings of the rheostat
should then be multiplied by the ratio of the
resistance of the standard in ohms (rather, the
difference between its resistance and that of the
shorting connector) to its resistance in bridge units
in order to convert the rheostat calibration to ohms.
The above procedure may be used for the cali
bration of the rheostat arm of any Wheatstone
bridge, in terms of steps of the lowest decade and
then in ohms, by comparison with a standard
resistor. When used with equal ratio arms, the

"zero" resistance of the rheostat need not be

determined, if two bridge balances are obtained,
the first with the unknown connected to the Xterminals and the second with the unknown re
sistance short-circuited.
Many precision thermometer bridges provide a
shorting plug for short-circuiting the resistor con
nected to the ^-terminals. The galvanometer
connects to the center of the shorting connector
so that equal amounts of its resistance are inserted
in the X and rheostat arms thus giving the same
balance as if the shorting connector had a negligible
resistance. With this method it is unnecessary to
know the resistance of the shorting connector,
although such resistance may usually be esti
mated with sufficient accuracy from its length and
gauge size.

VI. Resistivity of Solid Conductors

1. Resistivity, Definition and Units

2. Measurement of Resistivity

In the experimental work that led to the formu

lation of his law, Ohm found that the resistance,
R, of a conductor is directly proportional to its
length,
and inversely proportional to its crosssectional area, A. These experimental facts may
be written in the form of an equation as

For a uniform conductor, it is merely necessary

to measure the resistance of a known length of the
conductor and then measure its cross-sectional
area in order to determine its resistivity. For con
ductors of small cross section, it is usually possible
to use a sufficient length that the resistance may
be accurately measured with a simple Wheatstone
bridge. For conductors of large cross section, it is
customary to measure the resistance with a Kelvin
double bridge in order to avoid errors from contact
resistances. Wherever possible, the cross-sectional
area is calculated from micrometer measurements.
For conductors of small or irregular cross section,
micrometer measurements are not sufficiently
accurate, and the average area is determined from
mass and density measurements.
For a uniform conductor the cross section, A, is

(22)
where p is a constant of proportionality whose
value depends upon the material of the conductor
and upon the units used in measuring I and A.
This constant of proportionality is called resis
tivity.
The above equation, which defines resistivity
may be written
P=Rj-

(23)

No name has been assigned to the unit of resis

tivity, and consequently the unit is specified by
stating the units used in measuring R, A, and I.
This has resulted in the use of a large number of
units, as each of R, A, and I may be expressed in
more than one unit or subunit. From the above
equation for p, it is seen that the value of p is
numerically equal to that of R for a conductor
having unit length and unit cross-sectional area.
A cube is such a conductor, and this has led to the
rather common expressions for the unit of resis
tivity "ohms per cubic inch" or "microhms per
cubic centimeter". These expressions are unde
sirable, because they imply that resistivity is the
ratio of resistance to volume. It is logically bettor
to say "ohms times square inches per inch",
"microhms times square centimeters per centi
meter" or more briefly "ohm-inches" and "microhm-cen timeters" .

(24)
where V is the volume and / the length of the con
ductor. But since, from the definition of density.
D,
V=

(25)

where m is the mass, the preceding equation may

be written
m
A=
(26)
ID'
For many purposes it is sufficient to assume the
density as that given in tables. If this is not
sufficiently accurate, the specific gravity is deter
mined from weighings in air and in water, and
from these data and the density of the water, the

182-31

density of the conductor is determined by the

equation
wawu

(27)

where w and wv are the weights of the specimens

in air and water, respectively, and Z> is the
density of the water. At a temperature of 21 C,
the density of water is 0.998 gm/cm3 and this value
decreases uniformly to 0.997 gm/cm3 at about
25 C.
Instead of calculating A separately, the value of
A from eq 26 may be substituted into eq 23, giving
i? m 1
'-|XTX2>'

(28)

From this equation it is seen that the resistivity, p,

equals the product of resistance per unit length
and mass per unit length, divided by density.
The length need not be the same for the resistance
and mass measurements if the material is uniform.
For two conductors having the same density, the
ratio of their resistivities is the same as the ratio
of their values for the product R/lXm/l. This
product is called "mass resistivity", and is a
constant that is characteristic of the material of
a conductor, being D times as large as the or
dinary, or volume, resistivity. The mass resis
tivity is often specified in the purchase of con
ductors for electrical uses, and if. is sometimes more
readily measured than is volume resistivity.
However, it is doubtful that this advantage is
sufficient to compensate for the confusion that
arises from the use of two types of resistivity.
It is usual commercial practice to specify
percentage conductivity rather than resistivity,
especially in the purchase of copper conductors.
Conductivity is the reciprocal of resistivity, and
percentage conductivity is obtained by dividing
the resistivity of the given sample into that of the
standard ana multiplying by 100.
By international agreement, the resistivity of
annealed copper is taken as 1.7241 microhm-cm
at 20 C [15]. This value was selected from
measurements of the resistivity of a large number
of samples of high-purity commercial copper wire
from both American and European refiners. The
value was agreed upon as a standard for reference
and was not intended to be the value for abso
lutely pure material. Copper has been produced
with a conductivity of several percent greater
than 100, which probably indicates a material
of higher purity than that of the standard.
Some effort has been made to secure interna
tional agreement for a standard for the resistivity
of aluminum, but so far copper is the only metal
for which such a value has been adopted.
The measurement of the resistivity of a liquid
may often be made by comparison with another

315-411 O - 68 - 13

liquid of known resistivity. In this case measure

ments of dimensions may be avoided as the known
and unknown are given the same dimensions by
placing them in turn in the same container or
"conductivity cell." The ratio of their resistivi
ties is then the same as the ratio of their resis
tances. Mercury is often used as the liquid of
known resistivity, since its value is accurately
known.
Although the resistivity of mercury is known
to a high accuracy, its use as a liquid of known
resistivity may lead to errors of considerable
magnitude. In comparing the resistivity of
liquids by placing them in turn in the same cell
and measuring their resistance, it is tacitly as
sumed that the current distribution through the
cell is the same for both liquids. This may be
incorrect unless the liquids have very nearly the
same resistivity, as the distribution of current in
the cell depends to some extent upon the resistance
of the metal electrodes. For high-resistivity
liquids the resistance of the electrodes may play
a negligible part in determining the current dis
tribution, but in the case of mercury, which is a
relatively good conductor, the current distribu
tion may depend to a considerable degree upon the
dimensions of the electrodes and upon the resis
tivity of the metal of which they arc made.
In determining the resistance of a liquid, it is
necessary to use an alternating-current bridge.
When a direct current is used, the ions in the
liquid will drift towards the electrodes thus making
the density nonuniform. Moreover, polarization
will often be produced by the liberation of gases
at the electrodes. When measured with alter
nating current a conductivity cell is found to be
electrically the equivalent of a resistor and a
capacitor in parallel. This requires a balance
of reactance as well as of resistance when the cell
is measured in an alternating current Wheatstone
bridge.
VII. References
[1] N. B. Pilling, Trans. Am. Electrochcm. Soc. 48, 171
(1925).
[2] R. S. Dean and C. T. Anderson, Trans. Am. Soc.
Metals 2, 899 (1941).
A. A. Somerville, Phvs. Rev. 31, 261 (1910).
J. L. Thomas, J. Research NBS 16, 149 (1936) RP863.
J. L. Thomas, J. Research NBS 13, 681 (1934) RP737.
F. Wenner and J. L. Thomas, BS J. Research 1?, 147
(1934) RP639.
[71 P. H. Bridgman, Proc. Am. Acad. Arts Sci. 52, 573
(1917).
8] J. L. Thomas, J. Research NBS3, 107 (1946) RP1692.
91 P. H. Dike, Rev. Sci. Inst. 7, 278 (1936).
10 K. B. Rosa, BS Bui. 5, 413 (1908-09) S107.
11 F. Kohlrausch, Wied. Ann. 2, 76 (1883).
12 E. F. Mueller, BS Bui. 13, 547 (1916) S288.
13 E. F. Mueller and F. Wenner, J. Research NHS 15,
485 (1935) RP842.
[141 F. Wenner and E. Wcihel, BS Bui. 11,27 (1914) S223.
[151 BS Circular C31 (1914).
Washington, March 30, 1948.

183-32

Precision Electrical Measurements

Calibration

of

Potentiometers

by

Resistance

Bridge

Methods

DAVID RAMALEY
National Bureau of Standards
Boulder Laboratories
Convenience and ease in calibration of generalpurpose potentiometers is provided by a Universal
Ratio Set. Any moderate-sized electrical standards
laboratory should consider the procurement of a
URS for potentiometer calibrations, especially if
its also are required.
Instruments & Control Systems\ol 37

T t might seem that the simplest way to calibrate

* a potentiometer would be to set up the instrument
under test to furnish potential differences which can
be measured by a potentiometer of similar rnnge
which previously has been calibrated. This method
does have the advantage of calibration under actual
operating conditions and can be used advantageously
on instruments of moderate accuracy. The method
requires that currents in both potentiometers be held
as nearly constant as possible for optimum calibra
tion accuracy. Also, thermal electromotive forces con
stitute a possible source of calibration errors.
If the appropriate resistance portions of a poten
tiometer are composed of resistance bridge circuits,
then the precision of measurement is not limited
by the extent to which currents can be maintained
constant. The resistance ratio data can be converted
rapidly into potential corrections for switch settings
and dials on the potentiometer under calibration.
The resistance bridge method requires step changes
in resistance as small or smaller than one part per mil
lion (Fig. 1). The total resistance between terminals
A and B must be constant, and independent of the
positioning of the contactor terminal C. The ratio of
resistance between terminals A and C to the resis
tance between terminals C and B is thus variable in
small increments. Such a tapped-resistor function is
provided by ratio sets or dividers.
The Universal Ratio Set tURSI
A conventional six-dial UBS provides the equiva
lent of a resistor of 2111.11 ohms, having 2,111,110
separate taps, providing step changes of 0.001 ohm.
It contains twenty 100-ohm steps (right in Fig. 2),
and five dial switches of the Fuessner or double-action
ganged type. Decades have resistance elements of
10, 1, 0.1, 0.01, and 0.001 ohms, respectively. The
total resistance between terminals A and B is in
dependent of the positions of the ganged dials. Turn
ing any dial adds as much resistance on one side of
contact C as it removes from the other side of contact
C. The resistance from C to A (or from C to B) can
be varied over the entire range in increments of 0.001
ohm. The UBS is calibrated and corrections applied
to its readings when using the instrument.
A Kelvin-Varley type of resistive voltage divider
with six dials could be used as the tapped resistor.
1.000,000 equal-valued steps are available, but even
more steps are desirable for calibrating potentiome
ters in bridge circuits.

Comparison of Resistances
The resistances to be compared are connected in a
series circuit, which in turn is connected to the A
and B terminals of a URS. The parallel circuit thus
formed is shown in Fig. 3. A battery or other source
of direct current is applied across this parallel circuit.
A Wheatstone bridge can be formed by connecting a
galvanometer between terminal C of the UBS and
any desired point along the resistances to be com
pared, such as Pi, P2, etc. A simple analysis'-4 in
dicates that resistors, a, b, and c can be compared bymeans of readings taken on the URS for balancedbridge conditions when the galvanometer is connected
184-106

c
9
ONE MH.LK3N OR MORE DISCRETE, EQUAL STEPS

FIG. I. SCHEMATIC CIRCUIT diagram of a resistor suitable for bridge circuit calibration of
potentiometers.

0001
fl/STEP

B J10

0.01
ft/STEP

10

\hJ^

^Cw^

1.0
ft/STEP

0.1
fl/STEP

K>

10

V***?

10
ft/STEP

|0

100
fl/STEP

VvL^

FIG. 2. SCHEMATIC DIAGRAM of six-dial universal ratio set (URS).

FIG. 3. ARRANGEMENT for

comparing resistances by
means of a URS, a slidewire,
or a resistive divider.

I R0R

in turn to resistor terminal points Pi, Pi, P$, P, etc.

The relationships can be expressed as follows:
R. - R,
R.- R.

(1)

R.- R.
R.-R.

(2)

Ri> Rj, R*> R<> etc. are the URS readings, and a, b,
and c are the resistances of resistors, a, b, and c.
Because these are truly Wheatstone bridge arrange
ments, the power source and the galvanometer can
be, and commonly are, interchanged in the circuit.
Pofntfomfr Calibration
This same type of Wheatstone bridge can be ar
ranged by connecting the appropriate portions of the

potentiometer circuit to a URS. Usually the resistance

comparisons required in calibrating a potentiometer
include (1) intercomparison of all step resistors in
the "measuring" or "EMF" circuit, (2) determination
of the uniformity or linearity of the standard-cell
(SC) circuit (if it is adjustable), and (3) determina
tion of the ratio of resistance of the EMF circuit to
that of the SC circuit for all ranges of the instru
ment. In case the potentiometer is of the Lindeck
type, certain resistances usually will have to be meas
ured directly in terms of a known standard actually
inserted in series with the potentiometer.
Fig. 4 is a schematic diagram of a simple poten
tiometer connected for calibration. The URS forms
two arms of a Wheatstone bridge, and the potentiom
eter circuit forms the other two arms, thus com
pleting the bridge. Bridge balance readings are taken
at all desired potentiometer settings of the EMF and
SC dials by connecting the power source in turn to

185-107

BA +

BAL2 I.I 1.0 .9 B .7 .6 .5 A .3 .2 .1 .0

KEY
(CLOSED)
sc
GAL
(SHORTED)
*EMF +
SCSC +

'KEY
(CLOSED)
EMF
GAL
(SHORTED)
EMF-

FIG. 4. ARRANGEMENT of
universal ratio set (URS) for
calibrating a potentiometer.

REVERSIBLE
| ADJUSTABLE
/\ DC
DC. POWER SOURCE |
^^VWvVWvVVNAVi^VVVVVVVVVvWvVWVv^
URS REPRESENTED AS A SLIDE WIRE
--

the SC-f, SC-, EMF+, and EMF- binding posts

of the instrument. Connections to all circuit resistors
requiring calibration can be made through these
four terminals. The required resistance comparisons
are obtained from the URS readings by means of
equations similar to equations (1) and (2).
The computation of resistances and the ease of
making the balances with the URS can be simplified
by making the URS direct reading in terms of a de
sired portion of the circuit under calibration. Usually,
the choice corresponds to the voltage expected from
a typical standard cell, such as a voltage of 1.0190
volts. To make the ratio set direct reading for this
selected voltage setting on the standard-cell dial,
the difference in readings of the URS when the power
source is connected to SC+ and SC terminals,
respectively, should be 1019.000 ohms, if the ratio set
is calibrated in ohms. This is achieved by adjusting
R by trial and error, usually with the help of pre
viously calculated data.
Assume that the URS has been made direct read
ing for the SC circuit and that the readings for the
circuit of Fig. 4 are those shown in Table 1. The
uniformity of the standard cell circuit has thus been
determined for points 1.01800, 1.01850 and 1.01950.

TABLE ICalibration Data for

Standard Cell Dial Circuit of Fig. 4.
POWER TO POTENTIOMETER
URS
TERMINAL
SETTING
READING
READING
(VOLTS)
(OHMS)
DIFFERENCE
SCSC +
SC+
sc+
SC+
sc+

1.0190
1.0190
1 01800
1.01150
1.01900
1.01950

1023.3631
4.363/
5.369
4.866
4.363
3.832

1019.000
1017.994
1018.497
1019.000
1019.531

The resistance of the circuit for setting 1.01800 is

lower than it should be by about 6 parts in a million.
For a setting of 1.01850 the circuit resistance is low
by 3 ppm and for 1.01950 the circuit resistance is
high by about 30 ppm. If we wish to achieve maxi
mum accuracy when using this potentiometer with a
standard cell of 1.01950 volts, we should set the
standard cell dial at 1.019469 instead of 1.019500.
The EMF circuit is calibrated in like manner. If
this has been accomplished and the results tabulated,
it becomes simple to compute corrections to the EMF
dial positions if the URS is direct reading.
Ranges or factors for multirange instruments are
obtained by taking readings on the URS for bridge
balances with the potentiometer factor switch set
on the available ranges. The exact procedure differs
with different makes and models of instruments. One
feature of importance in determining procedure is
the standardizing circuit. Some instruments stand
ardize on all ranges and some only on one range'
Special-purpose potentiometers and potentiometers
containing Lindeck-type circuits usually can be cal
ibrated by the URS and appropriate resistors in
serted into the circuit. The procedure in each case
depends on a circuit analysis of the instrument un
der consideration. Usually it is necessary to measure
the resistance of the Lindeck element resistors and
also to obtain a calibration of the indicating meter.'
References
1. D. Ramaley, "Practical Methods for Calibration of
Potentiometers," NBS Technical Note 172.
2. J. L. Thomas, "Precision Resistors and Their Meas
urement," NBS, Cir. 470 (1948).
3. F. K. Harris, Electrical Measurements, John Wiley
and Sons, Inc., New York, New York, 1952.
4. P. P. Brooks, "Calibration Procedures for D. C. Re
sistance Apparatus," NBS Monograph No. 39.
5. F. Wenner and E. Weibel, The Testing of Potentiom
eters, BS Bui. 11, 27 (1914) S223.
6. "Procedure for Certification of 7553 Type K-3 Uni
versal Guarded Potentiometers," Leeds and Northrup
Company, Directions Booklets Nos. 177151 and 177189.

186-108

NOTES ON THE CALIBRATION OF THE DIRECT READING RATIO SET

by
Paul P.

B.

Brooks

FIGURE 1. - Panel of Direct Reading Ratio Set

Figure 1 shows the essential features of the panel of a typical
Direct Reading Ratio Set. The ratio arms of the set are represented
schematically in figure 2. The arm CA is variable from 99.4445 ohms
to 100.5555 ohms. Adjustment of the resistance is made by the four
dials on the panel. The arm CB, with a resistance of 100 ohms, is
used for an approximate 1:1 ratio. Use of the binding post D, instead
of B, provides a 10-ohm arm, DC, for a 10:1 ratio. A' and B' are
extra binding posts at the ends of the respective arms to provide for
additional lead connections.

A'

A ioo

ioo a

b'

0<>MA/\/VWK>AA/^^
I

6d

FIGURE 2. - Schematic diagram of ratio arms

187-1

The dials for adjustment of the variable arm will be designated from
left to right as dial
dial
dial
dial

I,
II,
III,
IV,

0.1 ohm/step equivalent to 0.1% or 1000 ppra;

0.01 ohm/step equivalent to 0.01% or 100 ppm;
0.001 ohm/step equivalent to 0.001% or 10 ppm;
0.0001 ohm/step equivalent to 0.0001% or 1 ppm;

It is required to test each step on each dial to find the correction

for each dial setting to the nearest part per million. If a calibrated
Mueller thermometer bridge is available it may be used to check both the
fixed and adjustable arms. Another method is to set up another resis
tance ratio that can be changed in accurately known steps. For example,
two 1000-ohm resistance standards may be connected in series and, of
course, their ratio is "^qq* that is, y.

*f now a 1-ohm standard is

connected in series with one arm, the ratio is changed to ^qqqS

1.001. Thus the ratio is increased by 0.1% (1000 parts per million).
If this ratio is connected in parallel with the Direct Reading Ratio
Set a Wheatstone bridge can be formed. When the ratio formed with
the standards is increased by 0.1% the setting of the DRRS must be
also increased by 0.1% to keep the two ratios equal, that is, to keep
the Wheatstone bridge balanced. Any departure of this change from
0.1% is a measure of the correction to the DRRS reading. This, in
brief, is the method of producing known changes in ratio to calibrate
a DRRS.

FIGURE 3. - Bridge circuit for dial tests

A Wheats tone bridge circuit (fig. 3) is set up using the arms AC

and BC of the Direct Reading Ratio Set as the lower arms of the bridge.
A 1000 -ohm resistance standard forms the right upper arm of the bridge.
R in the left upper arm is a decade resistance box in series with a
finely adjustable resistance which will permit adjustment of resistance
to about 0.0001 ohm. R is used to balance the circuit. R' is a link
and a resistor or a combination of resistors used to change the resis
tance of the arm in increments accurately proportional to the nominal
values of the dial steps on the DRRS. A change of 1 ppm in the upper arm
corresponds to a change of 1 ppm in the lower arm. Since the ratio of
the respective arms is 10:1, a change of 1 ohm in the upper arm corresponds
to a change of 0.1 ohm in the lower arm. Hence, the increments in R'
corresponding to steps on the different dials, left to right, are 1 ohm,
0.1 ohm, 0.01 ohm, and 0.001 ohm, respectively, and they should be accurate
to better than 0.0001 ohm. The method for obtaining the desired increments
will be explained with the detailed procedure for each dial.
The dial being tested is initially on zero. Since corrections will
be taken from dial IV, dial IV is set near its middle position. The
settings of the other dials are immaterial provided they are not changed
during the test of the selected dial. The circuit is balanced by adjusting
R. Then R' is used to increase the resistance of the upper arm by the
appropriate increment (1 ohm for dial I, 0.1 ohm for dial II, etc.) and
the DRRS dial is advanced one step.
If the step resistor is sufficiently
accurate, and the circuit is still balanced the correction for the step
is zero. If it is necessary to advance dial IV to balance the circuit,
the dial step in question is too small and its correction is negative by
the number of steps that dial IV is advanced. If dial IV was advanced
one step, the correction is -1 ppm.
If dial IV has to be set back one
step, the correction is +1 ppm.
After the correction to each step has been found, a step-by-step
summation is made to supply the total correction applicable to each dial
setting.
The fixed arms, CB and CD of the DRRS can be measured with standard
resistors and another DRRS. Their values are not critical.
Resistance boxes do not supply the fine adjustment required of the
balancing resistor, R, so a fine adjustment in series with the resistance
box is required. The variable arm of another DRRS will supply steps of
0.0001 ohm. If a standard resistor (100 ohms for example) is connected
in parallel with the variable arm of the second DRRS, the parallel resis
tance can be adjusted in very small increments. Likewise, a decade box
adjustable in 0.1-ohm steps in parallel with a 1-ohm standard is usually
satisfactory. With 30 ohms in the box an increase of 0.1 ohm increases
the parallel resistance 0.0001 ohm.

189-3

Test of Dial I, 0.1 Ohm per Step

i-Qvwvf

'L'

looonr

B^g'

FIGURE 4. - Test circuit for dial I

Figure 4 shows the connections of the DRRS and accessories to a
mercury stand to form the Wheats tone bridge circuit shown schematically
in figure 3. Corresponding to R' in figure 3 is a 1-ohm standard and a
link arranged so that when the link is in position L the standard is out
of the circuit, but when the link is switched to position L' the resis
tance of the left upper arm is increased S ohms, that is, the 2-terminal
resistance of the standard.
An alternate method is to remove the link from the stand and to
replace it at L with the standard. The resistance of the arm is increased
S-L ohms, that is the 2-terminal resistance of the standard minus the
resistance of the link. The resistance of the link in most common use
at the National Bureau of Standards is about 40 microhms or 40 ppm of 1
ohm (NBS Monograph 39, Sec. 5.1). It is approximately the same as the
difference between the 4-terminal resistance and the 2-terminal resistance
of most 1-ohm standards of recent manufacture. Hence, if the 2-terminal
correction to the standard is small or the 4-terminal correction is nega
tive, it is preferable to use the link as a switch. If the 4-terminal
correction is positive, it is preferable to make the change by substitu
tion. Any inaccuracy in the 1-ohm increment should not exceed SO or 60
microhms.

190-4

The test of the dial proceeds as follows:

1.

Set dial I on 0, dial IV on some step in advance of 0. The setting

of the other dials is immaterial provided they are not changed during
the test of dial I.

2.

Place the link at L, to remove the standard resistor from the circuit.

3.

Balance the circuit with the balancing resistor, R.

4.

Transfer the link to L (or substitute S for L) increasing the resis

tance of the upper arm 1 ohm.

5.

Advance dial I to step 1. If the circuit is still balanced, the

correction for step 1 is zero.

6.

If the correction is not zero, adjust dial IV to obtain balance

(nearest balanced setting).

7.

Find the correction from the number of steps that dial IV was
changed and the direction of rotation. For example, if the initial
setting of dial IV was 5 and it was set back to 4 to balance the
circuit the correction was +1 ppm. If dial IV was moved to 7 to
obtain balance, the correction was -2 ppm.

8.

For the next step, leave dial 1 on step 1 and return dial IV to its
initial setting.

9.

Proceed as above for the test of step 2. Note that to rebalance the
circuit with the link at L, R must be increased about 1 ohm.

10.

After each step has been tested, make a step-by-step summation of the
step corrections to show the total correction applicable to each dial
setting.

11.

With the 1-ohm standard on the stand, test the total of the ten steps
of dial II by the same procedure. The result will serve as a check
on the total correction found in the next test.

12.

A check on the total correction for dial I can be made by comparing

the ten steps on the dial with a 10-ohm resistor on the stand in
place of the 1-ohm standard.

Table I illustrates the above procedure with data for a few steps
taken from laboratory notes. In the first column, the plus sign denotes
the unknown resistance of the fine adjustment. Since R serves only to
balance the circuit its numerical value is not important.

191-5

TABLE I. - Sample data from a test of dial I

Correction
Z
DRRS Rdgs
R
ppm
ppm
S out
S in
1000.6+
1001.6+
1002.6+

1009.6+

0003
1003
2003
mmm
9003

1002
2002
3002

+1
+1
+1

X002

+1

+1
+2
+3

+8

Ten Steps on dial II

-1
0X04
1000.6+ 0003
Ten steps on dial I with 10 -ohm Standard
X002
+8
1000.6+ ooox

Test of Dial II, 0.01 Ohm per Step

This dial may be tested by the procedure used for dial I if a 0.1-ohm
standard of the NBS type is available. This would be used on the stand
in place of the 1-ohm standard.
Otherwise an accurate 0.1-ohm reference resistance may be set up
from the usual laboratory equipment as shown in figure 5.

FIGURE 5. - Test circuit for dial II

192-6

The circuit is the Wheats tone bridge circuit described above with
a 1000 -ohm standard resistor in the right upper arm and the balancing
resistor, R in the left upper arm. Also in the left arm, there is a
10-ohm standard resistor which can be placed in parallel with 990 ohms
by the link. With the link in circuit the parallel resistance of 990
ohms and 10 ohms is 9.9 ohms. Removal of the link increases the resis
tance of the arm 0.1 ohm, which is the reference resistance required for
testing dial II. The reference resistance is accurate provided the 990
ohms in a resistance box is not in error by more than 1 ohm.
To obtain an accurate resistance of 990 ohms, if a calibrated plug
box is not available, a 1000 -ohm standard resistor can be used in the left
upper arm of the bridge, maintaining a ratio of 1000:1000. After the
circuit has been balanced by adjusting the DRRS, the standard is replaced
by a 10-ohm standard and a plug box in series. The bridge is rebalanced
by adjusting the plug box in 0.1-ohm steps. The resistance of the plug
box plus 10 ohms equals 1000 ohms, that is, the resistance in the plug
box is 990 ohms.
The following adaptation of the above procedure permits calibration
of the plug box without changing the test circuit.
a.
b.
c.
d.
e.
f.

g.

Disconnect the plug-box from the stand (fig. 5).

Replace the plug-box by a 1000-ohra resistance standard on the stand.
Place the link on the stand to connect the 1000-ohm standard and the
10-ohm standard in parallel.
Balance the circuit with the DRRS, or R, or both.
Remove the 1000-ohm standard, reconnect the plug-box, and replace the
link on the stand with an extra 10-ohm standard.
With 10-ohms plus the plug-box in place of the 1000-ohm standard, 990
ohms are required in the plug-box to rebalance the circuit. Adjust
the plug-box in 0.1-ohm steps until balance is restored.
Remove the extra 10-ohm standard and proceed with the test.
Test Procedure

1.
2.
3.
4.
5.
6.
7.

Place the link on the stand, set dial II on zero, dial IV on some
step in advance of zero.
Balance the circuit with R.
Remove the link which increases the resistance of the arm 0.1 ohm.
Advance dial II to step 1. If the circuit is still balanced, the
correction for the first step is zero.
If the correction is not zero, balance the circuit with dial IV, and
find the correction from the setting of dial IV as explained for dial I.
Test each step of the dial in the same way and make a step-by-step
summation of the corrections.
By the same procedure, find the correction for ten steps of dial III.

193-7

TABLE II. - Data from a test of dial II

DRRS
L in L out

STEP
1
2
3
4
X

0003
0103
0203
0303

0103
0203
0303
0404

0903

0X03

Cor. ppm E ppm

0
0
0
-1
0

0
0
0
-1
-1

Ten Steps on Dial III

0003

00X4

-1

Test of Dial III, 0.001 Ohm Per Step

This dial may be tested by the procedure outlined for dial II by
replacing the 990 ohms in the left arm with 9990 ohms. Since this value
is not critical, it may be replaced by a 10000-ohm standard.
The following method reduces the number of readings one half. No
link is used and the resistance of the upper arm is increased step by
step by increments of 0.01 ohm, corresponding to the 0.001 ohm steps on
the dial.

*3

FIGURE 6. - Test circuit for dials III and IV

194-8

The test circuit is modified as shown in figure 6 by using a

standard resistor, S, in parallel with a decade box, r, to produce a
variable resistance, R1, which can be increased in accurately-known
increments approximately proportional to the steps on the DRRS dial.
The decade box, r, is adjustable in steps of 0.1 ohm, accurate to 0.001
ohm. Values of S and r have been determined so that an increase of
0.1 ohm on the decade box will increase the parallel resistance, R',
0.01 ohm each step within the range of the dial with sufficient
accuracy. For dial III, the value of S is 1000 ohms and the values
of r range, in 0.1 ohm steps, from 2162.0 ohms for step 0 on the DRRS
dial to 2163.0 ohms for step 10. R' will be about 684 ohms, so about
316 ohms will be required on R to balance the circuit.
Procedure :
Step 0 - Set dial III on 0, dial IV in advance of 0.
2162.0 ohms. Balance the circuit with R.

Adjust r to

Step 1 - Increase r 0.1 ohm (2162.1). Advance dial III to step 1.

If the circuit is still balanced (or nearly balanced) the
correction to step 1 is zero. If a better balance is
obtained with a different setting of dial IV take the
correction from dial IV as for the preceding dials.
Step 2 - Return dial IV to its initial setting. Rebalance the
circuit with R. Increase r 0.1 ohm (2162.2). Advance
dial III to step 2.
If the circuit is still balanced,
the correction is zero. Otherwise, find the correction
from the change in dial IV.
Follow the above procedure for each step.
summation of corrections.

Make a step-by-step

The following table includes data from a test of dial III. Column
Dq shows the initial setting of the dials for the test of the step. After
the circuit was balanced with the balancing resistor, R, the decade resis
tance, r, was increased 0.1 ohm to the value shown in column r, dial III
was advanced one step, and dial IV was adjusted, if necessary, to restore
balance, giving the dial setting shown in column D^.
TABLE III. - Data from a test of dial III
STEP
0
1
2
3
X

D0

Dl

2162.0
2162.1
2162.2
2162.3

0003
0003
0013
0023

____
0013
0024
0033

2163.0

0093 00X3

195-9

Cor. ppm T, ppm

0
-1
0
-0

0
-1
-1
--1

If an error of one or two parts per million is not objectionable,

the same procedure may be used for dial II. For S use a 1000-ohra
standard resistor and advance the decade resistance, r, in 1-ohm steps
from 2157 ohms for step 0 to 2167 ohms for step 10.
Dial IV, 0.0001 Ohm Per Step
The procedure is the same as for dial III. Use a 100-ohm standard
resistor on the stand in parallel with the decade box, r. Increase r in
0. 1-ohm steps from 899.5 ohms for step 0 on the DRRS dial to 900.5 ohms
for step 10.
Dial IV can be tested by the procedure outlined for dial II. The
required resistances are 1 ohm in a standard resistor and 999 ohms. A
1000-ohra standard resistor instead of 999 ohms in a plug box is quite
satisfactory. The resistance of the two standards in parallel is 0.999
ohm. When the link is removed from the stand to disconnect the 1000ohra standard the resistance of the arm is increased 0.001 ohm.
Summary
As explained above, the Direct Reading Ratio Set forms the lower arms
of a Wheats tone bridge. The ratio of the 1000-ohm upper arms can be varied
to correspond to any change in the ratio of the DRRS arms. This is done
by resistance increments in the left upper arm which are proportional to
the nominal values of the dial stepson the DRRS. Resistance elements
which provide accurately known increments for each dial are as follows:
Dial I.

- A 1-ohra standard resistor in combination with a link to intro

duce the required increment into the arm.

Dial II.

- A 10 -ohm standard resistor in the arm. A resistance of 990

ohms in a plug box which can be connected in parallel with
the standard by a link. Removal of the link increases the
resistance of the arm 0.1 ohm.

Dial III. - A 1000-ohm standard resistor in parallel with a plug box. The
plug-box resistance is increased in 0. 1-ohm steps from 2162.0
ohms for step 0 to 2163.0 ohms for step 10.
Dial IV.

- A 100-ohm standard resistor is parallel with a plug-box. The

plug-box resistance is increased from 899.5 ohms for step 0
to 900.5 ohms for step 10.
Alternate Methods

Dial II.

- A 0. 1-ohm NBS type standard resistor (if available) to be

used as the 1-ohm standard was used for dial I.

Dial III. - A 10-ohm standard in the arm as used for dial II. A resist
ance of 9990 ohms in a plug box (or a 10000 ohm standard
resistor) which can be connected in parallel with the standard
by a link. Removal of the link increases the resistance 0.01
ohm.

196-10

Dial IV. - A 1-ohra standard resistor in the arm.

Instead of 999 ohms
in a plug box, a 1000 -ohm standard resistor can be connected
in parallel with the 1-ohm standard by a link. Removal of
the link increases the resistance 0.001 ohm.
Instead of the above procedures the fixed arms and the dial steps
of a DRRS can be measured with a calibrated Mueller bridge if one is
available.

USCOMM-NBS-DC

Text corrected November 1967.

197-11

DIRECT READING RATIO SET DIALS

o-oooi %,x i ppm aooi %,x 10 ppm ooi %,x 100 ppm ai %,x loooppm
' aooi a/step

aoi a/step

qi a/step

1.0a/step

10000 ppm

ioa/step '
USUALLY SET
ON STEP 9 OR 10

A>

raon/STEP

SET ON STEP 9

B _-io

%JUr^

,10

_^io

^wJUr

-jio

->J0>

^VvX^

^HvW^

B ITOI

DIAL SWITCHES ARE GANGED AS SHOWN

FIG. I. SIX-DIAL Universal Ratio Set (URS) with

provisions for converting URS to a Direct Reading
Ratio Set with ratios of I to 1 , 10 to I , and 5 to I .

DIRECT RATIO

READINGS

FROM

URS

DAVID RAMALEY and JOHN F. SHAFER

National Bureau of Standards

A
511

FIG. 2. SCHEMAT
IC of Direct-Reading
Ratio Sets resulting
from Modifications
of Universal Ratio
Set.

A SIX-DIAL Universal Ratio Set (URS) can be

quickly and easily modified for use as either
a four- or five-dial Direct-Reading Ratio Set
(DRRS) without altering its functioning as a Universal
Ratio Set. Such a modification may be a convenience
for those who have a limited use for both a Universal
Ratio Set and a Direct-Reading Ratio Set, and who
now have a Universal Ratio Set but cannot justify the
expense of an additional instrument. (It is assumed
that the reader is acquainted with the discussion on
ratio sets to be found in Section IV of NBS Circular
470. An abstract of this article is presented in NBS
Technical News Bulletin, Vol. 47, No. 9, September,
1963.)
Using a simple modification, the resulting Direct. Reading Ratio Set has a variable ratio arm of onethousand ohms, adjustable in steps of 0.001 ohm (one
part per million). The conventional four-dial DirectReading Ratio Set or VVenner Ratio Set, described in
NBS Circular 470 (Ref. 1) and NBS Monograph 39
(Rcf. 2) contains a 100-ohm ratio arm adjustable in
steps of 0.0001 ohm (likewise one part per million).
The Direct-Reading Ratio Set obtained by modify
ing a Universal Ratio Set is suitable for measuring rc-

90011 IN
9011 IN
X 100 DECADE X 10 DECADE

FOUR ADJUSTABLE
DIALS

100011

a ARRANGEMENT WITH 5 OHM RESISTOR ADDED.

10000
wyfo*
PPM
90011 IN
X 100 DECADE

9011 OR
lOOli
X 10 DECADE

I000PPM I00PPMI0PPM IPPM'

FOUR ADDITIONAL
ADJUSTABLE DIALS

b. ARRANGEMENT WITHOUT THE ADDED RESISTOR.

198-1

I000H

STANDARD
RESISTORS

GALV.
NOT USED
FOR DRRS
20 STEP DECADE A

FIG. 3. TYPICAL
MEASUREMENT
setup using Universal
Ratio Set as DirectReading Ratio Set.

c^1 1 tol

6t6
SET ON 9-

SET ON 9

SET ON 9 OR X
xkxxxj
XIOOOO

(xiooo)

^xia^

(^)

lOOii/STEP I0I2/STEP Ift/STEPO.Iil/STEP 0.0 l/STEP O.OOIil/STEP

read five dials instead of four. When the 5-ohm resistor

sistors ranging from about 1 ohm to about 1 megohm.
is omitted (as shown in Fig. 2b), some of the readings
To obtain a precision of a few parts per million, the
for the Direct-Reading Ratio Set will necessitate set
switch contacts in the adjustable arm should be reting the 10-ohm-per-step dial on step 9, and others
peatable within about 0.001 ohm. The ten-step decade
will necessitate setting this dial on step 10.
dial-switches of most Universal Ratio Sets meet this
This modified Universal Ratio Set will actually
requirement. The switch contact resistance is not crit
cover a larger ratio range when step changes in the
ical in the 20-step (100 ohms per step) decade dial
10-ohm-pcr-step dial are made. Therefore, when mak
because this resistance is either in the battery branch
ing computations of reading differences, it will be
or galvanometer branch when the instrument is in use.
necessary to record the readings of the 10-ohm-pcrThe modification required to make the adaptation
step dial, as well as the four lowest dials. (We are con
consists of a lead soldered to step 19 on the 20-step
sidering a ratio range of 0.995 to 1.005, such as is cov
dial decade, and terminated in an external binding
ered by most conventional Direct-Reading Ratio Sets.)
post B'i toi as shown in Fig. 1. A 5-ohm precision
A Direct-Reading Ratio Set capable of measuring
quality resistor (or two 10-ohm resistors in parallel)
ratios of 10-to-l or 5-to-l can be arranged by soldering
is connected to binding-post A and terminated in a
additional leads into other steps of the Universal Ratio
separate binding post. A'. With these additions, the
Set. For a 10-to-l ratio, a lead should be soldered into
instrument is usable for a 1-to-l ratio, 4-dial Directstep 10 on the twenty-step dial and, for a 5-to-l ratio,
Reading Ratio Set when the 100-ohm-per-step and the
a lead is soldered to step 11. These additional ratio
10-ohm-per-step dials are both set on step 9 for all
connections are shown to the added B'10 to i and
measurements.
B'b to i binding posts in Fig. 1.
Fig. 2a is the schematic diagram of the resulting
Fig. 3 is a schematic diagram of a URS modified to
Direct- Reading Ratio Set. A fixed resistance of 995
a DRRS, and arranged to compare standard resistors
ohms is obtained by the use of the steps of each of the
in a Wheatstone Bridge circuit using a 1-to-l ratio.
first two decades and the added five-ohm resistor. The
four lowest URS decades constitute the adjustable
References
DRRS dials. These dials become the X1000, X100,
1.
J.
L.
Thomas,
Precision
Resistors and Their Measure
X10, and XI part-per-million step dials, respectively.
ment, NBS Circular 470.
This modification can be simplified even more by
2. P. P. B. Brooks, Calibration Procedures for Direct
omitting the 5-ohm resistof, if the user is willing to
Current Resistance Apparatus, NBS Monograph 39.
Reprinted from the January 1966 iisue ofInstruments & Control Systems
315-411 O - 68 - 14

199-2

JOURNAL OF RESEARCH of the National Bureau of StandardsC. Engineering and Instrumentation

Vol. 64C. No. 4, October-December 1960

A Method
Link

of

Circuit

Controlling
of

the

the

Thomson

Effect of Resistance in the

or

Kelvin

Double

Bridge*

David Ramaley
(.May 12, 1960)
Circuitry of the double bridge is reviewed to emphasize the significance of the link
circuit as a part of the bridge network. A simple analysis demonstrates the need of some
equivalent or substitute for low link circuit resistance in situations where adequately lowlink circuit resistance cannot be achieved physically. An appropriate method of controlling
the potential difference in the link circuit is shown to accomplish the same effect as an
actual reduction of resistance in the link circuit. The simple circuit modifications required
to utilize this method are explained. A step-by-step procedure is outlined for making
measurements with bridges incorporating these modifications, and some typical examples
are discussed illustrating the advantages realized.

1. Introduction
Double bridges of the Kelvin type are convenient
not only for the measurement of low-resistance
standards, conductivity samples, and components,
but also for calibrating resistors incorporated as
intrinsic parts of circuits. When using the double
bridge for calibrating resistors permanently con
nected in circuits, it is sometimes not feasible to
achieve an arrangement with a link circuit of suffi
ciently low resistance. For such situations the
precision of measurement will be diminished if the
usual procedure is followed in balancing the bridge.
The method outlined in this paper utilizes a sup
plementary power source and an auxiliary galva
nometer to minimize the potential difference in the
link circuit.1 The application of this method has
been found to reduce substantially the effect of
excessive link resistance, and its practicability has
been demonstrated in several different situations.
Descriptions and uses of double bridges are given
in references listed at the end of this paper.

SHORTING PLUG
-EI}
Figure 1. Double bridge.

2. Review of Double-Bridge Fundamentals

Figure 1 is a circuit commonly used in double
bridges. X and S are four-terminal resistors; X
is the unknown and S is a reference standard of
known resistance. The link circuit is designated as
L and includes the resistance of the link itself and
idso the additional resistance in the circuit between
the resistor branch points m and n. A and B are
the main ratio arms, and a and b are the auxiliary
ratio arms. These same symbols are used also to
refer to resistance values. The ratio arms are
connected to the potential terminals of A' and S1
by leads designated as LA, LH, La, and Li,. When
these notations are used to indicate lead resistances,
thev include not onlv the resistances of the external
Contribution from the. Radio Standards Laboratory. National Bureau of
Standards, Boulder, Colo.
1 This method was suceesled by the lute Dr. Frank Wenner of the N'ational
Bureau of Standards in Jy40.

connectors but all resistances between the ratio

arms and the resistor branch points m, n, etc.
Leads LA and La are constructed to be conveniently
varied in resistance by means of built-in adjustable
rheostats. P is a direct current power source such
as a battery, R is a reversing switch, and G is a
suitable galvanometer.
The delta network consisting of the two auxiliary
ratio arms and the link circuit can be transformed
readily into the equivalent Y network. The re
sistance of that branch of the Y between the Y junc
tion point and m, adjacent to the resistor, A, is
L(a+La)/(a + Ltt+b + Lb+L). The resistance of
that Y branch, between the Y junction and n,
adjacent to the resistor, S, is J.(b-\-Lb)/(a+La+b-{Lr\-L). The third Y branch is in series with the

200-267

galvanometer and need not be considered in the

present discussion because it does not affect the
position of bridge balance.
This transformation enables one to treat the double
bridge as a simple Wheatstone arrangement, from
which one obtains the equation of balance :
L(a+L.)
(a+L.+b+Lb+L)
L(t>+ U
(a+L+ b +U+L)

A+LA
B+LB

(1)

When rearranged this equation becomes

means of making sufficiently precise lead adjust

ments, so that the accuracy of measurement is
definitely limited. A reduction in accuracy under
conditions of excessive link circuit resistance therefore
may be unavoidable unless one makes some modifica
tion of the usual double bridge. Such a modification
is described below. It minimizes the effect of link
resistance even though the actual resistance of the
link is abnormally high.
With the bridge in balance and power applied, the
ratio of the potential difference EL, developed
across the link circuit, to the sum of the potential
differences Ex+Es, developed across the resistors
X and S, is given by

_(A+LA)S^_
L(A+LA)(b+L>)
X= (B+LB) T (B+LB)(a+La+b+Lb+L)
L(a+La)
{a+Ltt+b+Lb+L)

Ex+Es

(7)

(2)

If the substitions:
A+LA_A(l + g)
B+LB
B
'

(3)

a+Ltt A{\+h)
b+U
B
'

(4)

L(b+Lb)
=D
S(a+La+b+Lb+L)

1(5)

When all circuits of the bridge have been properly

adjusted, the ratio of S to X is almost identical to
the ratio of B+LB to A+LA and also to the ratio of
the corresponding auxiliary branch arms. It may be
assumed with sufficient accuracy that
b+L>
X+S a+La+b+L,,

and

EL
Ex+Es
(6)

This can be considered a fundamental equation of

the double bridge and can be reduced to the simple
Wheatstone equation, X=SA/B, provided the terms
g and D(gh) can be made negligible. This is
achieved in practice as follows: With A and B
shorted out and with the bridge essentially balanced,
the term, g, commonly is made very small by ad
justing LA so that the leads LA and LB have nearly
the same ratio as X to S or A to B. The term, ft,
is made closely equal to g by making a/6 as nearly as
possible the same as AIB, not only by using a common
set of controls for the main and auxiliary ratio arms
but also by removing the link and then making a
balance by adjusting the ratio LjLb to nearly X/S
or A/B. The term D usually is much smaller than
one and normally is minimized by using a link
circuit of low resistance.
In some situations it is impractical to arrange a
link circuit with suitably low resistance. Then the
balancing procedure employed in minimizing the
term (p h) must be accomplished with much higher
precision to avoid deterioration in accuracy. This
necessity is evident from the term D{gh) in eq (6).
If L is large, D also is proportionately large, and the
term (g h) must be made ext remely small. However,
this requires extremely fine adjustment of the leads,
and some double bridges are not provided with

(8)

Combining eqs (7) and (8) gives

are made, eq (2) becomes

SA[l+g+D(g-h))
X=
B

Ka+La+b+L,)
{a+L.+ b+Lt+D
X+S

L(b+L>)
S(a+La+b+L+L)

(9)

which from eq (5) is the term D. For practical use,

then, the double bridge eq (6) may be written as

Consequently, reduction of the potential difference

EL across the link circuit would produce the same
effect as reduction of the link resistance and would
eliminate the necessity of adjusting the ratios L to
Lb and LA to LB with higher precision than ordinarily
used when the link resistance is small. This can be
accomplished by introducing into the link circuit a
potential difference of such magnitude and polarity
that the potentials at the branch points m and n are
brought to practically the same value. This applied
potential difference is introduced across an auxiliary
resistor of appropriate magnitude which has been
inserted into the link circuit.
3. Double Bridge Modification for Abnormal
Link Circuit Resistance
Figure 2 is the circuit of the modified double
bridge. P' is the auxiliary power supply. Z is the
resistor inserted into the link circuit across which
the auxiliary potential difference is applied. R and
R' are reversing switches which can be operated
simultaneously. G' is an auxiliary galvanometer

201 -268

shorting plug
an

Fioure 2. Double bridge with modified link circuit.

used in detecting potential differences in the link

circuit. The auxiliary galvanometer is connected as
shown rather than to m and n because of physical
difficulties sometimes encountered in connecting
directly to the branch points. By varying the poten
tial difference applied to Z from the auxiliary power
source, the. potentials at the branch points can be
adjusted to practically the same value.
The modified double bridge can be balanced by the
following steps:
(1) With both battery circuits closed, the potential
difference indicated by the auxiliary galvanometer is
minimized by varying P' .
(2) A preliminary balance of the bridge is made by
adjusting in unison the main and auxiliary ratio
arms while operating the two reversing switches
simultaneously until a balance is indicated by the
main galvanometer.
(3) The leads in series with t he main rat io arms are
adjusted to the proper ratio by shorting out the
main arms and adjusting one of these leads. Both
power sources are used simultaneously, and the
balance is indicated by the main galvanometer.
(4) With the main power supply disconnected,
the lead resistances in series with t lie auxiliary ratio
arms are adjusted to the proper ratio by varying one
of the leads until balance is indicated by the main
galvanometer. Current is supplied by the auxiliary
power supply. An alternate method of adjustment,
commonly followed for double bridge operation, is to
open the link circuit and use only the main power
supply while adjusting the leads to the auxiliary
ratio arms.
(5) Final balance is achieved by repeating step 2.
If the bridge is too far out, of balance at the outset,
a repetition of steps 3, 4, and 2 may be required.

The readings of the bridge ratio arms A and H

corresponding to the final balance are used to com
pute the value of the resistor X.
Except for step 1, involving the auxiliary galva
nometer, the above procedure is that commonly
used in balancing a double bridge. The selection of
circuit components as well as the precision necessary
in making the adjustments in steps 1 to 5 wdl be
governed by the particular case under consideration
and the desired accuracy of measurement. Ordinar
ily step 1 is not unduly critical. However, good
stable power supplies and switches are required,
since variations in either power source adversely
affect the accuracy obtainable with the modified
double bridge.
This modification was developed in connection
with the calibration of a special microvolt potentiom
eter. The measuring circuit of this potentiometer
utilizes low-resistance elements connected in series
within a thermally shielded enclosure. In the
measurement of one of these resistors using a double
bridge, current lead connections were found to be
feasible only by incorporating the adjacent, perma
nently connected, circuit resistors. This arrangementproduced an abnormally high link circuit resistance,
and the resulting ratio of the link circuit resistance to
the sum of the resistances of X and S was approxi
mately 10. With the modified double bridge,
however, it was feasible to measure the low-resistance
element to the desired accuracy.
4. Examples
The improvement in accuracy achieved by use of
this modification can be evaluated for each indi
vidual case. It depends upon the resistors under
comparison and upon the bridge parameters. Both
the main and auxiliary ratio arms commonly have
resistances of 50 to 500 ohms. Lead resistances
commonly range in magnitude from 0.01 to possibly
2 ohms. Resistors suitable for comparison in a
double bridge may range in magnitude from a few
microhms to about an ohm.
An arrangement that will serve as an example for
this ordinary range of measurements will now be
considered. Let us assume that four-terminal 0.001
ohm resistors X and S are to be compared in a
bridge with ratio arms of approximately 100 ohms
and lead resistances of about 1 ohm, and that the
ratio of A to B can be adjusted to or estimated to
about one part in a million. The accuracy of adjust
ment of the resistor S and the ratio of A to B will be
considered to be known adequately. The uncer
tainty associated with these terms outside of the
bracket in eq (6) is a systematic error inherent in the
measurement regardless of the arrangement of the
link circuit or lead connections and is not relevant to
the present consideration. We are concerned pri
marily with the effect of link resistance and the pre
cision of lead balances upon the correction terms g
and D{<jh) in eq (6). Let us assume that a pre
cision of 30 ppm is desired and that the expression
SA/B can be, treated as a constant during the adjust

202- 269

merits. The sum of the correction terms g and

D(gh) in eq (6) can differ with each successive
measurement depending upon how consistently
balances are executed but in no case should exceed
0.00003 at the time of reading if the desired pre
cision is to be maintained. Ordinarily a very lowresistance shorting link can be used and the term D
in eq (5) or its equivalent EL/(EX+ES) or L/(X+S)
is less than or equal to one. To maintain the
desired precision for this situation, g and h each can
be as large as 0.00001 without giving g+D(gh)
a value larger than 0.00003. The ratio arms
including the leads need not be balanced more
precisely than 10 parts in a million or 0.001 ohm.
This implies that the leads must be adiusted to one
part in a thousand because they are 1 ohm in magni
tude. If this is accomplished, each of the terms g
and h in eqs (3) and (4), will be not larger than
0.00001 and the sum of g and D(gh) will be not
larger than 0.00003.
Next consider the situation in which it is impossible
to arrange a link circuit so low in resistance that D
can be held to a value of one or less. The expression
D(gh) can be held within the prescribed limit only
by adjusting the leads more closely than 0.001 ohm
because (gh) must be decreased to the same extent
that D is increased. Contact resistance variations in
adjustable lead circuits commonly exceed 0.0001 ohm
and thus may limit the fineness of lead adjustment.
Furthermore, some, rheostats used for the adjustable
leads may not be variable in steps finer than 0.001
ohm. If we assume such excellent conditions that
variations in contact resistances of 0.0001 ohm con
stitute the factor limiting the fineness of adjustment
of leads, g and h in eqs (3) and (4), can be reduced
to 0.000001 and (gh) to a maximum of
0.000002. For these values D must be smaller
than 15 if diminished precision is to be avoided.

For D=15, g+D(gh) could be a maximum of

0.000001 15(0.000001 0.000001)or0.000031
which is slightly larger than the permissible
0.00003.
If in the comparison of the above 0.001 -ohm
resistors, the value of L cannot be physically reduced
below 0.05 ohm, D will be about 25, as computed
from eq (5). The terms g+D(gh) will be equal
to 0.000051 if the leads can be adjusted to 0.0001
ohm and to 0.00051 if the leads can be adjusted
to 0.001 ohm. If the link potential difference is
reduced by a factor of 25, then for making a meas
urement, the conditions arc the same as if D were
actually no greater than one. In considering again
the favorable condition in which the leads can be
adjusted to 0.0001 ohm, the improvement in precision
obtained by reducing the link potential will be in the
ratio of 0.000051 to 0.000031 or a factor of about
two. In the case of the bridge in the less favorable
condition, the precision gained will be about 0.00051
to 0.00003 or about 17. These same factors likewise
can be regarded as representing an unavoidable loss
in precision if this modification is not employed.
Other examples with different circuit values and
requirements can be treated in like manner.
5. References
[1] F. Wenner, The four-terminal conductor and the Thomson
Bridge, Bui. BS 8, 559 (1912) 8181.
[2] F. Wenner, Methods, apparatus, and procedures for the
comparison of precision standard resistors, J. Research
NBS 25, 229 (1940) RP1323.
[3] J. L. Thomas, Precision resistors and their measurement,
NBS Circ. 470 (1948).
[4] F. K. Harris, Electrical measurements, p. 282-288 (John
Wiley & Sons, Inc., New York, N.Y., 1952).

203-270

(Paper 64C4-44)

JOURNAL OF RESEARCH of the National Bureau of StandardsC. Engineering and Instrumentation

Vol. 67C, No. 1, January-March 1963

Method for Calibrating a Standard Volt Box

Bernadine L. Dunfee
(August 22, 1962)

A volt box provides several discrete ratios and permits the accurate measurement of
direct voltages (maximum about 1,500 volts) through their reduction by suitable factors to
values within the measuring capability of a potentiometer. The ratios of a volt box are
measured against those of a standard using a "difference" technique. The standard volt
box used at the National Bureau of Standards possesses those design features required for
a ratio standard and lends itself to a "self-calibration" technique. This paper describes a
method for measuring its ratios at rated and above rated voltage to an accuracy of 10 part*
per million or better using a Direct Reading Ratio Set and a group of resistance standards.
Although the method is described in its particular application to the NBS standard it can
be used equally well in the measurement of other ratio networks. Errors that might appear
in the measurements and procedures for evaluating their magnitudes are outlined and the
derivation of a general expression that defines all ratios in terms of measured quantities is
presented in an appendix. Comments and further details regarding the standard and its
use are included.

1. Introduction
The value of a measured voltage, either direct or
alternating, is derived from the national standard of
emf. At the National Bureau of Standards the
unit of voltage is maintained in terms of the mean
value of emf of 48 saturated cells. This emf is
only slightly larger than one volt so that the accurate
measurement of larger voltages of any frequency
usually requires the design of special ratio networks
that can serve as voltage ratio standards, because of
their stability and high accuracy. The volt box
and the potentiometer with which it is used are
classic examples in the measurement of moderately
high direct voltages.
The volt box referred to here, as differentiated
from other types of d-c ratio devices (or voltage
dividers), provides several discrete ratios so that
the voltage to be measured (maximum about 1,500
v) is reduced by a suitable factor to a value within the
measuring capability of a potentiometer. Insertion
of a volt box in a measuring circuit introduces an
error in the measurement unless the volt box divides
perfect ly or the corrections to its nominal ratios are
known and applied. In any event, each ratio must
be measured, preferably at rated voltage and at a
reduced voltage to evaluate the effects of self heating
and leakage resistance.
The method of calibration employed at NBS
utilizes a difference technique in which the potential
at each binding post of a volt box under test is
measured with respect to that at the corresponding
binding post of the standard volt box, with both
instruments connected in parallel and for the same
nominal ratio. Special features must therefore be
incorporated in the standard volt box to assure long

term stability and the resistance sections must be

arranged in a proper sequence with respect to mag
nitude so that a "self-calibration" procedure is
possible and accurate to better than 10 parts per
million (ppm). The standard volt box used at
NBS, as well as a method developed for its calibra
tion, have been described by Silsbee and Gross [l].1
A second method for calibrating the standard volt
box was recently developed and has been under
observation for more than a year. The associated
circuitry and components have been incorporated
in a semiportable test console and the only connec
tions that need be made at the time of test are those
to the d-c supply and to the volt box. The present
paper describes this method and uses the NBS
standard volt box to illustrate the technique. Addi
tional comments are included on the use of the
standard volt box in calibrating other volt boxes.

2. Comments on the Calibration of a Volt

Box Against the Standard
A volt box consists, in general, of a large number
of resistance coils connected in series. At boih the
low- and high-potential ends and at appropriate
junctions along the resistor, taps are brought out to
binding posts to form the high voltage side. The
maximum voltage that can be applied between the
zero post and any of the others is governed by the
design of the resistance coils and is so marked at
each binding post. In addition, the resistance of the
lowest section is tapped at an appropriate point and

204 1

1 Figures in brackets indicate the literature references at the end of this paper.

leads are brought out from the zero potential end

and this tap point to binding posts, forming the low
voltage side for connection to a potentiometer. Any
given voltage ratio is defined as the ratio of the
voltage applied to the high side to that appearing
across the low side. These ratios are integers and a
given volt box, for example, may cover values from
1,500/1.5 to 3/1.5 if it is to be used with a 1.5-v
potentiometer or values from 1,500/0.15 to 3/0.15
for a 150-mv potentiometer.
The test instrument is calibrated with reference to
the standard in the circuit shown schematically in
figure 1. The two volt boxes, having the same
nominal ratio, are connected in parallel through
leads li and 4 across a suitable d-c source. The
measuring branch, consisting of a detector, a low
range potentiometer, and a reversing switch (not
shown), is inserted, in voltage opposition, into gaps
identified at positions vu v3, v3, and v4. For each
position the output of the potentiometer is adjusted
for a null on the detector so that at balance the
potential difference between corresponding points
on the two volt boxes is equal to that of the poten
tiometer. The potential differences Vi and vt are
small since they are essentially the potential drops
in leads lt and U ; similarly, if the volt box under test
is of high quality, potential differences th and v3
will be small. Thus only moderate accuracy is
required in the measurement of the potential differ
ences in order to realize an extremely high accuracy
for the values of ratio.
The equation that defines the ratio of the volt
box under test in terms of the ratio of the standard
and the measured potential differences may be
derived as follows. Let the respective polarities
be as shown in figure 1. By definition, the true
ratio of the volt box under test is

Summing the potential drops around loop "badeb"

gives
Vat-V2-Vd.+Vt=0
or
V.t=V+(v,-vJ.
(2)

2V,l=iV[l+M,]=^V ab

On rearranging terms and remembering that

NIn=N,n^N,[l + u,] the above expression becomes
[1+M.l
[1+mJ1
Y T-t, y

(1)

where y.z is the correction in proportional parts and

N is the nominal ratio.

By definition, the true ratio of the standard volt

box is
N.t=Nn[l+.]=^
or
Vd=

V-M

where N,n is the nominal ratio of the standard and

n, is the correction in proportional parts. Substi
tuting this value in eq (2) gives
(3)
Summing the potential drops around loop "defcod"
we have
or
Vif=V-{v,+vt).
When this expression is substituted in eq (3) Vab
becomes
v _V-(Vi+vt) , .
s
V- N.n[l+u,}+{v>-V3)
and if substituted in eq (1) the true ratio of the volt
box under test becomes
v ri.
VN,[l+n,]
n[ ~V- (*,+) +AU1 +n,](v3-v3) '

so that
GUARD CIRCUIT
f-wwwwwvwwwwwwvwv?
STANDARD VOLT BOX
d'
*/vA/WWWWWWWWWWWV\^-tI

o=6 6
A
Lib

-1 POTR.

(Pl+Pj), (V2V2)
plus higher order terms.
V
V/N

(4)

If vu v?, v3, v,, and V are expressed in volts then,

neglecting higher order terms, the correction to the
nominal ratio is

O
Mx=Mi+

+ ^yj^ m proportional parts.

(5)

v0LT B0X UNDER TEST

Figure 1. Circuit for calibrating a volt box.

If Pj, Vt, v3, and vt are expressed in microvolts and u,

in ppm the correction is given in ppm. It should
be noted that if the relative polarities of the supply
voltage (V) and the potentiometer differ from tnose
assumed in figure 1, the appropriate sign must
be given to the recorded values of the measured
quantities.
205-2

If

As will be seen later in the discussion on the cali

bration of the standard volt box, the measured
resistances of all sections and their summations are
referred to the resistance of the first section which
is chosen to lie between junctions d' e'. Thus the
true ratio of the standard volt box, given as

u
in the above derivation, is consistent with the meas
ured ratio. (The effects of rod resistances,2 such as
o.d.', on the measured ratios are discussed in the
section on errors.)
It is advisable that the potential drops in leads lx
and l2 be small in order to achieve the highest ac
curacy in the measured ratios without placing undue
demands on the accuracy required in the measure
ment of the potential differences. A slight modifi
cation made in the test circuit would eliminate the
lead effect and at the same time simplify the com
putation of test data. Referring to figure 2, let
small resistances Rx and R2 be connected in the Z,
and It branches with the d-c supply connected be
tween their sliding contacts. The potential drops
in the two sections of branch lu as well as those in
branch Zj, are in opposition so that it is only neces
sary to adjust the respective sliders to achieve can
cellation, which is indicated by a null on the detector
connected at the Vi and vt positions, respectively
(see fig. 1). For this condition, the second term on
the right of eq (5) becomes zero.
3. Standard Volt Box
A volt box must have several special features
incorporated in its design if it is to serve as a stand
ard in making ratio measurements to an accuracy
of a few ppm. (1) Some provision must be made for
reducing the effects from leakage paths to negligible
amounts; (2) the selection and arrangement of the
resistance wire must guarantee freedom from the
effects of relative humidity as well as those arising
from self-heating ; (3) the corrections to the ratios
should be small and the resistances of the tapped
sections proportioned so that a "self-calibration"
procedure of sufficient accuracy is applicable.
3.1. Leakage
The volt box is a high resistance network operating
at rather high voltages so that careful consideration
must be given to minimizing leakage paths into which
a portion of the current may be diverted from points
along the working resistance. Effects from such
parallel paths may be reduced sufficiently by keeping
the insulation resistance high. For example, if the
total resistance of the high side is 250,000 ohms, the
resistance of a leakage path bridging this section
must be 2.5X10' ohms for an uncertainty no greater
than 0.01 percent, and 2.5 X1011 ohms if an error
no greater than 1 ppm is required. This protection
The rod resistances mentioned throughout the text refer to the leads which
extend from the resistance sections to the binding posts where connections are
made to the external circuit.

STANDARD VOLT BOX

'
^

VOLT BOX UNDER TEST

1,
V
Figure 2. Branches for compensating lead drops.

'

may not be adequate in maintaining long-time

stability because of possible accumulation of moisture
and contamination across the insulating surface or
possible deterioration of the insulation over long
use. Additional protection must be provided by a
guard network which, in effect, maintains the shields
surrounding the resistance sections at the same
potential as that of the resistance section within.
The standard volt box discussed here is equipped
with a guarding resistance whose corresponding
sections can be connected in parallel with the working
circuit when the 15-v and nigher ranges are used.
Metal guard rings surrounding each binding post
of the working resistance and connected to the
corresponding binding post of the guarding resistance
prevent leakage across and through the top panel.
The working resistance is divided into sections of
not more than 25,000 ohms with the resistance coils
oi each section mounted on separate insulating
panels. Each of these insulating panels is fastened
to the top panel through a metal block that is
connected to the guard resistance and maintained
at the appropriate potential to prevent leakage from
one insulating panel to another. In use, leads that
connect the standard volt box to the volt box under
test and to the other parts of the measuring circuit
carry their own shields which are maintained at
their appropriate potentials through connections to
the binding posts of the guard circuit.
3.2. Humidity and Self-Heating
Any effect from relative humidity on a properly
guarded volt, box originates at 1 he resistance coils. It
has long been known that certain insulating materials
used for wire covering and for protective coatings of
resistance coils are susceptible to changes in relative
humidity and produce changes in the resistance of the
coils. Although the effect might be expected to be
negligible in materials available today or, if not, to
be sufficiently suppressed in ratio devices, the matter
must be considered if accuracies of a few ppm are
desired. The particular standard volt box referred to
in the original paper was found to have seasonal
variations as large as 30 ppm as a result of changing
humidity, so that provision was made for continu
ously supplying dry air at a pressure slightly higher
than atmospheric. In 1960 a new unit was purchased
in which the individual resistance coils are hermeti
cally sealed. To date, no measurable seasonal
variations because of changing humidity have been
observed.

206-3

In theory, the effects from self-heating could be

measured and the appropriate correction applied;
but these effects can be avoided by designing the
unit for lower rated current, by selecting the proper
kind of resistance material and by providing the coils
with sufficient cooling surface within an adequate
exchange volume. Both the original standard and
the one recently acquired have a working resistance
of 333 % ohms/v and each coil is wound with a single
laver of manganin wire taken from the same spool.
There is sufficient freedom from self-heating that,
with double rated voltage applied, the correction
differs from that at reduced voltage by less than 5
ppm.
3.3. Self-Calibration Feature
The most difficult and critical problem encountered
with any standard is the determination of its correc
tions to an accuracy considerably better than that
required for the test instrument. The method should
not only satisfy the immediate requirements in
accuracy but should be capable of providing still
better accuracy in the anticipation of future demands.
One of the most powerful measuring techniques
available and one capable of the highest accuracy is
that in which small differences are measured between
like and nearly equal quantities. The circuit
suggested by Silsbee and used in the designs of both
standard volt boxes lends itself to this ' difference"
technique. Its arrangement and adaptation to dif
ference measurements can be best understood by
referring to figure 3. The circuit has 21 tapped
sections beginning with the "0 to 0.15" section of 50
ohms and continuing to the "750 to 1500" section of
250,000 ohms.3 This resistance chain can be con
sidered as being formed of four groups Mu M2, M3,
and A/4 in which group Mx serves as the first section
of group Mj, M2 as the first section of group M3, and
M3 as the first section of group Mt. Being so
considered the following pattern is observed. Each

> The last section of 290,000 ohms is not present In the standard volt boles at
NBS. It is included here for completeness.

group contains six sections. The first five sections

of any group have nominally equal resistances while
the sixth section has a value nominally equal to the
sum of the first five. It is possible, therefore, to
measure the small differences between the first 50ohm section and each of the remaining four sections
of group Mi, and follow this by measuring the differ
ence between the 250-ohm section and the sum of the
first five sections. Since Mn-X forms the first
section of Mn this measuring sequence can be repeat ed
for each group in turn. Because of the continuity
within a group and from one group to the next, the
resistance of any portion summed from the zero end,
compared with that of the first 50-ohm section can
be computed from the nominal value of their ratio
and the sum of the measured differences.
4. Direct Reading Ratio Set Test Method
With the rapid growth of standards laboratories
evident in recent years, voltage ratio standards of
the type discussed here are no longer the exclusive
property of the National standardizing laboratory.
The method recently adopted at NBS for calibrating
the standard volt box and described herein makes use
of equipment that is readily available in many of
these laboratories. Thus, from the information that
is given on both the theoretical and practical aspects
of the method, the standards laboratory can perform
the calibration without recourse to NBS. The
method is not limited solely to those volt boxes
having the particular sequence of sections indicated
in figure 3 but can be used for any network in which
the resistance of each section is nominally equal to
the resistance of some combination of lower-valued
sections. The only change lies in the defining
equation and its derivation.
4.1. General Comments
A Wheatstone bridge network is used in which the
successive difference measurements, referred to in
the previous section, are made by inserting in turn
each section or summation of sections in the unknown
arm of the bridge. Aside from a d-c supplv and
detector, the only equipment required is a Direct

1
OHMS-Ji 50 50 50 50 50 250 1 500 500 500 500 2500 11 5K 5K 5K 5K 25K '50K
50K 50K 50K 250KIi
mrrrrrrrrtr^^
volts-Jo .15 .30 .45 .so .75 i.s 3 4.5 s 7.5 15 30 45 to 75 150 300 450 600 750 1500
1
1
1
1
1
I
- GROUP M, - 50011- CROUP M2- 5000fi
-GROUP Mj - 50.000Q-CROUP *, - 500,00011 Figure 3. "Self-calibrating" network of a standard volt box.
207-4

Reading Ratio Set (DRRS) and a group of seven

resistance standards: one each having the values 50,
200, 500, 2,000, and 5,000 ohms, and two having a
value of 50,000 ohms. (Other types of resistors or
resistance assemblies could be substituted for the
standard resistors if they remain stable during the
sequence of measurements.) The execution of the
test is quite rapid and with the circuit incorporated
in a suitable console a complete calibration including
computation can be made in an hour by experienced
personnel.
In the original method the test voltage was limited
to some extent because of possible heating of the
components. For example, measurements on the
50,000-ohm sections, if made at their rated voltage
of 150 v, required a power dissipation of 2.25 w in one
of the standard resistors. Since it is advisable to
keep the power dissipation appreciably less than this
amount, periodic calibrations were made at reduced
voltage after it was initially established that changes
arising from self-heating were negligible. The
DRRS method has the advantage that all sections
can be measured at their rated voltage. Under this
condition the maximum power dissipation required
of the standard resistor is 0.45 w. With some of the
resistor assemblies now available the method can be
extended to measurements at 200 percent rated
voltage so that the overload capability of the
standard volt box can be easily determined.
The method has a further advantage in that the
positions of the supply and galvanometer (detector)
relative to the circuit are such that the change in
circuit resistance as seen by the galvanometer is
quite small. Thus the damping and galvanometer
response are little affected throughout the measure
ments.

may be so connected and the other located in the

S-arm, and (3) they may be arranged as shown in
figure 4a. The relative magnitudes of the rod
resistances in the NBS standard volt box indicate
figure 4a as the preferred connection.6
The remaining sections whose resistances are
250 ohms or more are measured in the network of
figure 4b. (The rod resistances are expected to be
negligible compared with the resistance of the section
to which they are assigned.) The guard circuit of
the particular standard volt box under discussion is
not equipped with taps below the 15-v range so
that it is used only on the ranges of 15 v and higher.
Thus, for all successive resistances measured after
ihe first five and to, but not including, the "15 to
30" section, the guard branch of figure 4b is not
used. For all remaining measurements the guard
branch is connected as shown in the figure.
The appropriate bridge network is balanced by
adjusting the dials of the DRRS for a null on the
detector as each section, or summation of sections,
beginning with Rul is inserted in succession in the
"unknown" arm of the bridge. The procedure is
as follows. With S=Sa\ = 5Q ohms, the bridge
of figure 4a is balanced with the first 50-ohm section
The resistance per section of the guard of the volt box is 0.8 that of the working
resistance.
1 The influence of these rod resistances can be completely avoided by modifying
the circuit so that the 50-ohm sections are measured as 4-terminal resistors.
This technique is described in reference 12].
Rl,l Rl,2 Ri,3 Rl,4 RI,S Rl,t

WORKING
RESISTANCE

0.'
.15' 30 .45 .60 .75 1.5

4.2. Calibration Procedure

The measuring circuits are indicated schematically
in figures 4a and 4b; the reversing switch for the
supply is omitted to preserve simplicity. Resistances
A and B represent the two arms of a DRRS (in
cluding leads) each having a nominal value of 100
ohms and S is a standard resistor always serving as
a "Dummy" resistance and having a nominal value
equal to that of the volt box section being measured.
The guard branch that appears in figure 4b includes
a resistance of about 80 ohms 4 in series with the
appropriate guard section of the volt box; and the
leads of the measuring circuit are of such length and
size that at balance the DRRS reads near midscale.
Measurements on the first five sections are made
with the circuit of figure 4a. The d-c supply is
connected to the 1.5 binding post and the detector
to the zero dot (0.) so that rod resistances I,, l2 . . .
/n_i appear in the A- and S-arms of the bridge
rather than as part of the resistance being measured.
Inequalities among the rod resistances can introduce
errors in the measured ratios and are of greater
significance when the low-valued sections are
measured. There are three alternatives for locating
the rod resistances: (1) both rods may be connected
as part of the section being measured, (2) one rod

Figure 4a.

Circuit for measuring the lower-valued ratios.

oovW^AA^^A/WWp/YVJvW

GUARD

0 .15 .JO .75 1.5 3

I son

208-5

1^^c^_

Figure 4b.

WWVA

Circuit for measuring the higher-valued ratios.

inserted in the unknown arm. In rapid

sequence, each of the remaining four 50-ohm sections
is inserted in turn and the bridge rebalanced in each
instance. The resistance of each 50-ohm section
can now be stated in terms of the resistance of
and the measured difference, and the group
t-5
can be summed to give ^2 Ri k (computed). The
t-1
next set of balances completes the series of measure
ments required for group Mx. It consists of two
measurements using the appropriate circuit of figure
4b and with S=SM = 250 ohms. The reference
balance is made with the sum of the 50-ohm sections
t-5
5j ^i * having a nominal value of 250 ohms,
t-i
connected into the unknown arm through the zero
and 0.75 binding posts and is immediately followed
by a measurement of Ri, ,. This set of measure
ments provides a value of 7?i, , (sixth section) in
t-5
terms of ^2 Ri t and the measured difference.
t=i
t-5
Since the value of
R\ * is obtained by summing
t-1
the values from the first five measurements, straight
forward substitution gives the value of
, in
terms of Ru i and the sum of the measured differences.
t-5
Adding this value of R{ , to the value of 2 Ri *
t-i
(computed) gives the total resistance of group Mx
referred to the reference section RXA. These two
sets of measurements expressed mathematically are
as follows:
Ri. ,= Sal [1 + (A.i -D0)]
i?..2 = a.[l + (A.J-.Do)]
/?..3 = Sal[l+(A.3-A)]
Ru*=Sal[l + (Dui-D0)]
Ri.s=Sal[l + (Du5-D0)]
where {Dyx D0), (D]2De), etc., are the differences
in proportional parts 8 as deduced from the readings
of the DRRS; D0 is the reading of the DRRS for
an exact 1 : 1 ratio. (The significance of D0 is
explained in appendix A.)
Subtracting the first equation from each succeed
ing equation and from itself, gives
*i.i=Jfi.i

Ri.*=Ri.i
Rl.* = Rl.l-\-Saldiii
Ri.s=Ri.i+Saidi_s
t-5
t-5
^2Ri,k=5R1,i+SaX S^i.*
t-1
k-l
S appendii A.

where dll=(D12 D1A); rfi.3=(A,3 A.i) etc.

Since Sat = RiA closel y enough,
t-5
r t-5,/ -\
k-l
L
t-1 o J
The second set of measurements stated in similar
fashion is
jgfli.*=Si[l + (A..-A)]
fi,.f=SM[i+(A.-A)].
On subtracting, this gives
i. .= *i.*+Sirfi. , where ds. ,= (Z?,. ~A. .)
t-i
Substituting eq (6) in the above expression and
remembering that
Sm^SRx i closely enough, RLt becomes
Ri. ,=5i?,. , [l +g d~f+di. ,]

(7)

Adding eqs (6) and (7) and arranging terms gives

g Ri.k+Ri. '=10Ru[i+g^+^']

(8)

which states the total resistance of group Mx in

terms of Rit i and the measured differences.
Group Mi, having a nominal value of 500 ohms
and a true value given by eq (8), serves as the first
section of group M2 which, as in the case of group
Mi, consists of five nominally equal sections plus a
sixth section having a value nominally equal to the
sum of the five sections (see fig. 3). The sequence of
seven balances is now obtained for group M2. In
the first set of this sequence, with S,= S<I2=500
ohms, each of the 500 ohm sections, beginning with
group Mi as the reference section, is inserted in the
bridge to obtain five measurements. In the second
set two balances are obtained, first with all 10
sections (2,500 ohms) in the unknown arm and then
with the eleventh section (2,500 ohms) inserted;
for this set, S= S2= 2, 500 ohms. Since group Mt
has a computed value given by eq (8), the summa
tion procedure can be extended through group M2
which contains all sections up to the 15 v tap and
becomes the first section of M3.
The same procedure is carried forward through
groups M3 and Mt with resistance S always having
a value nominally equal to that of the section being
measured.
The mathematical treatment presented in this
section attempts to clarify the significance of the
computation process associated with the step-up
procedure (sometimes called a bootstrap technique),
which may be masked in the more formal approach
given in appendix B.

209-6

4.3. Computations

5. Test Console

The readings of the DRRS having been recorded,

the summation of their differences can be computed
according to the following equation (developed in
appendix B) to give the correction to the nominal
ratios of the standard volt box
r k-S
k-S
k-S
S ^l.t+S <k.k +2 <-l.*
*-l
t-l
Nm.t=N'm. , *=1
~*~
5

After the test method was investigated for more

than a year with the circuit assembled in "bread
board" fashion, a small test console was designed
and built so that the set-up time for a test would be
reduced to a minimum. The console, shown in the
photograph and indicated in figure 5, houses the
DRRS, the galvanometer with its damping resist
ance, a mercury stand to accommodate the standard
resistors, the 80-ohm guard resistor and the battery
reversing switch. The DRRS, located in the lower
portion of the console, is connected to the circuit by
short leads brought to three binding posts located
on the vertical panel immediately above the termi
nals of the DRRS. (With this arrangement the
DRRS can be easily removed and made available for
other work.) The detector is a portable galvanom
eter mounted on a platform within the enclosure
immediately above and to the left of the DRRS.
The galvanometer scale is observed through an
opening in the front panel and the damping resist
ance, permanently mounted on the inside, is adjust
able in steps by a knob accessible to the operator.
The mercurv stand, located within the console on a
platform adjacent to the detector and above the
DRRS, can accommodate either a single resistor or
several arranged in either series or parallel combina
tion. A removable top permits ready access to all
components and wiring.
As shown in the wiring diagram of figure 5, the
battery reversing switch, accessible on the front
panel, carries a two-conductor shielded cable that
terminates at a fitting located on the side panel.

I
!

I . +<n-l,I , f-1
o
T
i.

(9)

Where jVm,t=true value of the ratios,

iVi,t= nominal values of the ratios;
(i's=measured differences in proportional
parts.
The schedule of computation given in table 1 was
derived, in effect, by applying eq (9) to successively
increasing values of k and m. Its arrangement
permits the computation of all corrections in a single
process without having to apply the equation directly
each time a ratio is determined. The summation
of the differences within a set (five-set or two-set)
are kept separate until the final manipulation. This
is indicated by columns 4, 5, and 6. The d's are
the differences within each set; the g's are the
corresponding summations; and the r's refer to the
separate quantities appearing in the final terms of
eq (9). Column 7 is the summation of the sets for
each successive ratio.

Table 1. Schedule for computing corrections for the standard volt box

Section in Resistor Rending of

bridKi:
S
DRRS
arm
0-0. 15
0. 15-0.30
.30-0.45
. 45-0. 60
.60-0. 75
0-0. 75
0. 75-1.5
0-1.5
1.5-3.0
3.0-1.5
4. 5-6. 0
6. 0-7. 5
0-7.5
7.5-15
0-15
1.5-30
30-45
45-50
80-75
0-75
75-150
0-150
150-300
300-450
450-500
600-750

Ohms
50
SO
50
50
50
250
250
500
SIX)
500
500
500
2,500
2,500
5.000
5.000
5,000
5,000
S.OOO
25,000
25,000
50,000
50.000
50.000
50.000
50,000

Tf8
= Sd'j

-Sd'i

d'

Correction referred to
0-0. 15 section

Range

ppra
Dm
01J
0,.i
Cm
0i.i
010i,.
0i.i
02.!
014
0i..
0i.i
0i..
02..
01.1
0!.:
01..
01..
0..1
01..
01.,
0*.i
0,.!
0..1
0...
0..1

di.i-0i.i-0i.i-O
*.1-014-01.1
di.j=0i.J-J}|.,
d,,,= 0,,,-0,.,
di.s-0i.i-0i.,
d,=0,.,-0i..-O
di.i-0l.i-0l..
dj.i-0i.i-0i.i-O
di.j- 01.1-01.1
dij-0i.i-0,.i
di..-0i,,-0j.i
dj.i- 01.1-0:.,
d;.,= 0!.,-0,.. =O
dj.,-01,,-01.,
di.i-0i.i-0j.i-O
di.:- 01.1-01,,
di.j-0j.j-0j.i
dj.,-0j.,-0!.,
dj.j-0j.j-Dj.,
di..-0i.,-0j..=O
di.,= 0j.,-0j,,
d..i-0..,-0..,-o
d,,i=0,.j-0,.i
d,.!-0,j-0,,i
A.1-01.1-01.1
d..i-0..j-0..i

(7,.!=di ,.
9i 4?i.i+d,.i
?,..-v,.!+d,..
(fi4?,..+di.j

r,.i->4,.i
ri.i-SSJu
r,..= V.?i.<
r,.i-ts?,.j

(114-0
MM
Mi J1*14
r,.j
Mi.. r,.
Ml.J 1*14

0-0. 15
0-O.30
0-0.45
0-0.60
0-0.75

Si.idi.i

ri.i-Wfi.i

W.IMl 4+1*1.1

0-1.5

92.1 dj.i
<*2.J=0J.l+dj.J
9i..=?:.j+dj.t
?i.j=?i..+di.j

rj.i- Viji..
ri.i ^ffj.s
*2.< = '
rj.l=>5(|i.j

M4Ml.l+rj.l
m:.j=MJ.i+ri.J
Mi..Ml.i+ri..
M2.1M2.I+1*14

0-3.0
0-4.5
0-6.0
0-7.5

ffl.idi,,

r2.i- Hji.i

MJ.lM2.J+1J.I

?!.! = dj.l
oi.i-?i.i+dj.j
ffi.i?i.i+di.<
7!.i-?j..+di,j

M.i l59j.i
ri4- lS?s,i
rj.. Ugs.i
n.i- Vs?i.i

MJ.J-MJ.i+ri.i
M4MJ.I+rj.l
Mi..MJ.i+ri..
MJ.J-MJ.i+ri.i

oj.,= di..

ri.,-J4?j.i

M.I|U4+1*I.I

?..i=d..j
9*.!fl.i+d.j
9*..-9.j-r-d...
9..iff...+d..i

r..j= hfl..i
r.4-M<4
r...-l.?...
r..i-ji9i.j

M..2 M4tl+1*4
M.4M..1+1*44
IM.1M4.1+ri4
M..SM*.i+r*.i

210-7

0-30
0-45
0-30
0-75

0-300
0-4SO
0-S00
O-750

MERCURY STAND FOR S

O.-LO ' LEAD L
LO
GUARD
LEAD H
2>
^'I.S-HI

to
DAMPING
RESISTANCE

p^|===g; supply
DRRS

Figure 5.

Test console and layout.

An external d-c supply is connected to this fitting

with a pair of shielded leads equipped with a mating
connector. The side panel also carries two singleconductor coax fittings, identified as "1.5-Hi" and
"Hi," and three binding posts, identified as "O.-Lo,"
"Lo,", and "Guard."
When the first five sections of the standard volt
box are measured separate unshielded leads are used
to connect the terminals of the volt box (1.5 post,
high-potential post of the section being measured,
0. post, and low-potential post of the section being
measured) to the respective terminals of the console
(1.5-Hi, Hi, O.-Lo, and Lo). For all other measure
ments special leads H and L are used to connect
the section under test as indicated in figure 5. When
the guard of the volt box is used, the pigtail of lead
H is connected to the high-potential terminal of the
guard section and a connection is made from the
"Guard" binding post to the low-potential terminal
of the guard section. Otherwise, these last two con
nections are omitted.
6. Consideration of Errors
An examination of errors requires a careful study
of all known effects that could render a measured
value different from the true value within the
specified accuracy. An appraisal of all components
of the measuring circuit as well as a justification of
approximations must be included, even though in
this instance one would expect the associated errors
to be negligible.
The first possible source of error to consider is the
DRRS. As in the usual case, this type of error can be
avoided by calibrating the instrument and applying
corrections. Apart from this procedure, however,
it should be emphasized that an error in the measured
value is diluted when the values of the ratios are
considered. For example, if each scale reading were
in error by t=l ppm and all errors were in such a
direction as to add when computing each difference
their contribution to the error of a given ratio, when

summed within any one of the five-sets, would be

(k-l)2(
k

2(k-\)
k

where k is the number of summed sections. For

k=5 the error would be 8/5 = 1.6 ppm. Similarly,
the error introduced in the measurements of the
two-sets would be
-2=1 ppm.
The maximum error would occur for the highest
range since it contains all the sets. This would
amount to only 9.4 ppm under these adverse con
ditions. Since the corrections to the DRRS used at
NBS are less than 0.3 ppm and their signs and
magnitudes have a random distribution, the errors
in the ratio measurements are almost certain to be
less than 1 ppm with no corrections applied to the
dial readings.
A source of error is concealed in the base (or
balance) equation used in appendix B to develop the
general expression for the volt box ratios. Referring
to appendix A, the exact base equation is given as
Rn.*=S

(Dm.k-D0)
1+
B

whereas, the approximate value used in the mathe

matical development in appendix B is
Rm.t=S[l + (Dm.t-DB)].
The latter departs from exactness by an amount
jt~ {Dm *D0) but is of second order.

For example,

if ~ were as large as 0.1 percent and (Dm tD0) rep

211-8

resented a change of 500 ppm in the A-arm, the value

for Rm_ i would he in error bv only 0.5 ppm.
Another error is concealed through approximation
in developing the general expression that defines
the volt box ratios (appendix B) when it is assumed
that the value of the "Dummy" resistor, including
leads, is exactly equal to that of the section being
measured.
The error introduced through this
assumption is second order since it is the product of
two small terms. For example, consider the summa
tion of the first five sections (refer to eq (6)).
*-8

If Sai differs from /?,., and is given by Sal = R1A

the summation becomes
S**-Wu.['+sV+6V]
where the magnitude of the error is ji, 1=5 dt.
If both ii\ and the summation were as large as 500
ppm, the error would be only 0.05 ppm.
Another possible source of error arises if the resist
ance in any one arm changes during the measurements
of a given set. This change could occur in the leads
and contacts, for example, and its effect would be
greatest when measuring the 50-ohm sections. A
change of this nature can be considered as a change
in DB and an estimate of the error it introduces in a
measured ratio can be obtained by considering the
measurements of the first set. Assume an extreme
case in which the resistance of a given arm increases
in equal increments after the first measurement so
that for subsequent balances D0 takes on new values.
The measured values for the set become

precautions are taken, the probability of this kind of

an occurrence is quite remote and could be easily
observed from the detector response. A single,
abrupt change would be more difficult to detect but
an error from this cause can be avoided by taking
a repeat set of measurements.
The final consideration involves those errors that
arise from the unavoidable presence of the rod
resistances. This source of error could be the most
critical of any because of its insidious nature. Two
errors must be examined when considering the rod
resistances. If the rod resistances are equal they
can introduce an error if, when the standard volt
box is used, the connections differ from those made
during the process of calibration. The second error
occurs if the rod resistances differ in magnitude.
Consider the cause of the first error. When in
use, the connections to the standard volt box are
as shown in figure 6. There are no currents in rods
li and l2; rod ln, however, carries current by virtue
of its connection to the supply. Thus the ratio of
the standard in use is

(10)

N'=^R
When the first five sections are measured
fig. 4a) the rods are in the A- and S-arms.
the ratio under consideration were one
solely to these sections, the measured ratio

(refer to
Thus, if
confined
would be

URn
(ii)
Ri.i
and eq (10) can be written as
N^+^N^l+j^

//i.i = S[1 + (A.i -#<>)]

tf..2=Sal[i+>,.2-(A>+)]
ff,.,=Si[i+A.s-(A>+2)]
Rl.t = Sal[l+D1.i-(D0+3t)]
tf1.5=Sai[i+A.5-(A>+4*)]

In the measurement and summing process through

the higher ranges, the rod resistances form part of
the section being measured. (See fig. 4b.) How
ever, if the rod resistances are equal, their effects

where c is the change in DQ between successive bal

ances. After taking differences, any given sum
within the set can be written as

f-l

f=l

t' = l

(12)

STANDARD VOLT BOX

4 /*^YV\AAAA/WVV\A/VVWWV

J
6

Thus an error in any ratio formed within the set is

(lc' \)(
S

For example, if a lead resistance in

O
<t

creased in increments of 0.0001 ohm in the S-arm

(50 ohms) the magnitude of the error introduced in
the ratio

would be 4 ppm.

O O
O
Q
1 1
VOLT BOX UNDER TEST
\\^AA/WWWWWVWWVV<

If the usual
Figure 6. Circuit emphasizing one effect from rod resistances.
212 -9

cancel and the difference between the ratio in use

and the one measured is the same as that given in
eq (12). It is apparent then that the ratio in use
differs from the measured ratio by an amount equal
to the rod resistance divided by the total resistance
of the range used and occurs even though the rod
resistances are equal. If the standard volt box were
used as a 4-terminal voltage divider the above cor
rection to the ratio in use would be jy^0 since l0 be
comes part of the high voltage side.
The magnitude of an error arising from inequali
ties in the rod resistances depends on where the rods
are located in the bridge network. As stated in sec
tion 4.2, there are three alternatives for connecting
the rods and the choice of orientation is not as clearly
indicated when measuring the 50-ohm sections as
when measuring the higher-valued sections. For
the purpose of analysis, consider the measurement
of the low-valued sections and the particular con
nection shown in figure 4a.
For this condition, differences among the rod re
sistances may be considered as producing changes in
Z>0 so that, if the respective rod resistances are
identified as lu l2, etc., the measured resistances are

7. Summary
A particular type of voltage ratio standard and a
method for its calibration are discussed. Although
the method is described in its particular application
to the NBS standard it can be used equally well in
other networks. Errors that might appear in the
measurement of its ratios and procedures for evaluat
ing their magnitudes are outlined. With a standard
volt box of proper design, the calibration method as
set forth should yield results that are good to an
accuracy of 10 ppm or better. This estimate of
accuracy is based on the theoretical considerations
of the circuit and its parameters, on the continued
good agreement with those values of ratio obtained
by the older method, and on other self-consistent
tests.
The author appreciates the assistance of Rita
McAuliff and Ronald Dziuba who made most of the
measurements and constructed the test console.
8. Appendix A
8.1. Comments on the DRRS

i.=SJi+A.i-A]

[i+A.3->o+^+^-2]

[i+A.s->o+^+^-J]
If the rod resistances are measured, their contribu
tion to the error of any ratio can be evaluated.
Comparable sets of equations can be formed for the
other two cases and corresponding errors evaluated.
If the successive values of rod resistances neither in
crease nor decrease, it cannot be stated categorically
that one connection is preferred to another. Thus
a comparison of the resulting errors for the three
possible connections must be made and the connec
tion chosen which yields the smallest errors. Such
a comparison applied to the NBS standard volt box
indicated that the circuit of figure 4a is preferred.
As measurements are continued through the higher
ranges, the rods are connected as part of the large
resistance being measured so that their effects are
further reduced!
In the NBS standard volt box recently acquired,
rod resistances from zero through the 6-v range are
about 5X10"4 ohm and their differences are suffi
ciently small and random in sign that the maximum
error they produce in a ratio measurement is less
than 0.3 ppm.

Complete discussions on the DRRS and its use in

the accurate measurement of d-c resistance are
contained in references [2], [3], and [4] so that an
elaborate treatment is not warranted. However,
certain aspects are emphasized here in order to
clarify the meaning of particular equations, as well
as justify their use.
The DRRS is used principally in comparing two
nearly equal resistances and under these conditions
has an accuracy of about 1 ppm. It comprises two
resistance arms of about equal magnitudes with one
arm B' having a single-valued resistance (in most
cases 100 or 1,000 ohms) and the other consisting of
a fixed resistance in series with an adjustable section.
The adjustable branch contains, in most cases, four
resistance groups, each provided with a selector to
vary the resistance and register the change on suit
ably marked dials. The resistance of the A'-mm
can be varied by about 0.5 percent on either side of
the B'-arm value to a least count of 1 ppm. The
dials register in increasing numbers as resistance is
added to the fixed portion of the A'-&vm. Since the
DRRS is used to measure differences between two
nearly equal quantities, the dial readings can be
translated to ppm or, as is often the case, the dials
may be marked in ppm.
The reader should note carefully the form in which
small corrections are recorded or used in derivations
and computations. As usually constructed, a change
of one step on the lowest dial of a four-dial DRRS
corresponds to a change of 1 ppm in the ratio. At
one particular balance of a measuring circuit the
reading of the DRRS should be thought of simply
as a four-digit number, such as 5472. If a second
balance of the same circuit yields a DRRS reading
of 5459 the measured ratio change is 13 ppm, this

213 -10

being the first point at which the abbreviation

"ppm" should be introduced. In recording data
and computations (see table 1) it is convenient and
space saving to enter DRRS readings as four-digit
numbers, and ratio changes in ppm. In deriving
equations, however, (and corrections based thereon)
ratio changes should be expressed in proportional
parts. A ratio change in proportional parts is one
millionth of the change expressed in ppm. In the
above illustration the ratio change is 0.000013
proportional parts. See reference [2] for further dis
cussion of these terms.

a 1 : 1 ratio. Next, let Si be replaced by a resistance

whose value is R and let AB be the resistance of the
.A-arm of the DRRS under the new balance condi
tion. The equation of balance is
(14)
Subtracting (13) from (14) and rearranging terms
gives the value of R as
R=S+S

(AR+h)
_(Bm+l3)

(A9+ll)
(Bm+l2.

8.2. Bridge Balance Equation

Figure 7 indicates a bridge network in which the
adjustable arm of the DRRS is represented as a
section of fixed resistance in series with a uniform
slide wire of high resolution. Let Bm indicate the
resistance of the Z?-arm of the DRRS, Am the resist
ance of its .4-arm for which AmBm, l2 and lt the
respective lead resistances with U^U, and S the
resistance of a standard serving as the "Dummy"
resistor. For convenience, let the DRRS be without
error as previously determined from calibration.
Assume first that a standard resistor Su having a
resistance equal to S, is inserted in the unknown
arm and the bridge balances at some point for which
the A- and 5-arms of the DRRS are not equal.
Let the resistance of the A-&rm of the DRRS be Ac
for this balance condition. The equation of balance
is given by

(15)
Since Am and Bm equal the respective resistances of
the A- and fi-arms of the DRRS when Am = Bm, let
AK=Am+AB
A0=Am+&o
where AR and A0 are small changes in the resistance
of the slide wire. Equation (15) becomes
(Am+*lt)-(Am+*Jl
Bm+h
J
which can be written as

(13)
Thus Aq is the resistance of the A-&nn of the DRRS
when the ratio arms (A- and B-) of the bridge are in

(16)

An
since -^-=1.
Now the dials of the DRRS (assumed
Bm
to be without error) indicate, in effect, that in going
from a position corresponding to the resistance A*
to that of AR, the resistance of the yl-arm of the
DRRS has changed by -. in proportional parts;
Am
and similarly, in going from A to Aa the resistance
has changed by -

Hence,

*-=(DB-Dm)md^=(D0- Dm)
Am

11

where DR, D0 and Dm refer to the respective readings

of the DRRS translated into proportional parts.
Substituting these values in (16) gives

!
DRRS

(D D0)

R=S
B
Figure 7.

Circuit indicating the DRRS with respect to the

balance equation.

or, if -J- is sufficiently small

Bm

214-11

(17)

R Sf \ + <J)m- D0)}.

(18)

or
Rm.k ^m, 1 "I" Sam(lra ic

In the notation used in developing the general

expression given in appendix B, this corresponds to

(21)

where
dm.k = [Dm,k-DmA]
The sum of the first five sections of group m is

Rm.t=Sam[l + (Dm.t-D0)).

S R~.*=5Rm.i + Sam d.k

t-i
*-i

(22)

It should be noted that the concept of Do has been

generalized for the purpose of this paper to indicate
that setting on the DRRS for which the ratio
arms between terminals 3 and 4 have a 1 : 1 ratio.
In most uses of the DRRS, particularly in the com
parison of resistance standards, arms A and B
contain only the DRRS and Do is a constant, defined
as the setting for which the A- and ii-arms of the
DRRSh&ve a 1 : 1 ratio.

since each section has the same nominal value and

Sam is used throughout the set of measurements.
If now the resistor Sam is replaced by Sum and the
bridge is balanced first, with the five sections (con
nected in series) and then with the sixth section
inserted in the unknown arm, we have

9. Appendix B

S Rm.t=SdM + (Dm.,-D0)}
k-l

(23)

Rm..= Stm[l + (Dm.,-Dt)]

(24)

and
Referring to the text and to figure 3, the working
resistance is considered as made up of four groups,
with each group containing six sections. The first
five sections of any one group have nominally equal
resistances and the sixth section has a resistance
nominally equal to five times this value (sum of the
five sections). Counting from the low-potential
end, the first section of each group (after the first)
is formed by all the preceding sections connected in
series. Any section, therefore, can be identified by
two integers; the first, m, specifying the group to
which the section belongs, and the second, k, specify
ing the position of the section within its group.
If the Wheatstone bridge of figure 4 is balanced
as each of the first five sections of group m is in
serted in succession in the unknown arm, then the
resistance of the first section and of the kth section
can be written as,

/here
Dm-, is the reading of the DRRS when the sum of
the five sections is inserted;
Dm,, is the reading when the sixth section is
inserted.
Subtracting (23) from (24) gives
t-5
or
I; = 5
m, 1 2j Rm. * 4" Sbmdm, I
fc = 1

(25)

where
m, ( " [^m. I

Dm i] .

(19)
Substituting (22) into (25) gives
and
Rm.t=Sam[l + (Dm,k~Dn))

Rm.t 5/^m, \-\-Sam

(20)

where
Sam is the resistance of S and has a nominal value
equal to that of the volt box section being measured;
Dm.\ and Dmk are the readings of the DRRS in
proportional parts 7 when the first and Jfcth sections,
respectively, of group m are inserted in the unknown
arm ;
Do is the reading that would be indicated on the
DRRS if the Wheatstone bridge were in exact 1 : 1
ratio.

(/m.t + 'Som^m. (

(26)

t=l
which defines the resistance of the sixth section of the
group m in terms of the sum of the first five sections
and the measured differences.
The total resistance of group m is the sum of the
resistances of the six sections that comprise the
group. Its value is found by adding (22) and (26)
so that
*=.i
(.-=r,
R,a = i: Rm.*+Rm.t=lORmA+2Sam 2 d,k+ S><lm, ,
(27)

Subtracting (19) from (20) gives

H.k-Itm.i = Sam[Dm.k-Dm.l]
7 Strictly speaking, as emphasized elsewhere in the paper. (Dm.\Dt) and
(l'm.t lh) are the differencei in proportional parts as deduced (rom the readings
of the DRRS.
315-411 O - 68 - 15

Remembering that the total resistance of group m

is also the first section of group (m+1) and noting
that
S=10"-,i?1.,
and
SM=5X10--1i?l.1

215-12

we have from (27)

Rm+l. , = 1 0Rm. ,+ (2X 10

,) YJ dm.t
k=l
+ (5Xl(T-1R1,1)dm,l.

Any given ratio of the volt box is defined as the ratio

of the sum of all resistances up to the kth tap of
up to to the resistance of the first section of the
t group. Thus, by definition, any ratio of the
(28)

Equation (28) serves as the connecting link between

one group and the next. If it is applied successively
to increasing values of to, beginning with m=l,
one finds that the resistance of the first section of
any group to can be expressed in terms of the re
sistance of the first section of the first group and the
measured differences by the general equation
Rm.,= 10 'J?, ,+(2X10*#,.,)
[t-6
*-5
2</i.*+S*.*-
t=l
*=1
+ (5X10

t-5

~|

k-l

volt box is jjr^ so that, since (fcXlO ') is the

i. i
nominal ratio, eq (31) can be written in the form

where N'm_t are the nominal ratios and nm k are the

corresponding corrections in proportional parts, com
puted from the difference terms that appear in eq
(31). The corrections can be stated in ppm by
multiplying by 10".

10. References

+<*,.,. . . +*.-!.,]. (29)

Equation (21) states that the resistance of the kth

section of group to is
R*.*=Rm.i+S,*dm,l=Rm,i+(.l(r-lRl.l)dm,k
so that the sum of all resistances up to and including
the kth section of group m is
2 s*S-.*=*fl.o+(io ,.,) 24.-.

[1] Francis B. Silsbee and Francis J. Gross, Testing and per

formance of volt boxes, J. Research NBS 27, 269 (1941 1
RP1419.
[2] Paul P. B. Brooks, Calibration procedures for directcurrent resistance apparatus, NBS Mono. 39 (1962).
[3] James L. Thomas, Precision resistors and their measure
ment, NBS Circ. 470 (1948).
[4] Forest K. Harris, Electrical measurements (John Wiley &
Sons, Inc., New York, N.Y., 1952).

(30)

Substituting (29) in (30) and factoring out /tX10 1

, gives
au*=*xio"-,jj,.,

(Paper 67C1-114)

216-13

Human

Engineering

for

Comparison

The

P. H. Lowrie, Jr.
NBS Boulder Laboratories
Boulder, Colo.
The principles of human engineering are as applicable
to making precise measurements as they are to the
production line. Too often the human element is not
given adequate consideration in the design of precise
measuring equipment and as a result, accuracy de
clines. To reduce this source of error, a calibration
console was designed in which the human engineering
factors were given the same consideration as the
technical requirements. In this system the console
operator is considered to be a decision maker, and
those functions not requiring judgment are automati
cally processed by the console. Most of the calcula
tions are performed automatically by internal circuits,
and the results are displayed digitally upon command.

OFTEN, in the zeal to produce a measurement de

vice that is technically superior, the designer does
not place enough emphasis on the fact that the de
vice must be operated by a human. As the goal of
technical perfection comes closer to attainment, the
role of the operator as an uncertainty factor be
comes more and more apparent. Certainly a technical
ly perfect device that is incompatible with the op
erator is less than desirable.
Prior to the design of a new console for the meas
urement of volt boxes at the NBS Boulder Labora
tories, the relationship between the operator and the
equipment was studied. The conclusions resulting
from this study led to the extensive use of the prin
ciples of human engineering in the design of the
console. At no point, however, were the technical
requirements subordinated to the human-engineering
requirements. The result is a man-machine system
that introduces measurement uncertainties of less
than 10 parts per million (ppm) while allowing meas
urements to be made with greater efficiency than was
previously possible and allowing a more extensive
study to be made of each measured volt box.
Method of Measurement
The design is based on the method first described
by Silsbee and Gross1 in 1941. In this method, the
ISA JOURNAL -July 1965

of

Console

Volt

Boxes

input terminals of unknown and standard volt boxes

are connected in parallel, and the difference between
the voltages at the corresponding output terminals
are measured. This measurement is usually made
with a Lindeck-Rothe potentiometer.2 The circuit
is as shown in Fig. 1.
To correct for voltage drop in the leads between
the corresponding input terminals, these voltages
are also measured. The resulting equation for the
corrected ratio, Xr, can be readily derived by the
application of Kirchhoff's Voltage Law to the cir
cuit of Fig. 1. A derivation of this equation has
been published by Dunfee.3 If the voltages v,, v2,
and v3, are defined as positive when of the same
polarity as v4, the correction equation is

--^-)"-]

en

In equation 1
X is the nominal ratio of the unknown volt box.
u, is the previously determined correction to the
ratio of the standard volt box, in ppm.
v, and vs are the voltages between corresponding
output terminals in microvolts.
v, and
are the voltages between corresponding
input terminals, in microvolts.
S is the nominal ratio of the standard volt box
and is equal to Xn.
V is the applied voltage, in volts.
Human Factors Influencing the Design
Before starting the design, it was necessary to
determine what the design should promote and what
it should prevent, with regard to its effect on the
operator. A short study into the matter resulted in
the following conclusions:
When an operator becomes tired or bored, he un
consciously becomes less careful.
He is much less likely to misread a digital display
than he is an analog display.
He is less reliable than a machine in the perform
ance of repetitive calculations. As Sinaiko and Buck
ley put it, he "is not very effective in performing
routine calculations. He is rather slow and likely to
make errors."4

217-67

 Irritating situations, such as the necessity of fit

ting many leads under one terminal post, produce
a state of mind that is not conducive to highest ac
curacy.
LindeckRothe
potentiometer
V

Figure I. Basic voltbox measurement circuit.

Figure 2. Console at the NBS Boulder Laboratories

Figure 3. Main control panel

Figure 4. Output display panel

Design Requirements
In the design of the console, human engineering
aspects were subordinated only in those instances in
which they would jeopardize the technical quality of
the console. The design objectives included the fol
lowing:
1. The operator should be able to change ranges
on the standard volt box quickly and easily. This
not only increases efficiency and reduces irritation,
but also minimizes the time during which the boxes
are cooling. This is important since the boxes are
measured under temperature equilibrium conditions.
2. Controls should be centralized. Operations that
must be done simultaneously should be performed
by the movement of a single control.
3. Digital presentation should be used both for
controls and displays.
4. As much of the computing as practicable should
be done automatically. This is important in the
measurement of volt boxes, since corrections may
have to be calculated by means of Equation 1 as many
as 32 times for an ordinary volt box and as many as
96 times for a master volt box. Though the math
ematics is simple, the repetition is tedious. At best,
it will lower the morale of the calculator and at worst
it will result in the increasing occurrence of errors
toward the end of the calculations.
5. Recopying of data should be reduced to a min
imum.
6. An automatic record should be kept of control
settings and displayed values to provide a check
against gross error.
7. Adequate safety measures must be provided be
cause d-c voltages as high as 1,500 volts may be
present during a measurement.
Additionally, there was the overlying technical
requirement that the measurement uncertainty in
troduced by the console should be held to a few ppm.
Among other things, this necessitated that guarded
circuitry be used throughout to minimize leakage
currents, and that a reversing technique be used
to minimize the effects of thermoelectromotive forces.
Features
In the console shown in Fig. 2, most of the de
sign objectives were realized. The design features
of the console include:
1. A specially designed plug-panel which allows
the operator to change ratios on the standard quickly
and easily. The panel is completely guarded and uses
heavily gold-plated connectors.
2. All controls used during a measurement are
grouped in front, or slightly to the side, of the
operator. No control is more than eighteen inches
away. Those operations that must be done at the
same time are performed by the movement of a single
control. This was accomplished by the extensive use
of relays. The operator thus needs to move very
little during the measurement of any range. However,
he must get up and move around the end of the con
sole to change ranges. This change of pace helps
to reduce fatigue.
3. Wherever possible displays and controls are in
digital form. Some examples are shown in Fig. 3.
4. Console circuits reduce the measured voltages
to parts per million of the ratio of the unknown
218-68

Parameters

Set-volts
Enter
Sn

Enter
and
set
V

Digit
Decimal
point
- -H

Digitally
displayed
results
Results
sign
Sign
adjust

Digital

Multiplication and division circuit*

Switching
circuit
Digits
Decimal
point

Results

Measuring
potentiometer
Measure
selected
voltage

}ecimol
-lultiplier

Operator
line

X
voit
box
s
volt
box

Electrical
power
Electrical

Decimal point positioning circuits

decimal point

information

Figure 5. Block diagram of measurement system.

volt box and display them as shown in Fig. 4.
5. The operator records the data on an adding
machine, adding or subtracting as indicated on the
display. At the completion of the measurement, he
then enters correction to the standard, u,, in the
machine and depresses the totaling key. The adding
machine sums the values, and thereby completes the
calculations. When the calibration of a volt box
is complete, the tape from the calculator is re
produced photographically on 8 x 10 "6 sheets which
become the permanent records. Recopying of the
data is thus reduced.
6. A digital printer monitors the controls on the
main control panel and the digital display. A com
parison of this tape with that from the adding ma
chine detects operator error and shows up equipment
malfunctions.
7. Safety measures include a photoelectric fence
that disconnects the high voltage whenever the high
voltage area is entered, an illuminated sign that
is energized whenever the high voltage is applied,
and neon lights connected across the high-voltage
circuit to provide a warning even if all other safety
circuits fail.
The Measurement
The objective of the measurement is, of course,
to determine the parameters and variables of Equa
tion 1 and then to use them to determine the cor
rection. Equation 1 may be rewritten in the form
Xc = X.[1 + fu. + V, + V2 + V, + V^10-] (2)
where V,, V,, V3, and V, are the equivalents of the
terms in v v v3, and vk respectively. If v,, vit v and
vt are in microvolts, and V is in volts, then VlT VIr
V,,, and V, will be in ppm of ratio of the unknown
box. X X, and u, have the same meaning as in
Equation 1. In order for Equation 2 to hold, the
signs of the values of V,, V* V3 and V, must be ap
propriately adjusted.
From Equation 2 it can be seen that, provided the
values of the difference voltages can be automatical

ly reduced to ppm of ratio as they are measured, the

calculations involve only addition and subtraction
and thus lend themselves ideally to performance on
an adding machine.
To assist the operator in making the measurement,
the main function switch indicates what the operator
should do, or what the console is doing, at each
position. This switch, which can be seen in the
lower center of Fig. 3, controls the measurement
sequence and hence is the program control for both
console and operator.
In the performance of a measurement, the operator
first enters the value of S, the nominal ratio of
the standard voltage box, into the console by means
of the S switches shown at the upper right of Fig.
3. These switches are connected to the computing
circuits shown in Figure 5. He then selects the
desired value of V, applied voltage, by adjusting
the four "V=" dials shown in Fig. 3 to display that
value. In so doing he also enters the value into
the computing circuits. When the 0-1500 volt sup
ply has been adjusted to the selected voltage, the
"set V" galvanometer will be at null. This type of
indicator was chosen because, though the eye, in
general, does not notice small changes in the angle
of a needle, it readily detects such changes when
they are from the vertical.
Since the galvanometer indicator is within
the operator's field of vision while he is concen
trating on the measuring circuit galvanometer, shown
immediately above the main control panel in Fig. 2,
he is able to continuously monitor the supply voltage
during the measurement.
After having entered S and V into the console,
the operator rotates the main function switch to
select one of the four difference voltages. He then
adjusts the measuring potentiometer to obtain a null
on the measuring circuit galvanometer. While he is
making this adjustment, the output from the measur
ing potentiometer is continuously being reduced to
ppm of ratio by the computing circuits. See Fig. 5.
Note that he may select any one of a number of

219-69

a voltage by R6 rather than being displayed on a

milliameter. The voltage across R& + R is applied
also to the multiplier-and-divisor resistor chain. This
resistor chain corresponds to the R, and R, of Equa
tion 4. The "hundreds," "tens," and "ones" decades
shown in Fig. 6 are automatically adjusted to be
proportional to V as the set-volts potentiometer is
adjusted to measure V. Consider for example, the
adjustment to the value V = 325. The adjustment
of the hundreds decade to 3 introduces 20 kilohms
into the series resistance circuit. The adjustment
of the tens decade to 2 introduces 2 kilohms, and
the adjustment of the ones decade to 5 introduces
500 ohms. Note that there always is a minimum resist
ance of 10 kilohms in the circuit. With the ad
justments as described, the current through R, is

Multiplication and
division circuits

Measuring
potentiometer

23
Hundreds

To
Lindeck switching
circuit I
shunt
'g

Figure 6. The computing potentiometer

measurement ranges by rotation of the Lindeck Shunt
switch (Fig. 3). The position of this switch is fed
to the decimal point positioning circuit as a decimal
multiplier (Fig. 5). As shown in Fig. 5, the cir
cuits in the console perform the required multiplica
tion and division, position the decimal point, and
adjust the sign of the value so that the digitally
displayed result can be directly recorded on an add
ing machine by the operator.
Computer Principle
Human-engineering considerations suggested the
use of digital controls and displays. However, analog
circuits were preferrable for computing because
of their simplicity in this application. Hence, the
computing circuits were designed to allow digital
entry of the parameters, and digital display of the
results. The calculations, however, are performed
by electrical analogs of the v v v and v4 terms
in Equation 1. As an illustration of how such an
analog was developed, consider the v, term in Equa
tion 1. When this term is compared with the equiv
alent term in Equation 2, the result is
(3)
An electrical analog of Equation 3 is the common
voltage divider circuit:
E = E

(4)

10,000 +20,000 +2,000 +500

E
(5)
32,500
which, considering only significant figures, corre
sponds to dividing by the voltage V = 325.
The voltage E (Fig. 6) is proportional to Ir,xR
hence the result will satisfy Equation 4. It is seen
from Fig. 6 that either R, or R, can be varied with
out affecting the value of the other. Thus the re
quirements previously stated are satisfied. Since at
balance vd is equal to either v v vs, or vit which
ever is being measured, the relationship between
E and vd is
(6)
where M is a multiplier, determined by the ratio
R6/R0, and is equal to some integral power of ten.
A complete analysis of the circuit will be published
shortly.
Decimal Point Positioning Circuit
The use of a digital display reduces the problem
of positioning the decimal point to one of simple
switching. The circuit is shown in Fig. 7. From the
figure it can be seen that switching to a larger
Lindeck element, R, or to a larger ratio, S, will
move the display decimal point to the right, but
moving the "V=" decimal point (Fig. 3) to select a
S, ganged to
Lindeck shunt
switch

D-c volts

S2 ganged to
decimal point
switch on
set volts

where R, is the portion of R, across which E ap

pears. The only requirements for the analogy to hold
are that:
E is proportional to v3,
R, is proportional to V,
Rj is proportional to S, and
R, and R, can be varied independently.
Computing Potentiometer
Fig. 6 shows the computing potentiometer. The
difference voltage, vd, at galvanometer null, is de
veloped across the Lindeck element, R. At this
point the resemblance to a Lindeck potentiometer
ends however, because the current is converted to

Figure 7. Decimal point positioning (simplified).

Z20-70

Present procedure
Enter

Set and j ^jMeosure LIMeasure

IMjMeosure
Measurement
v.
V
I
Depress| Calculations
Record Record Record Record
total complete
v,
key
Previous procedure
.^Meosure^JMeaaureljMeasureHMeasure Compute CcrnputeJJComputeUcomputelJ
VS<MV) Vz(u.V) V,(u.V) m y,(U.v)r'

Record
V

ISA JOURNAL -July 1965

Calculate
JCorrectiot
a*b*c

Record
Record
Record
Record
nnilljamps -nilticmps miHiamps
Imillromps
i
Record
Record
Record Measurement
Lindeck Undeck Umdeck
shunt
shunt
shunt
|
Record
Record
position of positionof position ofj Ipositionof
switch switch
Record
t_indecK
shunt

Figure 8. The procedure used with the console contrasted with the previous procedure.
larger value will move the display decimal point
to the left. Thus the effect of adjusting the Lindeck
element or S switches is to multiply by powers of
ten, and the effect of adjusting the set-volts, "V="
switch is to divide by powers of ten. The result is
automatic placement of the decimal point. In the
actual circuit, provision is included for the fact that
v, and w4 are not multiplied by S.
The Guard Circuit
The console is completely guarded, ar.d the meas
urement circuit is guarded at ground potential. This
allows the guard to act as a shield also. A relay
grounds the side of the power supply connected
to v, (Fig. 1) to ground when v,, v or v, is being
measured, and grounds the side connected to w4 when
V) is being measured. This arrangement allows the
cabinet to be grounded without the need for exten
sive high-voltage insulation.
Problems Encountered
Before active electronic equipment could be added
to the measurement circuit, it was necessary to deter
mine its effect on that circuit. The digital voltmeter
was found to disturb the circuit when sampling. To
remedy this problem, sampling is disabled during the
actual measurement. At the completion of the meas
urement, on command of the operator, the voltage,
E0, is sampled by the digital voltmeter. However, the
interaction of the digital recorder was found to be
even more severe, and it had to be extensively
modified before it was usable.
During the experiments with the console, it was
established that the insulation resistance between
the working and guard circuits must be kept high.
If difference voltages between the two circuits are
only as high as 1.5 volts, resistance values below about
100 megohms can result in excessive leakage cur
rents.
It was found that extremely high transient volt
ages existed under some conditions of reversing the
supply voltage. The use of capacitive-resistance spark
suppression circuits solved this problem. Of course,
the fact that the full supply voltage, 1500 volts, can

be used, makes high voltage wiring techniques neces

sary in some of the circuits.
The J500-volt supply is not adequately stable for
use below about 50 volts, and a chopper-stabilized
supply is therefore included for use up to 100 volts.
The voltage adjusting circuits of this supply were
replaced by a decade resistor of the type found in
the standards laboratory, thereby making the supply
satisfactory for use down to voltages of less than one
volt.
Conclusion
The new console has been in use for more than a
year. Through the use of the console, measurement
uncertainties have been reduced from approximately
25 ppm to less than 10 ppm.* The reduction in
the amount of time and effort required in the calibra
tion of a volt box is diagrammed in Fig. 8. Not
only has the number of steps been significantly re
duced, but the complexity of the steps is also much
less. From information gained during the past months
of console use, we conclude that the judicious ap
plication of the principles of human engineering
can be of significant value in the design of precise
measuring equipment.
This uncertainty figure, which is consistent with experiment,
is the result of the simple addition of the errors that would be
introduced in the worst case If all components differed from
nominal value by the full amount of the manufacturers' stated
tolerance limit and is considered to be the pessimistic limit.
It does not include the uncertainty in the correction to the ratio
of the standard volt box, nor does it Include uncertainties re
sulting from the effects of ambient conditions on the volt box
undergoing measurement.
References
1. Silsbee, F. B., and Gross, F. J., "Testing and Per
formance of Volt Boxes," RP1419, J Research NBS,
Vol. 27, Sept. 1941, pp 269-287.
2. Harris, F. K., Electrical Measurements, John Wiley
and Sons, Inc., New York, 1952, pp 176-177.
3. Dunfee, B. L., "Method for Calibrating a Standard
Volt Box," J. Research NBS, Vol. 67C, No. 1, Jan.-Mar
1963, pp 1-3.
4. Sinaiko, H. W., and Buckley, E. P., "Human Factors
in the Design of Systems," NRL Report 4996, Naval
Research Laboratory, Washington, D.C., Aug. 29, 1957.

221-71

71

A METHOD FOR CALIBRATING VOLT BOXES, WITH ANALYSIS OF VOLT-BOX SELF-HEATING CHARACTERISTICS
Ronald F. Dzluba and T. Michael Souders
Electrical Instruments Section
Electricity Division
National Bureau of Standards
Washington, D. C.
Summary
The direct comparison of a "working" volt box
with a standard volt box of the same nominal ratio
is well known and widely used, since the Guarded
Standard Volt Box has recently become commercially
available to measurements laboratories . An alter
nate technique, the Julie ratiometric method, ^
has been recently described, which permits cali
bration of volt boxes without the need for this
standard. The method described herein, suggested
by F. L. Hermach, is a modification of the Julie
method that has been under study at NBS for well
over a year. It proves to be a useful calibration
method, featuring speed of operation and high ac
curacy while requiring only an inexpensive and
versatile "ratio standard."* The method is limit
ed to the calibration of volt boxes whose selfheating errors are within the desired accuracy
range. However, with the production of volt boxes
designed to decrease the self-heating errors by
decreasing the rated current (higher "ohms per
volt" characteristic) and by using resistance
materials having lower temperature coefficients,
the method could be increasingly useful for cali
brating working volt boxes quickly and easily, to
an accuracy within 10 ppm.
Test Method
The Julie method uses a calibrated Kelvin
Varley voltage divider in a bridge network to
measure successive "step-down ratios"** of a volt
box. (An s-d ratio is defined as a ratio of any
volt box range to its immediate lower range; for.
example, 3/1.5, 7.5/3, 15/7.5, . . . 1500/750.)
Bridge balance Is achieved by adjusting the divi
der for a null on the detector; the s-d ratio of
the volt box is then computed from the corrected
divider setting, applying its corrections as
measured in a separate test. Any volt box ratio,
referred to its lowest section, Is obtained by
multiplying all the s-d ratios up to and including
the first one containing the desired voltage range.
If the s-d ratio corrections are small, the cor
rection for the volt box ratio is simply the alge
braic sum of the individual s-d ratio corrections.
*Not to be confused with a Guarded Standard Volt
Box. Standard in the sense that it is used to
calibrate a test instrument.
**Step-down ratio is the inverse of step ratio,
the notation used by Julie. The "step-down" ratio
lends itself better to the derivation of voltage
ratios .

The modification introduced at NBS simplifies

the measurement procedure, preserves the precision
needed for the s-d ratio measurements, and permits
measurements at rated voltage. A guarded ratio
standard, see figure 1, comprised of a string of
thirteen nominally equal, fixed resistors and a
three-dial resistance box, is used directly to
measure the s-d ratio of a volt box. Suitable
combinations of the thirteen 100 kw resistors pro
vide the ratios required to measure s-d ratios
from 2/1 to 20/1.* The ratio resistors in the NBS
assembly are connected to guarded binding posts
mounted on the top panel. The decade resistor,
which includes the 10-, 1- , and 0.1-ohm dials of a
resistance box, provides the means for balancing
the bridge circuit when performing either a selfcalibration of the ratio standard or a calibration
of a working volt box. A resolution of 1 ppm is
possible when performing either test. This is not
always feasible when a Kelvin Varley divider is
used to measure large s-d ratios.
Calibration Procedure
Ratio Standard
The NBS ratio standard is evaluated, follow
ing Julie's method, by using three resistors as
fixed arms of a square bridge to measure the re
maining ten in succession (see figure 2). By
using a different set of three resistors for the
fixed arms, the deviations of all thirteen resis
tors can be referred to one reference resistor.
The resistance decade is used to balance the bridge,
and is located either in the unknown arm or one
fixed arm as necessary, depending on the individual
corrections of the resistors. A high resistance
detector is necessary, but its impedance does not
limit the accuracy, as it might if it were used as
a deviation indicator rather than a null device.
The guard resistors eliminate errors from leakage
resistance of the d-c source to ground, and enable
the detector to be at ground potential. The decade
dial setting becomes direct reading in ppm when
multiplied by 10. From the data, ratios of the
standard can be computed. A ratio of the standard
is defined as the resistance of a series-connected
string of k resistors (high side) divided by the
resistance from the low voltage end to any tap
point along the string consisting of n resistor's
(low side). As shown in Appendix A, if &i is the de
viation in ppm of the i-^k. resistor in the string
from a reference resistor, the expression for com
puting ratios of the standard is:
*The 20/1 ratio is needed to calibrate a "s-d ratio"
of 3/0.15, the largest "s-d ratio" encountered at
NBS. All thirteen resistors are needed to provide
this ratio.

222-17

At
5>i
S - S'(l + 8)* fi + i=ig- ,
" L
106k
106nJ

(1)

where s Is the correction to the nominal ratio,

S , of the standard. Higher order terms are
neglected. It should be noted that k>n and the n
resistors In the low-side are a part of the k
resistors in the high-side.

The guard circuit increases the effective

leakage resistance that shunts each 100 kQ coll to
greater than 10" ohms. Consequently, the ratio
error caused by leakage is much less than 1 ppm.

Volt Box
A calibrated ratio of the ratio standard is
then compared to an s-d ratio of a volt box (see
figure 3). If dfc is the decade resistance reading
expressed in ppm of 100 kfi when the detector is
located at position a in figure 3 and
is the
decade reading in ppm when the detector is located
at position b, then, as shown in Appendix A, the
expression for computing an s-d ratio of a volt
box is :
d,+d
d
M - M'(l + m)s'fl +
S_ + s , (2)
L
106k 106n
J
where m is the correction to the nominal s-d ratio,
M . Higher order terms are neglected. (Note that
either dk=0 or dn= 0, depending on the connection
to the decade resistor.) Generally only three
different ratios of the ratio standard are requir
ed to calibrate a working volt box.
A voltage ratio of the volt box is obtained
by multiplying all the s-d ratios up to and in
cluding the first one containing the desired
voltage range. The expression for computing the
voltage ratio is:
J
Hf- (l+up^tM^-M^M^ 'l^Mj][l +Vmt] , (3)
where p, is the correction to the nominal voltage
ratio, Ny Higher order terms and the limitations
of detector sensitivity are considered in Appendi
ces A and B.
Evaluation of the Ratio Standard
When the resistors of the standard are con
nected in a ratio configuration to measure a volt
box s-d ratio, they experience different condi
tions than when measured individually in the intercomparison circuit. They are subject to ratio
errors caused by self-heating, leakage, ambient
conditions and high voltage effects.
The resistors used in constructing the ratio
standard are hermetically sealed, Evanohm resis
tors, each having a temperature coefficient not
exceeding 5 ppm/C and a self-heating coefficient
not exceeding 10 ppm/watt. The worst case en
countered at NBS was in measuring the 1500/750
ratio at rated voltage, where each ratio resistor
was required to dissipate nearly 0.3 watt. At
this power level, tests indicate that the maximum

ratio error caused by self-heating is within 3 ppm,

if the standard is calibrated at low voltages. If
each resistor is calibrated at the same power level
used in the test of a volt box, the uncertainty of
ratio caused by self-heating is within 1 ppm.

Ratio errors, resulting from the effects of

high voltage applied to the ratio standard, were
determined by a method similar to the one recom
mended by Julie. 3 Leakage currents were measured
when a d-c voltage of 1000 volts in series with a
vacuum tube electrometer was applied between the
metal case and the series-connected resistors of
the ratio standard with the guard resistors not
connected. The leakage currents from the resis
tors to the case of the standard were thus measur
ed by the electrometer which had an input resis
tance greater than lO1^. The measured currents
were less than 10" *2 A. The resulting ratio error
due to high voltage effects would be negligible.
Rod and Lead Resistance Errors
A volt box is either a three or four-terminal
network depending on the number of connecting
posts it has at its low end. Julie has consider
ed the effects of rod* resistances on ratio error
for the three-terminal type.4 It is a calibration
error inherent In the "bootstrap" procedure of
the method. Each s-d ratio measurement necessarily
includes the effect of the rod resistance at its
high voltage tap. The s-d ratios along with their
rod resistance terms are multiplied to give a
voltage ratio. However, the voltage ratio of a
volt box includes only the rod resistance at its
highest voltage tap, not those associated with the
Intermediate, unused taps. Thus, the calibration
error caused by the rod resistances can become ac
cumulative, increasing for higher voltage ratios.
This error can be reduced by a proper connection
procedure and can be evaluated as discussed by
Julie.5
The calibration error of a four- terminal volt
box Includes not only the effects of rod resis
tances at the high voltage taps, but also the rod
resistances at the low input and output terminals.
An analysis of these errors and a test procedure
for determining their magnitude is given In Ap
pendix C. The error term is the ratio of rod re
sistance to low-side resistance. The worst case
encountered at NBS is the 3/1.5 ratio of a 200
"ohms per volt" volt box. For this situation the
error was within 3 ppm.
*Rod resistance is the resistance between the ex
ternal binding post and the internal tap-point at
the resistance string of the volt box. In a stan
dard volt box it is a rod, but in a working volt
box it is usually a wire lead.

223-18

The leads connecting the test volt box with

the ratio standard are in the standard circuit
(see figure 3). The maximum error occurs for the
2/1 ratio measurement. Since all lead resistances
are less than 0.01 ohm, the resulting error is
within 0.1 ppm.
ggsulti
The validity of this method was verified by
comparing the results with those obtained by dir
ect comparison with a calibrated guarded standard
volt box. At low voltages (negligible selfheating), the agreement was within 10 ppm. This
also held true for rated voltage calibrations with
volt boxes having negligible self-heating errors.
However, rated voltage calibrations differed more
than the inherent accuracy of the method with volt
boxes having self-heating errors greater than 10
ppm (see Table I). A further Investigation indi
cated that the method did not simulate the manner
in which a volt box is normally operated, as did
the direct comparison method. To resolve these
differences, a study of volt box self-heating,
then in progress, was accelerated.
Volt Box Self-Heating
The self-heating errors of over fifty working
volt boxes have been observed at NBS during the
past two years. Their characteristic curves re
semble those shown in figure 4. The magnitude of
ratio change and the time required for a state of
"ratio equilibrium"* to be established varies with
volt box construction and on operating and ambi
ent conditions. Representative values for various
volt boxes are given in Table II. Over 80 percent
of the volt boxes tested were of manganin con
struction and had self-heating curves on the higher
voltage ranges represented by c in figure 4.
To determine why some heating curves went
through a maximum value; resistance and tempera
ture measurements were performed on the high and
low sections of several volt boxes. Results of
some typical measurements are given in figures 5
and 6. The curves Indicate that the heating of
the high section is significant, with temperature
changes approaching 12C. The major portion of
this resistance heating and the major changes in
ratio occur within 10 minutes after the volt box
is energized. In contrast, the resistance heating
of the low section is small. Eventually, the low
side temperature Increases by external heating
caused by the transfer of heat from the high side.
This effect, called "proximity heating," has a
long time constant because of poor air circulation
within the box. When ratio equilibrium is estab
lished, the high side, low side, and box tempera
tures are constant.
*Some volt boxes, which have large self-heating
errors, do not come to a true state of equilibrium
after a reasonable warm-up period. When the ratio
changes less than 2 ppm/15 minutes, a state of
"ratio equilibrium" is said to exist.

A peak can occur as a result of these two heat

ing effects and as a consequence of the resistancetemperature (R-T) properties of manganin wire. Fig
ure 7 depicts a typical R-T curve for manganin, di
vided into several zones. The maximum, zone B, of
this curve is in the neighborhood of room tempera
ture, being generally between 20* and 50C, while
the minimum, zone D, is at 350C.6 The majority of
the commercial volt boxes calibrated at NBS were
constructed of manganin wire whose maximum was spec
ified to occur around 28C. Figure 5 indicates that
the resistance of the high side, initially in zone A
on the R-T curve, passes through zone B into zone C.
The low side resistance, which increases slightly it
always in zone A. Since the voltage ratio is di
rectly proportional to a resistance increase in the
high side and inversely proportional to a resistance
increase in the low side, curve c in figure 4 is the
resultant. The temperature measurements, shown
graphically in figure 6, further substantiate the
existence of this peak value.
The other self-heating curves of working volt
boxes constructed of manganin can be explained in
a similar fashion. Curve a in figure 4 results
when the resistance material is initially in zone
A and never leaves this zone. Curve d, the mirror
image of a, is observed when the resistance mater
ial is initially in zones B or C and never leaves
zone C during a heat run. In any practical con
struction the manganin will not heat sufficiently
to pass Into zone D. Curve b has been observed for
one volt box and the reason for its shape is not
certain. A reasonable explanation is that initial
ly Its resistance material is located In zones B
or C. During the heat run, the high side reaches
temperature equilibrium quickly, then follows the
proximity heating effect of the low side causing
the reversal of the volt box ratio.
Working volt boxes constructed of Evanohm or
of a similar alloy have self-heating curves repre
sented by a or d in figure 4. This results from
the linear temperature coefficient of Evanohm.
Generally, the magnitude of the self-heating error
is less than that for manganin volt boxes of the
same ohms-per-volt characteristics.
The general procedure for calibrating a work
ing volt box at NBS by the direct comparison method
is to observe the ratio at rated voltage on the
highest voltage range as a function of time and
then proceed to the lower ranges in descending or
der. A moderate amount of time is spent waiting
for ratio equilibrium at these lower ranges, ap
proximately five to fifteen minutes. Tests per
formed in the reverse order (ascending) could dif
fer significantly from the decending order tests
on ranges where self-heating is a problem. This is
observed with working volt boxes having no ventila
ting ports. The slow rate of heat transfer in these
boxes accounts for these differences. To limit this
uncertainty, calibration procedures should be
specified.
A further complication to the self-heating
problem is the effect of ambient conditions. Tem
perature, humidity, and air circulation, modulate

224-19

the magnitude of the error and the time for ratio

equilibrium to be established. For high accura
cies In volt box measurements, these conditions
should be known during calibration and duplicated
during use.
The above analysis pertains to working volt
boxes but can also be applied to Guarded Standard
Volt Boxes. Some standards can have significant
self-heating errors as large as 30 ppm. Some of
the heating effect within a standard can be at
tributed to the self-heating of the guard resis
tors. Their curves usually resemble a or d In
figure 4.

resistors (high-side) to any tap point along the

string consisting of n resistors (low-side). The
ratio equals:
k
2>i
S - S'(l+s) -

, where k>n.

(4)

The R^'s can be expressed in terms of a refer

ence resistor, R^, and ppm deviations from R, A^'s,
as :

Conclus Ion
R 1 - R 1 (1 + -^t)
106

The accuracy of a volt box calibration Is

limited by the uncertainty of the self-heating
error of the volt box. To reduce this uncer
tainty test conditions and procedures should be
strictly specified and then duplicated by the user.
This is not always practical or desirable. As a
result, the tendency on the part of the manufac
turer has been to decrease the rated power dissi
pation of a volt box to decrease this self-heating
error.
Consequently, working volt boxes having
negligible heating errors can be calibrated by
methods that do not necessarily duplicate the
conditions of use. The technique described herein
serves as such a useful calibration method. Work
ing volt boxes can be calibrated to accuracies
within 10 ppm of ratio by this method, which re
quires only an inexpensive and versatile standard.

Therefore, the ratio S becomes:

k

s<(1+8)

is
^

3.

Julie, op. clt.

4.

Ibid.

5.

Ibid.

106k

106n

An s-d ratio of a volt box Is compared to a

standard ratio S. To balance the bridge network,
a decade resistor is Inserted at point a or point
b of the standard (see figure 3). If this decade
resistance is expressed as r^ when the detector Is
at point a in figure 3, and rn when the detector Is
at point b, then the equation of an s-d ratio is:

M - M'(l-hn)

Julie, op. clt.

The resistor deviations, A^'s, are less than

10 ppm. The error introduced when neglecting
higher order terms is less than 10" 1".

6. Harris, Forest K., Electrical Measurements ,

John Wiley & Sons, Inc., New York, New York, 1952.
7.

(6)

*i

^1

2. Julie, Loebe, "A Ratlometrlc Method for Precise

Calibration of Volt Boxes," Julie Research Labora
tories, Inc., 1964. A similar technique has been
used for several years in the Standards Laboratory
of the Scintilla Division of the Bendlx Corporation.

:lA

Expanding the denominator by the binomial

theorem, S becomes:

Re ferences
1. Dunfee, Bernadine, "Method for Calibrating a
Standard Volt Box," Journal of Research, National
Bureau of Standards, C, Engineering and Instru
ments, Vol. 67C, No. 1, Jan. -Mar., 1963.

(5)

R. + r,k + rn
1=1 i

(8)

>R.
1=1 1 + r n
Substituting in equation 5 gives:
k

Appendix A

kR, + R, = r + r. + r
1
1 106
k
"

Derivation and Evaluation of Calibration Equations

As mentioned in the text, s ratio of the stan
dard is defined as a series-connected string of k

(9)

M'(l+m) nR.1 + R,1 7

1Q6 + r n

225-20

Factoring out kR^ and nR^, expanding by the

binomial theorem and using equation 7, M becomes:
r, +r
M'(W-)*8'[l+-^-a-^+.+

The worst case would be for the highest ratio,

where J might be as large as 10. For this case,
even if (fa - 100 ppm the error introduced by neglec
ting higher order terms is less than 0.5 ppm.

(10)
Appendix B

where S - k/n. If r^ and rn are expressed In ppm

of R^ (100 kfl) as d^ and dn respectively, M then
becomes :
d, +d
M'(l+m)wS'[l + _k n
(11)
+ I
106k 106n
where, as mentioned in the text, d^sO or dn5 0,
depending on the connection of the decade resistor.
The s-d ratio corrections for a volt box are
usually within 100 ppm. An error less than 10~8
is introduced when nelgecting higher order terms of
this equation.
The s-d ratios can also be expressed in terms
of the resistance sections of a volt box. If yf
refers to the number of sections, the general
term for M would be:

Evaluation of Errors Associated with Sensitivity

Limit
The accumulation of errors in the voltage ratio
equation that are due to the limit of sensitivity
of the detector in the determination of each s-d
ratio can be significant. If each step-down ratio
has an error of a; as a result of the sensitivity
limit, the voltage ratio equation can be expressed
as :
1-2
The total error could apprxlmate (1-1)0^ ,
where cfa is the maximum resolution error for a s-d
ratio measurement. However, a more realistic
evaluation of the total error would be to combine
the independent errors by the square root of the
sum of the squares, as they are as likely to com
pensate for one another as to add. The total er
ror would be:

51
1-1 K1

(19)

- Mf'(l+mf)

(12)
Assuming a a, of 1 ppn>, the total error when
J - 10 would be 3 ppm.

i-i
where X > 2.

Appendix C

The voltage ratio of a volt box Is obtained

by multiplying all the My's up to and including
the first one containing the desired voltage range.
The expressions for computing this ratio are:
n, - ty'<i+uv) - ly*^

M*-rMf

(13)

Nj'(l+uv) - M2(l+m2)-M^(l+m3)- -M^l+m^)

;o*^>-[h^

(14)

Evaluation of Errors Associated with Rod Resistors

A four-terminal volt box in an operational
setup is shown in figure 8. Rod resistance rQ
is part of the high side resistance only. Rod
resistance r0 and r^ are in the potentiometer cir
cuit and consequently are not significant in the
voltage ratio of the volt box. The voltage ratio
is expressed as:

H>;][i+fm1+ . . ](15)
1=2

Equation 15 is identical to equation 3 in the

text. To analyze the higher order terms, assume a
maximum s-d ratio correction 6m, in equation 14.
Therefore equation 15 becomes:
N?'(X+u,) -[M^- M;/M;Jl+6j"

(16)

Expanding the correction term by using the

binomial theorem,
becomes:
tf-l)(f-2)6_
Nj^M^-M^- Mj][l+(f-l)6|j+21

. .](17)

0 +i\
i-l

(20)

where f> 2.
There are four possible ways to connect leads
to the low end of a four- terminal volt box when
calibrating an s-d ratio. The lead connection employed at DBS is sh own in figure 3 and is the best
arrangement for sup resslng the ratio error caused
by rod resistances r^ and rQ, if the resistance of
each unit of the ra tio standard Is greater than
that of the lowest section of the volt box. The
s-d ratio for this circuit arrangement is:

226-21

Table I
A.
M, -

(21)

A voltage ratio expressed in terms of the s-d

ratios Is:
(22)
After substituting, rearranging and then divi
ding by equation 20, the voltage ratio for the
connections of figure 8 can be expressed as:
,

I'1
(23)

i-i 1

200 "ohms per volt" volt box

Voltage Ratio Correction in ppm

At 20% Rated Voltage At 1007. Rated Voltage
Direct
S-D
Direct
S-D
Voltage Comparison Ratio
Comparison Ratio
Range
Method
Method
Method
Method
750
300
150
75
30
15
7.5
3

- 65
+ 11
+ 30
- 3
+100
+141
+100
+ 9

-140
- 42
- 10
- 28
+110
+144
+ 98
+ 9

- 58
+ 2
+ 22
- 13
+ 98
+143
+ 99
+ 7

- 99
- 12
+ 15
- 19
+118
+148
+ 98
+ 8

z i-i 1

where/ > 2.
The error term,
(24)

is maximum for 1-1; and is greatest for the lowest

"ohms per volt" volt boxes. The error was measured
by using the bridge circuit shown in figure 9,
where d is the three-dial resistance decade read
ing and R a 100 kfl resistor. The error term can
be computed from the following expression:
(25)

B. 1000 "ohms per volt" volt box

500
- 2
- 3
300
- 3
200
- 4
100
- 10
50
- 23
30
20
- 22
- 18
10
- 7
s
+ 1
3
+ 2
2

+
+

5
6
5
5
13
25
24
19
8
1
2

Table II

Lead resistance L, which is located In the

decade arm, must be sufficiently small or known
so as not to introduce an added error.

Ratio Correction in ppm

Referred to

The term,
(26)
k-2
i=l
is the accumulative error caused by the rod re
sistances at the high end. It is identical with
the calibration error associated with a threeterminal volt box as discussed and evaluated by
Julie.
It can be suppressed by placing the rod
resistances in the ratio standard circuit. Since
rod resistances are usually less than 0.01 ohm,
the error introduced by this term is negligible.

Volt Box
Voltage
ohms /volt
Range
200
750
750
750*
300*
750

1500**
750**
750

At
Peak
+45
+48
+59
+44

At
Equilibrium
-190
-104
- 67
+ 12

+57
+50
+18

+ 8
+ 18
- 8

* Same Volt Box

** Same Volt Box
Two hour heat run for 200 ohms /volt volt boxes
One hour heat run for 750 ohms/volt volt boxes

227-22

TENS

i
pW/VJ

i
a-W^

O^VVN^

o-wvo
Re

o-vwo
R?

t<wv
Re

o^/vVo
R

vw
Rio

Rc

nS^HI

Ga-Hi

UNITS

'/ R,
(WW-O

R2
OA-WO

Rs
OA-WO

R4
OA>WO

R9
IWM

LvWrl

LwNrl

iwvl

Lvwl

Lwvl

TENTHS
Shield
DECADE
RESISTOR
Figure 1

R' - RATIO RESISTORS

r's -GUARD RESISTORS
Layout Diagram of Ratio Standard

Rq i "biRc ~ pixed Bridge Arms

R|, R2...R|q- Rotio Resistors
r's - Guard Resistors
d - Decade Resistor
Figure 2

Intercomparison of Ratio Resistors

k
Standard Ratio - S

228-23

1 + i-1
10 k

Mil
10 n

RATIO STANDARD
r

0.

75 15 30

75

150

300+

'
0 J|*5

750L^^"00
J
VA * W1

WORKING VOLT BOX

^1
Figure 3
S-D Ratio Measur
,n
S-D Ratio - M'(l+m)sTl +
L
106k

S_ + ,1
106n
J

Voltage Ratio - n/(1+u^) [Mj'Mj. ^'M('Jl + V^]

1-2

E
*50-i

Figure 4

Self-Heating Curves

Figure 5

229-24

Resistance -Time Curve of a Volt Box

Reprinted from:
1966 IEEE INTERNATIONAL CONVENTION RECORD
Part 10

230-25

U. S. Department of Commerce

National Bureau of Standards

RESEARCH PAPER RP1323

Part of Journal of Research of the National Bureau of Standards, Volume 25,
August 1940

METHODS, APPARATUS, AND PROCEDURES FOR THE

COMPARISON OF PRECISION STANDARD RESISTORS
By Frank Wenner
ABSTRACT
Reference is made to some of the more important contributions that have
been made to the subject of precise measurements of electrical resistance. The
sensitivity of bridges when used with the modern high-sensitivity moving-coil
galvanometer is discussed rather fully. Special consideration is given to the meth
ods and apparatus used and to the procedures followed in the National Bureau
of Standards in those comparisons in which the precision desired is of the order of
1 part in a million.
The more important factors limiting the precision of the comparisons, such as
load coefficients, terminals, and contacts, thermoelectromotive forces, insulation,
and the optical system of the galvanometer, are discussed rather fully. A meth
od of analysis of networks containing both linear and nonlinear four-terminal con
ductors is given, and the theoretical basis for the experimental procedure used in
determining the effect of slight defects in the insulation is pointed out. This is
followed by a brief discussion of Ohm's law from the standpoint of precise resist
ance measurements and by a brief discussion -of units of resistance. Finally,
reference is made to more than 100 publications having a more or less direct bear
ing on the subject of resistance comparisons.

Research Paper 133

NBS Journal of Research
Volume 4, January 1930
NOTES ON THE DESIGN OF 4-TERMINAL RESISTANCE
STANDARDS FOR ALTERNATING CURRENTS
By Francis B. Silsbee
ABSTRACT
The design of a resistance standard for use with alternating current involves
not only the consideration of the value, definiteness, and permanence of the
resistance and the adequate cooling of the metal parts, which are encountered
in direct-current standards as well, but also the consideration of the inductance
and of the possible change in the resistance of the standard with the frequency
of the current flowing in it. These notes give the theoretical basis for the com
putation of the inductance of resistance standards which consist of a system
of straight conductors long in comparison to their diameters. Most standards
used in the measurement of large currents are of this type. Formulas are
fiven for the skin effect in various combinations of flat strips and coaxial tubes,
lethods of attaching and locating the potential leads so as to minimize the
possibility of error from stray magnetic fields and yet permit of convenient
adjustment of the resistance are discussed. As examples of the principles
here set forth, detailed descriptions are given of two groups of resistance standards
which have been constructed at the Bureau of Standards for testing current
transformers. These standards range in resistance from 0.05 to 0.0002 ohm
and in current capacity from 10 to 2,500 amperes.

315-411 O - 66 - 16

Z31-1

Journal of Research of the National Bureau of Standards

Vol. 50, No. 3, March 1953

Research Paper 2402

Measurement of Multimegohm Resistors

Arnold H. Scott
The method by which multimegohm resistors are measured at the National Bureau of
Standards is a null method using an electrometer as the null detector. The charge flowing
through the resistor during the time of measurement is obtained from a variable air capacitor
maintained at a fixed potential. The potentials across the variable air capacitor and thus
across the specimen are maintained constant as indicated by the null reading of the electrom
eter by decreasing the capacitance of the air capacitor at just the right rate. The capacitance
of the capacitor is varied by a small direct-current motor geared to the shaft of the capacitor
and whose speed can be controlled. Several multimegohm resistors of two different makes
have been studied over a period of about three years. Although these are the most stable
multimegohm resistors available, it was found that they had erratic fluctuations of 0.5 to 1
percent and were generally voltage sensitive. With the impressed voltage varied from 1.5
to 180 volts, various resistors showed resistance changes ranging from 0.4 to 26.9 percent.

N3S Monograph 39
March 1962

Calibration Procedures for Direct- Current Resistance Apparatus

Paul P. B. Brooks
This monograph was written in response to a demand for information
on methods used at the National Bureau of Standards for the calibration
of d-c resistance apparatus. The paper was written primarily for use
by laboratory technicians and therefore contains much detail that is
unnecessary for those with scientific training. However, the techniques
described have been found helpful by physicists and engineers concerned
with accurate electrical metrology.
The monograph is divided into three parts. Part 1 entitled
"General Outline of Equipment and Procedures" discusses errors, correc
tions, tolerances, standard resistors, galvanometers, the algebra
of small quantities, the Direct Reading Ratio Set, and the Universal
Ratio Set. Practical examples of the use of these ratio devices are
given. Part 2 given detailed information on the calibration of a
Wheatstone Bridge using the DRRS and techniques described in Part 1.
Part 3 discusses the calibration of general purpose potentiometers
using the URS and gives detailed instructions regarding the calibration
of a specific type of potentiometer.

Chester Peterson
August 24, 1967

232-2

NBS Technical Note 172

March 1963

Practical Methods for Calibration of Potentiometers

David Ramaley
Potentiometer circuitry, particularly as related to
calibration, is discussed with the primary consideration
given to the required circuit measurements.

The more

feasible means of calibrating potentiometers are described

in considerable detail.

Emphasis is placed upon the use

of the Universal Ratio Set as the basic implement for

accomplishing the major portion of potentiometer calibra
tions.

NBS Technical Note 220

August 1964

SOME MODIFICATIONS IN METHODS OF CALIBRATION

OF UNIVERSAL RATIO SETS
David Ramaley

Universal Ratio Sets can be calibrated by a

number of different methods. The well established
methods are very briefly outlined and emphasis is
placed on some more recent developments. The
choice of methods will depend upon available labora
tory equipment and other considerations.

233-3

JOURNAL OF RESEARCH of the National Bureau of Standards C. Engineering and Instrumentation

Vol. 70C, No. 3, July-September 1966
A Versatile Ratio Instrument for the High Ratio
Comparison of Voltage or Resistance
Alfred E. Hess*
Institute for Basic Standards, National Bureau of Standards, Boulder, Colo.
(April 12. 1966)
A 9-dial resistance ratio instrument capable of precise high ratio comparison of voltage or re
sistance (from 1/1 to 107/1) is described. Also described is the modification of a 6-dial universal
ratio set to permit its additional use as a versatile ratio instrument. Paramount to the accuracy of
these high ratio instruments is the carefully adjusted "common point" junction which is briefly
discussed.
Key Words: DRRS (direct reading ratio set); junction; precision measurements: ratio, resistor,
four-terminal; resistance decade: transresistance: URS (universal ratio set): VERI
(versatile ratio instrument).

234-4

Capacitors

and

Inductors

Papers
Improved ten-picofarad fused silica dielectric capacitor, R. D. Cutkosky
and L. H. Lee
Voltage dependence of precision air capacitors, J. Q. Shields
Capacitor calibration by step-up methods, T. L. Zapf
Calibration of inductance standards in the Maxwell-Wien bridge cir^ cuit, T. L. Zapf
Some techniques for measuring small mutual inductances, D. N . Homan.

237
244
254
259
265

Abstracts
New apparatus at the National Bureau of Standards for absolute
capacitance measurement, M. C. McGregor, J. F. Hersh, R. D.
Cutkosky, F. K. Harris, and F. R. Kotter
Active and passive direct-reading ratio sets for the comparison of audio
frequency admittances, R. D. Cutkosky
Four-terminal-pair networks as precision admittance and impedance
standards, R. D. Cutkosky
Variable capacitor calibration with an inductive voltage divider bridge,
T. L. Zapf
A new type of computable inductor, C. H. Page.
235

271
271
272
272
272

JOURNAL OF RESEARCH of the National Bureau of StandardsC. Engineering and Instrumentation

Vol. 69C, No. 3, July-September 1965
Improved Ten-Picofarad Fused Silica Dielectric Capacitor 1
R. D. Cutkosky and L. H. Lee
(March 8, 1965)
Some defects in a set of fused silica dielectric capacitors constructed in 1961 are listed,
and methods for their reduction or elimination are described. The construction of a new
set of capacitors completed in 1964 is described in detail. Preliminary stability data
presented indicate that the typical drift of the 1964 capacitors with respect to their mean
was less than two parts in 107 in live months.

1. Introduction
Capacitors utilizing fused silica as the dielectric
have been under investigation at NBS and else
where [1] 2 for many years. Early indications were
that suitably constructed capacitors of this type
would be at least as stable as air capacitors, and addi
tionally would be sufficiently rugged to withstand
shipment between laboratories.
A set of four fused silica dielectric 10-pF capacitors
was constructed at NBS in 1961 [2]. The capacitance
elements of these capacitors consisted of fused silica
disks 7 cm in diameter and 1 cm thick. Silver elec
trodes were fired on the two faces of the disks and a
third electrode was fired on the cylindrical surface.
The cell in which the element was mounted shielded
the face electrodes from each other in the region
external to the fused silica element, and was con
nected electrically to the cylindrical electrode of the
element. The housing contained a four-terminal
resistance thermometer for temperature measure
ment and was hermetically sealed.
One of the 1961 capacitors was successfully handcarried to several distant laboratories and returned.
These laboratories included XRC (Canada), NPL
(England), and NBS Boulder. The largest change
in capacitance observed upon completion of the round
trip was 0.2 parts in 10e (ppm). A shipment of this
capacitor by commercial carrier was not successful
and resulted in a change in its capacitance of 36
ppm. This change was traced to a sensitivity to
severe mechanical shock.
The 1961 capacitors had a number of undesirable
features, the shock sensitivity being by far the most
important. Further investigation indicated that all
of the capacitors in the set suffered from this flaw,
and that the change in capacitance was caused by
motion of the fused silica element in its cell. The
element was supported in the cell by means of
phosphor bronze springs which permitted motion of
the element along the cell axis. The element could
not return to its equilibrium position after displace' This work was supported in part by the Metrology Engineering Center,
Bureau
of Naval
Weapons,
Pomona.
Calif, references at the end of this paper.
i Figures
m brackets
Indicate
the literature

ment because of frictional forces between the cylin

drical surface of the element and the spring contacts
attached to the cell. This mechanism could not
have caused a capacitance change if the direct elec
trical field between the face electrodes had been
entirely within the fused silica. Unfortunately, the
direct capacitance included a small contribution due
to fields from the back of a face electrode which
entered the fused silica element at the insulation
gap between this face electrode and the cylindrical
shield electrode, and terminated at the opposite
face electrode. Motion of the element within its
cell altered the leakage field and consequently
changed the total direct capacitance.
A second flaw in the 1961 capacitors was a pro
nounced dependence of capacitance upon voltage.
It was necessary to restrict the voltage applied to
these capacitors to 10 V rms to reduce the measure
ment uncertainty from this cause below 0.1 ppm.
Since sufficient sensitivity is available at this voltage
to see one part in 107, the voltage dependence did not
seriously limit measurement accuracy. An investi
gation into the mechanism of the voltage dependence
was made, and it was found that the dependence
could be almost eliminated by increasing the thick
ness of the fired silver electrodes. The evidence
indicates that the existence of islands of silver on
the dielectric surface which are not in good electrical
contact with the body of the electrode is in some way
responsible for the large voltage dependence. A
similar effect has been observed in mica capacitors
with fired or deposited silver electrodes [3].
Measurements of capacitance and dissipation fac
tor as a function of frequency indicated a capacitance
difference from 50 Hz to 20 kHz of about 25 ppm,
with a maximum slope and a maximum dissipation
factor at about 1600 Hz. The shape of the curve
suggested a relaxation mechanism involving polar
impurities in the fused silica. An investigation of
several optical grades of fused silica disclosed that the
use of a grade possessing exceptionally good ultra
violet transmission characteristics resulted in capaci
tors with very much smaller frequency dependencies
and with dissipation factors not greater than 3X10""
throughout the audio-frequency region.
Examination of the published optical characteris
tics for the various grades tested disclosed no correla

237-173

tion between optical absorption and audio-frequency

behavior. It is felt that the presence or absence of
the audio-frequency relaxation phenomenon is an
accidental result of the technique for producing
fused silica. It is possible that changes in manu
facturing technique which have no effect on the
optical properties of the fused silica may greatly
affect the audio-frequency characteristics.
The temperature coefficients of the 19G1 capacitors
were about 14 ppm/C. The large values were be
lieved to be the result of a large temperature coeffi
cient of dielectric constant in fused silica. Resist
ance thermometers were built into the capacitor
housings to eliminate the measurement uncertainties
caused by unknown temperatures.
A set of twelve 10-pF fused silica dielectric capaci
tors was built in 1964. Care was taken to reduce
greatly all of the flaws listed above. The capaci
tance elements were constructed of type II Suprasil,3
and were supported in their cells by means of tightly
fitting polytetrafluoroethylene (PTFE) rings. Care
was taken to keep the gaps between the face elec
trodes and the cylindrical shield electrode as small
as seemed consistent with operation up to 200 V rms.
The voltage dependence of each capacitor was meas
ured before final sealing, and the electrodes were
replated when necessary. The 1964 capacitors are
physically smaller and much lighter than the 1961
capacitors, but were not designed with this as a
principal objective.
The construction of these new capacitors is de
scribed in some detail to allow their reproduction
elsewhere. Some characteristics of the completed
set are then presented.

Figure 1. Cross section of the capacitance standard.

2. Construction
2.1. Housing
A cross section of the capacitance standard is
shown in figure 1. The exterior of the assembly is
constructed entirely of stainless steel. The hermeti
cally sealed region at the bottom containing the
fused silica element is welded shut upon completion,
and after baking, the container is filled with dry N2
through a copper sealoff tube. Electrical connec
tions are made to the capacitor element and to the
four-terminal resistance thermometer by means of
glass-Kovar seals soft soldered to the sealed
compartment.
Electrostatic shields not shown in figure 1 isolate
the two face electrode leads from each other and from
the resistance thermometer leads. The face elec
trode leads pass through stainless steel tubes as
shown to a pair of coaxial connectors located above
the liquid level of a constant-temperature oil bath.
The resistance thermometer leads are attached to
four binding posts located below the liquid level.
Details of the construction can be seen in figure 2.
1 Available from Amersil Quartz Company.

Figure 2. Details of the capacitor housing and superstructure.

The dimensions of the fused silica element and its
mounting are shown in figure 3. The PTFE radial
support is inserted into the slot of the brass cell be
fore the inside diameter of the PTFE is finish
machined. Each fused silica element is individually
fitted to its cell by machining this radial support so

238 174

BRass CELL
062

that the element with its silver electrodes in place

can just be removed from the cell with a rubber
suction cup. The axial (vertical) constraint on the
element is controlled by the use of shims or by
machining, so that an axial interference of about
0.001 in. is obtained. This results in a slight com
pression of the PTFE axial supports.
Electrical connections to the face electrodes on the
fused silica element are made by means of wires
soldered to the phosphor bronze disk springs shown
in figure 3. The disks are cut to the shape of an X
and bent as shown in figure 4. Electrical connection
between the cylindrical electrode and the cell is
made by inserting three 0.030-in. diam phosphor
bronze wires about 2 in. long into the circular gap
between the cell and the element. The ends of
these wires lie in the slot at the sides of the cell with
the PTFE radial support. The three wires are
positioned uniformly around the circumference of
the cell and provide three points of electrical contact.

- 160
- 640
- diom
J
-2
-2 820
840 - diom
^
2 938-diom
-diom
*
- 3562 - diom
PTFE RADIAL SUPPORT

PTFE AXIAL SUPPORT

I- i t2S-diai
L -142
020
1 626- diom
i 875-diam
OiO-iNCH PHOS BRONZE SPRING. I 620-0
FUSED SILICA ELEMENT
-385;gga
005 8F.VEL

2.2. Resistance Thermometer

3
The four-terminal thermometer is located on the
outside of the cell containing the fused silica element,
as shown in figures 1 and 4. The cell is prepared
by cementing a layer of linen cloth over it with
shellac and baking at 100 C. A length of B & S
No. 36 Formvar insulated copper wire with a re
sistance of 25 il is doubled over and wound bifilarly over the linen. More shellac is then applied
and baked. Leads to the current and potential
terminals are connected to the resistance wire with
soft solder. The junctions are tied down securely
and shellac is applied. The completed thermometer
is baked for 24 hr at 75 C.

BRASS COVER
"-094
-1880-diam
-3 b62-diom
Figure 15. Fused silica capaci
tance element and cell dimensions

2.3. Fused Silica Element

FIGURE 4. Partially assembled capacitance element cell.

The fused silica element is ground to the dimensions

shown in figure 3 with conventional toolroom grind
ing equipment. Some care is necessary to prevent
chipping of the edges. After inspection for chips,
the element is cleaned witli alcohol, and then with
soap and water. It is next immersed in an ultrasonic
cleaning bath containing a wetting agent, and rinsed
with water. Finally, it is immersed in an ultrasonic
cleaning water bath and rinsed with distilled water.
After drying at 100 C, Dupont Silver Paint #4666 is
sprayed onto one face and the cylindrical surface,
and allowed to air dry. The other face is then
sprayed and dried, and the silver is fired onto the
element at 4S0 C. It is found that single silver
deposits thicker than 0.0005 in. are often defective,
and that a single application is not sufficient to
prevent a dependence of capacitance upon voltage.
A satisfactory remedy is to buff the electrode surface
with fine steel wool and apply two additional coats
of paint to the electrodes. The preferred final
electrode thickness is between 0.0010 and 0.0015 in.
If the electrode thickness is increased to 0.002 in.,
the large thermal expansion coefficient of the silver
often causes the fused silica element to chip when it is

239-175

cooled from the firing temperature. No noticeable

chipping occurs when the silver thickness is less than
0.0015 in.
After the third electrode coat has been fired, the
edges of the element are very lightly beveled to
separate the cylindrical electrode from the face
electrodes. This is done by hand with a diamond
wheel and a special jig, using water as a lubricant.
Care must be taken at this point to avoid making the
gap excessively wide and to remove all traces of
silver from the beveled portions of the fused silica.
If the diamond wheel is allowed to become con
taminated with silver, a thin layer of silver will
remain on the fused silica, and a large dependence of
capacitance upon voltage will result.
The element is next fitted to its cell by machining
the PTFE radial support. The element is then
baked 24 hr in vacuum at 175 C to eliminate
adsorbed water, placed in its cell, and the cell is
evacuated. The baking and evacuation are es
sential to produce a small dissipation factor and a
stable capacitance.
After the capacitor has reached room temperature,
it is measured and the adjustment necessary to pro
duce exactly 10 pF at 25 C is estimated. Using the
element and cell dimensions shown in figure 3, the
capacitance will probably not depart from 10 pF bv
more than 0.05 percent. An adjustment as small
as this can be made quite easily by cutting a cavity
into either a face or the cylindrical surface of the
element with a small diamond wheel. The cavity is
then sprayed with silver paint and refired. The
adjustment sensitivity depends upon the diamond
wheel diameter and thickness, the depth of cut, and
the position of the cavity on the element. A jig may
be readily set up to regulate depth of cut, and a
calibration of capacitance change versus depth of cut
measured with a trial element. Usually two or three
successive adjustments suffice to produce a capaci
tance of 10 pF 50 ppm. A typical fused silica
element after adjustment is shown in figure 5.
After final adjustment, a test is made for voltage
dependence, and if necessary, the entire element is
replated. The stainless steel housing is then welded
shut, and the system is baked at 65 C with a vacuum
pump attached to the sealoff tube. After cooling,

Figure 5. Completed fused silica element with an adjustment

of minus 50 ppm.
240

the housing is filled with dry N2 at atmospheric pre>sure, and the sealoff tube is pinched off and soldered.
A series of temperature cyclings is made between 50
and 0 C, and the unit is placed into operation.
Figure 6 shows a completed capacitor.

3. Performance
3.1. Capacitance Stability
All measurements of capacitance stability reported
here were made at 1592 Hz using a 10:1 bridge, with
all capacitors in an oil bath maintained at 25.0 C.
The measurements seemed to indicate that these
capacitors were at least as stable as any other
capacitors in our laboratory. Accordingly, the mean
of five of these capacitors, numbers 108, 109, 110, 112,
and 1 13 was taken to be the basis for the comparisons.
No temperature corrections were applied since the
capacitors were at nearly the same temperature and
had nearly identical temperature coefficients. The
application of individual corrections for temperature
slightly reduces the scatter in the measurement series
reported below. Initially, all measurements were
made with 10 V rms applied to the capacitors to al
low ready comparison with the 1961 set of capacitors
which had large voltage dependencies. Bridge
readings were recorded under these conditions to the

nearest part in 107. Beginning November 9, 1964.

the new capacitors were intercompared at 100 V
rins. This allowed considerably greater precision,
but prevented simple comparison with the 1961 fused
silica capacitors.
Capacitance differences between each capacitor in
the set and the mean of the select five are listed in

table 1. The table demonstrates that most of the

capacitors are stable with respect to each other, but
they could possibly all be drifting at the same rate.
Comparisons of the 1964 set of capacitors with the
1961 set indicate no appreciable relative drifts, but
conclusive proof of stability must await the comple
tion of a new calculable capacitor.

Table 1. Helatire stability of eleven fused silica capacitors

Decimal points indicate parts in 10".
Capacitor
Date
B/i/M
8/4
8/17
S/26
9/8
9/11
9/22
10/5
10/20
10/22
10/-J0
11/5
UI9
11/24
11/30
12/7
12/14
12/21
12/28
1/4/65
1/12
1/19
1/28

108

109

110

111

112

113

114

115

116

117

118

-2.5
-2.5
-2.5
-2.5
2.8
-2.6
-2.6
-2.6
-2.5
-2.5
-2. 5
-2.5
-2. 57
-2.53
-2.53
-2.58
-2.59
-2. 62
-2.52
-2.58
-2.51
-2.50
-2.50
-2.50

+1.6
+ 1.8
+1.5
+1.5
+1.6
+ 1.5
+ 1.5
+15
+1.6
+1.6
+1.
+ 1.5
+1.58
+1. 57
+1.54
+1.55
+ 1.58
+ 1.58
+1.55
+ 1.55
+ 1.M
+1.54
+ 1.54
+ 1.M

-28.3
-28.3
-28.3
-28.3
-28.2
-28.2
-28.2
-28. 2
-28.2
-28.2
-28. 1
-28.2
-28. 18
-28. 18
-28. 15
-28. 13
-28. 13
-28. 12
-28. 15
-28. 14
-28. 15
-28. 15
-28. 15
-28. 15

+5.0
+5.0
+5.0
+4.9
+5.0

+ 18.2
+18. 2
+1S.3
+1H. 2
+ 18.2
+18.2
+ 18.2
+18.2
+ 18.2
+18.2
+18.2
+18.2
+ 18. 19
+ 18. 18
+ 18. 10
+ 18. 11
+18. 12
+ 18. 12
+18.08
+ 18.09
+ 18.08
+18.07
+ 18.06
+18.07

+10.8
+10.9
+ 10.9
+10.9
+ 10.9
+ 10.9
+ 10.9
+10.9
+11.0
+ 11.0
+11.0
+ 10.9
+10.98
+10. 97
+11.04
+11.06
+ 11.05
+ 11.07
+ 11.04
+11.08
+11.04
+ 11 04
+ 11.04
+ 11.04

+31.5
+31.5
+31.5
+31.6
+31.5
+31.5
+31.6
+31.5
+31.5
+31.5
+31.49
+31.46
+31.38
+31.40
+31.41
+31. 41
+31.37
+31.37
+31.36
+31.36
+31.38
+31.37

+21.6
+21.4
+21.4
+21. 5
+21.6
+21.6
+21. B
+21.5
+21.59
+21.59
+21.00
+21.62
+21.60
+21. 59
+21.60
+21.56
+21.54
+21. 5ti
+21. 56
+21. 56

+ 15.6
+ 15.5
+ 15. li
+ 15.6
+15. 8
+15.8
+ 15.8
+15.7
+ 15.76
+ 15.73
+15. 67
+ 15.69
+ 15.68
+ 15.68
+ 15.66
+ 15.64
+ 15.64
+ 15.65
+ 15.65
+ 15.65

-23. 2
-23.1
-23.1
-23. 0
-22.9
22 9
-22. 9
-22. 9
-22. 85
-22. 87
-22. 76
-22.73
-22. 74
- 22. 69
-22. 70
-22.69
-22.68
-22.68
-22. 67

+6.1
+6.1
+6.1
+6.1
+6.23
+6.20
+6. 22
+6.26
+6.24
+6.27
+6.21
+6.21
+6.20
+6. 21
+6 19
+6.21

' I '' 1 I 'T |

Capacitor number 117 exhibits a fairly steady in

crease of capacitance with time. The electrodes of
this capacitor were about 0.002 in. thick. This
capacitor also has a large voltage dependence of
capacitance, as will be seen later. It is believed that
the diamond wheel used to bevel the edges of the
element was contaminated with silver.

NW

3.2. Resistance Stability

Measurements of each resistance thermometer
are plotted in figure 7. The effect of variations in the
mean oil bath temperature was eliminated from the
data by plotting the differences between each resist
ance and the mean of the resistances of capacitors
10S, 109, 110, 112, and 113.
Figure 7 indicates that the resistance thermometers
are quite stable relative to each other, but shows a
flaw in our oil bath. Between September 20 and
October 10, a blast of air from a fan was directed at
the southwest corner of the oil bath, where capacitor
number 1 15 was located. This blast of air introduced
a temperature gradient of about 0.02 C into the
bath and is believed responsible for the relative varia
tions observed on September 22 and October 5.
The stabilities of the copper resistance thermome
ters were determined from a series of measurements
made with a Mueller bridge. A calibrated platinum
resistance thermometer in conjunction with measured
data for the dependence of the copper resistances
upon temperature was used to refer all measurements

5 15 25 I 5 15I 25. I 5 . O 25 I 5 O 25 I 5 IS 25 I S 1 5 25 '

SEPr
OCT
MOV.
OtC.
JAN
fES.

Fioure 7. Relative resistance thermometer stability.

Insert shows relative positions in the oil bath. The southwest corner was ab
normally cold between September 20 and October 10. A resistance change of
0.001 Q corresponds to a temperature change of about 0.01 0, or a capacitance
change of about 0.1 ppm.

241 -177

to a standard temperature. The limited information

currently available indicates that the mean of the
five selected resistances at a constant temperature
has not changed more than 0.0001
in 2 months.
This corresponds to about 0.001 C; or, referred to a
capacitance correction, about 0.01 ppm.

Table 2. Dependence of capacitance upon prior thermal

history
Septemljer 8 and September 1 1 data were taken at 25 C after subjection to 50 =C;
Septeml>er9datawerc taken at 25 C alter subjection too "C (see text). Capac
itance differences from the September 8 datu are given in parts per million.
Capacitor

3.4. Hysteresis
The capacitances of most air or solid dielectric
standards at a given temperature depend upon the
previous temperatures to which the standards have
been subjected. Table 2 shows the magnitude of
this hysteresis effect in the 1964 set of capacitors.
Prior to September 8 all of the capacitors were
subjected to a temperature of 50 C, and then
replaced in the oil bath at 25 C. On September 8,
their capacitances were measured with respect to a
capacitor maintained at 25 C throughout the experi
ment. Following these measurements, the capaci
tors were cooled to 0 C, and then placed back into
the oil bath. They were measured a second time
on September 9. Table 2 shows that the capaci
tances increased an average of 0.28 ppm following
the 0 C treatment. A second subjection to 50 C
and a return to 25 C resulted in the September 11
data in table 2. The capacitances returned most of
the way to their original September 8 values, but an
average capacitance increase of 0.06 ppm remained.
Since the data were only recorded to the nearest part
in 107, this overall change may not be real. The
effect illustrated by table 2 represents one of the
most serious faults in the capacitors, and one for
which the mechanism is not completely understood.
It is believed that the large differential expansion
coefficient between the fused silica element and the
silver electrodes may be partially responsible.

September 9

September 11

0
0
0
0
0
0
0
0
0

+0. 15
+.35
+.45
+.35
+.25
+.25
+.35
+.35
+.05
+.28

+0.1
.0
+.1
+.1
+.1
+.1
-.1
.0
+.1
+.06

108
1U9
no
112
113
114
115
116
117
Mean change

3.3. Temperature Coefficients

The temperatures of the capacitance standards
may be measured by means of the internal 25-S2
resistance thermometers. The capacitances at a
specific temperature may then be determined if the
capacitance temperature coefficients are known.
In practice, the temperature is usually treated as
an implicit variable. A report of capacitance cali
bration then contains values for the capacitance and
for the thermometer resistance at a temperature
near 25 C. A later measurement yielding a different
thermometer resistance is corrected to the standard
resistance thermometer value by reference to a plot
of capacitance versus resistance for the standard.
Measurements of the capacitance-resistance func
tion have been made over the temperature range
20 C to 30 C. The curves deviate from a straight
line by about 2 ppm at the extremes and are all con
cave upwards. The slopes at 25 C range from 100.4
ppm/fi to 110.0 ppm/ft. The actual capacitance
temperature coefficients are about 10 ppm/C which
is substantially smaller than the 14 ppm/C tempera
ture coefficients of the 1961 fused silica capacitors.
No explanation for this difference has been found.

September 8

3.5. Voltage Dependence

Some techniques and apparatus were developed
recently at NBS for measuring the dependence of
100 and 1000 pF capacitors upon voltage, with un
certainties in the order of 1 part in 109 [4]. The
equipment was used for measuring the voltage
dependencies of the 10 pF capacitors with slightly
larger uncertainties. Table 3 shows the capacitance
differences observed when the voltage was changed
from 100 V to 200 V. The computed probable
error of the data in table 3 is 6 parts in 1010, and
the systematic error is estimated to be less than 2
parts in 109. Since the voltage dependence of a
standard depends upon which electrode is at ground
potential, all of the standards are marked with a
"D" near the coaxial connector chosen for the
ground potential lead.
Table 3. Dependence of capacitance and dissipation factor
upon voltage
The increases i'i capacitance and dissipation factor due to increasing the volUire
from 100 V to 20U V mis at 1592 Hz aic. shown.
Capacitor

AC
C

AO
C

108
ion
no
112
113
114
115
116
117
118

+1.3X10-"
-3.6
-5.2
-5.0
-2.6
-1.0
+4.8
-1.8
+12. 4
+9.8

-3. 1X10-'
2 2
+ 1.2
-4.6
+0 6
+1.0
-0.1
-.3
-2.2
-1.4

3.6. Frequency Dependence and Dissipation Factor

The result of comparing the fused silica capacitors
with two 10 pF air capacitors at three frequencies
is reported in table 4. Capacitor A was constructed
with concentric cylindrical electrodes, and capacitor
B with rigid parallel plates. Identical results were
obtained at 159 Hz, using the two capacitors, but
a significant difference was observed at 15900 Hz.
The cause of the discrepancy is not known, but
probably involves either mechanical resonances or
excessive series inductances in one or both of the

242 -178

air capacitors. The measurements at 159 Hz were

extremely difficult to make due to poor bridge
sensitivity. The computed probable error of the
159 Hz data in table 4 is 2X10-7.
Table 4. Frequency dependence of the fused silica capacitors
relative to two air capacitors (see text)
Capacitor

159 Hz

10S
109
110
112
113
114
115
UH
117
118

ppm
+1.3
41.0
+ 1.8
+0.2
+1.0
+0.1
+1.2
+0.3
+l.f)
+1.S

1582 Hz

carried to NRC and returned between August 4 and

August 17, 1964, and again between December 8 and
December 11, 1964. Comparisons of the two ca
pacitors at NRC indicated no relative change in the
two capacitances larger than 0.2 ppm; and as can
be seen from table 1, no change in number 108 oc
curred during either of the two round trips.

15000 Hz A 15900 Hz B
4. Conclusion

0
0
0
0
0
0
0
0
0
0

PP'n
-0.C
. 2
-.8
+.2
-.9
+.2
-.8
-.3
-1.2
-0.8

ppm
-1.9
-1.5
-2.1
-1.1
-2.2
-l!. 1
-3.1
-?.6
-.1.5
-1.9

Accurate measurements of dissipation factor can

not be made at NBS at this time. Comparisons of
the fused silica dielectric capacitors with several
types of air capacitors suggest that the dissipation
factors of the fused silica capacitors are probably
between 0 and +3X10"6 at 1592 Hz.
3.7. Shock Sensitivity
The capacitors were tested upon completion for
shock sensitivity by dropping them onto a hardwood
table. Various angles of impact and various heights
up to 8 in. were tried. The largest capacitance
change observed was 1 part in 107 which occurred
when a capacitor fell about S in. in an upright posi
tion. Falls through smaller distances usually pro
duced no observable change in capacitance.
3.8. Transportation Experiences
A fused silica dielectric capacitor designated num
ber 107 and which was identical with those described
in this paper was shipped to the National Research
Council of Canada (NRC) on July 28, 1964, via air
parcel post. Prior to shipment, it was compared
with capacitor number 108. Number 108 was hand-

Much more data is needed to evaluate conclusively

the reliability of the 1964 set of fused silica dielectric
capacitors. Existing evidence suggests that sta
bilities in the order of a few parts in 107 may be ex
pected, and that the capacitors can probably with
stand normal handling during shipment between
laboratories.
Some further modifications of the design may re
sult in an improved standard. A study of the
mechanism which produces the capacitance hysteresis
effect with large temperature changes would be of
value, and consideration might be given to improving
the support system for the fused silica element.
It is felt that the choice of 10 pF for a fused silica
dielectric capacitor results in a design of nearly
optimum stability. Other values would be useful
as secondary standards and should not be difficult to
construct.
5. References
[1] M. Kanno, Researches on standard capacitors, researches
of the elect rotechnical laboratory (Japan), No. 597,
April 1961 (in Japanese).
[2] R. D. Cutkosky and L. H. Lee, The construction and
behavior of a transportable ten picofarad capacitor,
comite consulatif d'electricite, Aupres Du "Comitc
International Dcs Poids et Mcsures," 10 Session 1963,
p. 44.
[3] Standard capacitors and their accuracy in practice, Notes
on Applied Science 13, (NPL) 1955.
[4] J. Q. Shields, The voltage dependence of precision air
capacitors, to l>e published.

243-179

(Paper 69C3-196)

JOURNAL OF RESEARCH of the National Bureau of StandardsC. Engineering and Instrumentation

Vol. 69C, No. 4, October-December 1965

Voltage Dependence of Precision Air Capacitors

J. Q. Shields
(August 1, 1965)
Methods are presented for measuring both the relative and absolute voltage dependence of admit
tance standards. The practical circuits based on the methods are composed of precision bridges
which are described in detail. Measurements with these circuits yield values for the proportional
changes in admittance which result from specified changes in voltage. The results of measurements
on selected capacitors are presented. Uncertainties are of the order of one part in 10*.
A general instability in the voltage dependence of precision air capacitors was observed at the
higher accuracy levels. Most of the instability is believed to be caused by changes in the electrode
surfaces.
Key Words: Bridge, capacitor, coaxial choke, difference transformer, electrode surfaces, thin
films, three-terminal, transformer, voltage coefficient, voltage dependence, voltage
ratio.

1. Introduction
Several years ago a computable cross capacitor,
based on a theorem by A. M. Thompson and D. G.
Lampard [l],1 was constructed at NBS [2]. The
value of this capacitor is computable, in esu, within
an estimated accuracy of 2 parts per million (ppm).
A new cross capacitor of improved design is now being
constructed. It is expected that this capacitor will
be computable, in esu, within a few parts in 108 or
better.
The expected increase in accuracy not only requires
better bridges and capacitors, but also requires that
many sources of error be reevaluated. One such
source of error is the voltage dependence of precision
air capacitors.
To meet our particular needs, the quantity chosen
as a measure of voltage dependence is termed y and
is defined by the equation
Y1=Y1(l+y)
where Y\ is the admittance of a given standard at
voltage V\, Y2 is the admittance of the same standard
at voltage V2, and the condition V2 > V\ is imposed to
restrict the meaning of y. A given standard may
have any number of values of y, each with a speci
fied voltage change. Since y represents a propor
tional change in admittance (AY/Y), this measure of
voltage dependence is in the preferred form for cor
rections to precision measurements.
In the past, most measurements of voltage depend
ence have been confined to a determination of the
ratio, (1+ ya)l( l+yi>), where ya and yi> apply to differ
ent capacitors but have the same specified voltage
change. One exception to the above is the work of
N. L. Kusters and 0. Petersons [3]. They developed

1 Figures in brackets indicate the literature references at the end of this paper.

a method for measuring the voltage dependence of

an individual capacitor, provided the voltage depend
ence results only from electrode deflections.
The accuracy requirements described earlier re
sulted in the need for a method which would be inde
pendent of the mechanisms which cause voltage
dependence. Such a method is described in section 2.
The method yields values of y corresponding to a
change in voltage from V to IV, where V\% a variable.
The practical circuits based on the method are
described in sections 3 and 4. These circuits were
used to measure values of y for several capacitors
corresponding to the following voltage changes: 12.5
to 25, 25 to 50, 50 to 100, and 100 to 200 V, rms, at
1592 Hz. Measurement uncertainties had to be kept
small since the above values of y are used to calculate
other values of y in different voltage ranges such as
from zero to 200 V. The latter value is needed for
the determination of the absolute unit of capacitance.
Small uncertainties also provide the means for quickly
detecting instabilities in the voltage dependence of
capacitors. It was found that the voltage dependence
of precision air capacitors was, in general, unstable
at the accuracy level described. This subject is dis
cussed in section 5.
2. TheoryOne method for measuring the voltage ratio of a
transformer is to connect admittance standards to
the transformer so as to form a bridge as shown in
figure 1 [4]. In general, if the transformer has a
nominally m:n ratio, then m + n nominally equal admit
tances are used, and m + n detector balances are
required corresponding to a cyclic permutation of
these admittances. In this sequence of balances
each admittance appears with voltage E\ m times,
and with voltage E2 n times. If 1 t4 E2, a ratio meas
urement error results since the admittance of a practical standard changes when the voltage applied to it

244-265

Ecf@
de r

<S>:

lDET-2f

T
t
DET-I
FIGURE 1. Circuit for measuring the voltage ratio of a trans
former [4].

4DET-3h

Y2

FIGURE 2. Basic three-bridge-circuit.

is changed. The difference between the measured
ratio and the true ratio is related to the average change
in admittance of the standards.
If the transformer ratio can be determined from a where EB 1 +8. Eliminating Y\ and F2, we obtain
second set of measurements which is independent of
Ea
voltage, the combination of the two sets of measure
ments will yield a value for the average change in ad
Sb =(0i + 92).
(3)
mittance of the m + n standards corresponding to a
change in voltage from E\ to 2. The change in admit
The balance equations for bridge 2 are
tance of each standard can then be determined by
Y3 Eb + e3 1 + 6B + q3
measuring the relative changes between various pairs
(4)
of standards.
F,
<
1+8,
The basis of this paper is a method of the type de
Y4^EB + e<=l + dB + q<
scribed above in which the second set of measurements
(5)
consists of determining a group of 1 : 1 ratios which
Y3
Ec
l + 8c
can be combined to determine the m:n ratio of a
transformer. For simplicity, a transformer having a where 1
jr 1 + 8C. Eliminating Y3 and F<, we obtain
2 : 1 ratio was chosen so that only two 1 : 1 ratios need
H'A
be determined.
It is assumed that taps are brought out from the
s _03
8c
= :j + 2i28fl
^
l"i
(6)
secondary winding of a bridge transformer [5] so as
to obtain three secondary voltages of approximately
equal magnitude. These voltages are represented where 1 is a small correction term consisting of second
by Ea, Eh, and Ec in figure 2. The circuit shown in and higher order terms in q.
figure 2 consists of three bridges, each of which may
The balance equations for bridge 3 are
be balanced, in turn, by the small, adjustable voltage
e. For the case in which the circuit elements are
closely matched, elE.\=q<\, and only a low accu
Ye+Y'7
Ea + Ec + e* 2 + 88 + 8f+95
'
racy calibration of q is needed for high accuracy
determinations of ratios such as EbIEaY^(l+y6)_ EA-es
\-q6
It is assumed that the changes in Y\, F2, Y3 . . . Y7
F;+Fi
EB + E<+ee 2 + 8b + S(+<?6
W
which result from changes in e can be neglected.
Following from this, we may consider the voltages
YHl+yi)_ EA-e7 _
l-q7
applied to F5, F8, and Y7 to be either EA or 2EA and
Fi+Fi
EB + Ec + e7 2 + 8B + 8c + q7
W
define the values of F5, F6, and Y7 to be Y'it Fg, and Y'7
at voltage EA, and Y'Jil + yt), Y'jl + y,), and Y'+l + y,)
Eliminating F5, Fe, and Y'7, we obtain
at voltage 2EA.
The balance equations for bridge 1 are
8b + 6V + qs + qe + q7 .
y=
2 +e'(10)
Y\ _ EA ei __ 1 qx
(l)
Yt EB + e, l + 8 + g,
where
Yt_EA-et
Y\ Eb +

11 + Sg + q^

(2)
245-266

-_-ys+ys + -y7
(11)

and 62 is a small correction term consisting of second

and higher order terms in q and y.
Solving eqs (3), (6), and (10), we obtain
-V 4gt 4- 4g2 - <?3 <74: 2qh 2qe 2q7
4

e,2e2
2

(12)

In order to separate real and imaginary parts, let

y-y'-rf

(13)
Figure 3a. Bridge I of the practical three-bridge-circuit-

and
q^q'-jq"

(14)

so that

Ch

-,_4gi + 4ri-gj-ri-2gj-2ri-2ri
y
4

c!-26Z
2

(15)

and
- _ 49;' + ql - q'j - g'l - 2q'j - 2gg - 2g?
y
4

e',' - 26^' (16)

2

For the capacitors and accuracies described in this

paper, y' is equal to the proportional change in capaci
tance (AC/C), and y" is equal to the change in dis
sipation factor, corresponding to a specified increase
in voltage.
The values of 75, y, and y7 can be separated from
the value of y by measuring the differences, ys ys,
yt y7, and y7 y<,. Measurements of this type are
described in section 4. Combining the differences,
we obtain
y5 = y + (-y5-"y6)/3-(y7-75)/3

(17)

ye = y + (y6 - y7)/3 - (y, - y8)/3

(18)

y7=r+(77-y5)/3-(76-77)/3.

(19)

FIGURE 3b. Bridge 2 of the practical three-bridge-circuit.

3. Practical Three-Bridge-Circuit
The method described in the previous section was
used to determine the voltage dependence of three 100
picofarad (pF) air capacitors, designated 100A, 100B,
and 100C. This section describes the equipment and
techniques used to determine the values of
yiOO =

(20)

corresponding to the following voltage changes: 12.5

to 25, 25 to 50, 50 to 100, and 100 to 200 V, rms, at
1592 Hz.
If small capacitors are to be well defined, they must
be shielded. The addition of shielding to the circuit
of figure 2 is shown in figures 3a, 3b, and 3c. Each of
the admittance standards represented in figures 3a,
3b, and 3c consists of a completely shielded threeterminal air capacitor connected to a pair of coaxial

FIGURE 3c. Bridge 3 of the practical three-bridge-circuit.

cables. The defining terminals of each standard
are located at one of the detector junctions and at one
of the coaxial connectors designated a, b, c . . . p.
The required circuits are obtained by joining con
nectors a. b, c . . . g to connectors. A, B. C . . . G
in various configurations. When connector a is joined
to connector A and b to B, the balance equation is
given by equation (1), section 2. When a is joined to
B and b to A, the balance equation is given by eq (2).
Similarly, eqs (4) and (5) are applicable to figure 3b
and eqs (7), (8). and (9) are applicable to figure 3c.
For purposes of evaluating the effects of self and
mutual impedances in the leads (not shown in the
figures), it will initially be assumed that: (1) net cur

246-267

rents through the "coaxial chokes" [6], designated Ch

in the figures, are negligibly small, (2) the voltages at
connector A are, to sufficient accuracy, independent
of whether A is joined to a or to b, and similarly for the
voltages at connectors B, C, D . . . G, (3) only negli
gible effects result from externally induced emfs and
external loading, e.g., capacitance from the bridge
shielding to its surroundings.
Subject to the above assumptions, the two values of
t? corresponding to the two balance conditions for
bridge 1 yield a measure of the voltage ratio, VaIVh
where Va is the voltage between the inner and outer
terminals of connector A with e = 0 and Vb is the
voltage between the inner and outer terminals of con
nector B with e = 0. Similarly, a measure of the volt
age ratios, Vt lVii and Vt:/Vy are obtained using bridges
2 and 3, respectively. The value of yioo is calculated
from the measured values of Va/Vh, VcIVd, and Ve\Vp.
It is assumed that the relationship between voltages
Va, Vb, Vc . . . V(; is known. Ideally, the voltage
drops in the leads which precede connectors A, B,
C . . . G should be zero, but if the voltage drops are
small, eq (12) is sufficiently accurate when the rela
tionship between the voltages is as follows: V a=Vi>,
VH = Vh = Vv Ve=Va + Vc. Since voltage ratios
rather than voltages are measured, other satisfactory
relationships between the voltages can be described,
e.g., Vv and Vu can each be changed if V'c\Vo remains
unchanged. Thus, the error in the measurement of
yxoo which results from voltage drops in the leads will
be small if the impedances in the leads which join
various terminals of connectors A, B, C . . . C are
small or properly matched. These features are present
in the junction box shown within broken lines in figure
4. Parallel lines represent copper strips separated by
thin strips of insulation.
The error in the measurement of y,oo resulting from
lead impedances within the junction box and resulting
from contact impedances between connectors A, B,
C . . . G and connectors a, b, c . . . g was found to be
approximately 1 X 10-8. Most of this error was cor-

rected for by changing the currents at the terminals

of connectors a, b, c . . . g and observing the change
in the measured value of yioo. Since the current at
the inner or outer terminal of connector a was nearly
proportional to the internal load. Ki + YA, and similarly
for the currents at the terminals of b, c, d, . . . g, the
following procedure was used: (1) Yt. Yb, YV . . . Kg
were adjusted so that each of the seven internal loads,
Y, + YA, Y2+Yu, Y3+Yc . . . Y7+Yc, were increased
by a known factor, (2) the change in the measured
value of yioo was used to calculate a third value of yioo
corresponding to the internal loads being zero. The
above procedure amounts to an extrapolation of the
effect of internal loads to zero.
The internal loads were measured by disconnecting
A, B, C . . . G from a, b, c . . . g, shorting the detec
tors, and measuring the two-terminal admittances be
tween the inner and outer terminals of connectors a,
b, c ... g. Since the lead impedances within the
junction box were small, low accuracy measurements
were sufficient for purposes of extrapolating the
internal loads to zero. The larger equivalent lead
impedances in that portion of the circuit which pre
ceded the junction box (see fig. 4) required that more
accurate measurements be made for purposes of ad
justing YB, Yd, YF, and KG- so that Y, + YA = Y2 + YB,
Y3 + YC=Y4 + Y, and K5+ Ys = K6+ Yt= Y7+ YG.
The accuracy of adjustment needed was determined
by connecting additional loads between the inner and
outer terminals of A, B, C . . . G, each in turn, and
noting the changes in e required to rebalance the
bridges. It was found that if the internal loads were
adjusted within 1 X 10"* mho, the voltages at con
nector A were, to sufficient accuracy, independent of
whether A was joined to a or to b, and similarly for the
voltages at B, C, D . . . G.
The coaxial chokes used consisted of a number of
turns of the coaxial cable threaded through a high
permeability core. The resulting impedance to net
current in the cable was about 400 ft at 1592 Hz.
Since the impedance in the path, P\ A a P2, of
figure 3a was about 0.1 ft, net current through the
coaxial choke was of the order of a few parts in 104
of the current in the inner or outer conductor of the
cable [6]. The impedances in the paths. Pi A a
and P\ B b, differed by less than 2 mft. The
maximum internal load was about 3000 pF.
Thus far in the discussion the effects of externally
induced emfs and external loading have been ne
glected. Although these effects were small in the
practical circuit used, they were not negligible. Ap
propriate corrections were determined from additional
measurements which consisted of recording the un
balance at the detectors when the admittance stand
ards of each bridge, in turn, were joined to the shorting
connectors. A' , B' , C' , . . . G', (see fig. 4) while the
other standards were joined to A, B,C, . . . G. Since
the accuracy of the corrections depended on the size
of the unbalance at each detector, it was desirable for
the effects of externally induced emfs and external
FlGURE 4. Source of bridge voltages.
Junction bo* Iwithin broken lines) connected by coaxial cables to an "active direct- loads to be small. The techniques used to reduce
reading ratio act" [7|.
these effects will be considered next.
315-411 O - 68 - 17

247-268

FIGURE 5. Bridge 2 redrawn to illustrate external loads and sources

ofinduced emf.

Bridge 2 has been redrawn in figure 5 for purposes

of illustrating the different types of induced emfs
and external loads. Since the shielding of bridge 2
is not at ground potential, the largest external loads
are associated with this bridge. External loads are
represented by Ci, C2, Cj, and C4 in figure 5. Induced
emfs are represented indirectly by showing areas in
which a time-varying magnetic flux is assumed to exist.
These areas are designated A\, A-i. A3, and At in figure
5.
Most of the flux through A, existed in the region
preceding the junction box and thus had no effect on
the measured value of y\oo- Flux through A2 was re
duced by using coaxial cables of sufficient length so
that the three-terminal capacitor could be positioned
at a distance from sources of appreciable magnetic
field. Flux through A3 was reduced by using coaxial
cables and magnetic shielding in that portion of the
detector circuit which precedes voltage amplification.
The effects of flux through A\ were reduced by the
use of coaxial chokes, e.g., if the high permeability
core positioned between connector r and P* were
removed from the circuit, the current through Y3 which
results from flux through A4 would be increased by a
factor of a few thousand.
The effects of external loads were reduced by the
use of coaxial chokes in conjunction with low-imped
ance leads which bypass the junction box (hereafter
termed bypass leads). The lead connecting P3 to Pa
and the lead connecting P3 to /5 are examples of by
pass leads in figure 5. These leads, which have
impedances of about 0.1 SI. supply most of the current
to C3 and C4. Thus, if coaxial chokes and bypass
leads are properly positioned, voltage drops will be
reduced in critical portions of the circuit such as the
junction box leads and the cables of the admittance
standards and detector circuit.
The effects of external loads were also reduced
by the use of an electrostatic shield which is repre
sented by a broken line in figure 6. This figure
represents more closely the actual circuit used. The
electrostatic shield, maintained at a potential of
approximately Eb with respect to ground, encloses
those circuit elements which would otherwise have a
large capacitance to the grounded surroundings.
Current to the shield is supplied by the bypass lead
connected to P3.

Ch

FIGURE 6. Bridge 2 redrawn to show the difference

and additional shielding.

WIRE (Coincident with axis)

PRIMARY WINDING (100 Turns)

COAXIAL
V CONNECTOR

COAXIAL
CONNECTOR
SUPERMALLOY CORE
SHIELD NO. I
SHIELD NO. 2
SHIELD NO 3
SHIELD NO. 4
SHIELD NO. 5
CROSS -CAPACITANCE AREA
FIGURE 7. Cross section through the axis of the diffe
former.

The transformer shown in- figure 6 is termed a di!

ference transformer2 and is used to change the poten
tial of the detector cables with respect to ground from
approximately Eb to zero. This was necessary in
order that a phase-sensitive detector having a grounded
power supply could be used. An adequate signal-!"
noise ratio was maintained by using a battery-operated
preamplifier having a voltage amplification of approxi
mately 103 (see fig. 6). Coaxial chokes assured thai
the bypass leads in figure 6 supplied most of the cur
rent to the large capacitance which existed between
the shielding of the difference transformer.
A cross-sectional view of the difference transformer
is shown in figure 7. The secondary winding is a
' The basic ideas in the transformer are similar to those in a transformer built bs A M
Thompson and termed by him a difference transformer.

248-269

The external loads, Ct and C-i in figure 5, represent

single turn consisting of shield No. 3 connected to a
wire through the center of the core. With the excep the capacitances between the grounded surroundings
tion of shield No. 4 which is mu-metal, all shielding is and those cables which are located between the co
axial chokes and connectors c and d. The circuit
brass.
Most of the current resulting from the capacitance was designed so that Cr and C% would be small and
between shield No. 1 and shield No. 2 does not encircle nearly equal. Measurements of Ci and Ci and of the
lead impedances in the junction box indicated that
the Supermalloy core. The current which does en
circle the core, if not compensated, would induce a the resulting error in the measurement of Vioo was
small. This error is accounted for in item 1, table 1.
voltage in the detector winding. The two capaci
tances which result in core excitation are associated Also accounted for in item 1, table 1 are errors re
with the corners of the shield gaps. With reference sulting from capacitance between the copper strips
to the cross-capacitance area shown in figure 7, the of the junction box and errors resulting from imperfect
two capacitances are (1) from the upper left to the balancing of the internal loads.
lower right, (2) from the upper right to the lower left.
It can be seen that the two currents encircle the core Table 1. Uncertainties in the measurement of yioo = '>'!oo j"y"iio
in opposite senses, and hence their effects can be
made to cancel by proper adjustment of the gap
Sources of error
Systematic errors in Systematic rm>r in
geometry. The gap was adjusted experimentally by
y
observing the detector voltage with the transformer
Part, in I0'
l-artu in 10"
input open circuited and with the bypass leads
1
1
2
connected.
3. Dclecli>r zero
1
2
1
As a result of the techniques described above, the 5. Drift of admittance* and transformer
2
1
unbalance at the detectors was small when the ad
mittance standards of each bridge, in turn, were
6
6
joined to the shorting connectors, A', B' , C . . . G',
while the other standards were joined to A, B, C
Standard deviation of <fIM or Y1m is 3 X 10 *
... 6". Thus, accurate corrections were obtained
for most of the effects of induced emfs and external
loads. Those effects which were not accounted for
Table 1 lists only the significant errors which were
in the corrections will be considered next.
not corrected for by extrapolating the internal loads
Since currents were reduced when the standards to zero or by using the shorting terminals. A', B',
were connected to the shorting terminals A', B' , C . . . G', as described earlier. Item 2, table 1,
C , . . . C, induced emfs resulting from currents accounts for small changes in the contact impedances
in that portion of the circuit shown in figure 4 were between a, b, c . . . g and A, B, C, . . . G which
not entirely accounted for in the above corrections. were found to occur with continued use. Item 3,
The use of coaxial cables, parallel-strip leads, and table 1, accounts for possible systematic errors which
magnetic shielding assured that the resulting errors result from observer bias in determining the null
were negligible. It is interesting to note that if the balance condition.
Item 4. table 1. accounts for errors in the calibration
errors had not been negligible, they would have been
of
the instrument used to obtain Ea, Eh, Ec, and e.
accounted for in the extrapolation of the internal
loads to zero. The effect of induced emfs resulting The essential features of the instrument, termed an
from currents within the standards was negligible "active direct-reading ratio set," are described in
since the phase of the induced emfs was shifted by reference [7]. The adjustments of q
180 when the standards were permuted.
were made by adjusting two sets of dials, one cor
Magnetic fields in the vicinity of the junction box responding to changes in q and the other to changes
may induce emfs which are selective with respect in q". The maximum range of q' or q" was 10 ppm
to connectors A, B, C . . . G, thus causing errors with a minimum adjustment of 2 X 10~10.
in the measurement of yioo. Induced emfs of this
Item 5, table 1, accounts for errors which result
type were kept small by using closely spaced copper from the admittances and transformer ratio having a
strips in the construction of the junction box (see fig. 4). nonconstanl drift rate and from observer bias in timing
The following procedure was used to assure that the the bridge balances which were made so as to obtain
remaining errors were negligible: (1) A small probe values of q in the following sequence: 93, 94. q3; 91, 92,
was used to measure the magnetic field in the vicin
Qu 95, q, 97, 9, <7r; 91. qi, 9i, 93. q*, 9s- The above
ity of the junction box, including the region where set of values could be obtained in about 12 minutes
changes in geometry occur due to changes in the with Ea = 100 V, rms. The standard deviations of
connections of the standards to the junction box, yl00 and yioo, obtained from one such set, were about
(2) the current in a loop of wire was adjusted so as to 6 X 10~10. Slightly larger standard deviations were
increase the measured field by a factor of approxi
obtained at lower voltages. A sufficient number of
mately 100, (3) the changes in e required to rebalance measurements was made so that the final values of
the bridges were related to changes in yioo.
y"\oo a"d Ti'oo had standard deviations of about 3 X 10~10.
249-270

4. Difference Measurements
The difference in voltage dependence between two
standards whose admittances are approximately equal
can be determined from ordinary bridge measure
ments, e.g., simple substitution combined with a
change in bridge voltage. The particular circuit
used for this type of difference measurement is de
scribed in section 4.1.
When the admittances of the two standards are
radically different, additional measurements are re
quired. These measurements are described in section
4.2.
4. 1 . Difference Measurements Involving
Approximately Equal Admittances
The circuit shown in figure 3a is typical of the
circuits used for difference measurements. The
two general balance equations for this bridge have
already been presented, namely eqs (1) and (2). For
the case in which the voltage, Ea, has two specific
values, Em. and Eaii, four specific balance equations
may be written:
Yu.
Yu.

Eu. eu.
E,. + cu.

1 - qu.
1 + 8m. + qu.

(21)

Y:l.
Yu.

Eu. fit.
Ehi. + eu.

!-</,,
1 +S/,/. + q<i.

(22)

Yu,
Y,

Em, em
E nil + I'm

1 qui
1 +8i,n + q,n

(231

Y,
Y,

E Ml e>n
E mi + I'm

1 - qui
1 + o/i// + q-.H

(24)

II
Yu, = > w.U + yi)

(25)

Y-. = Yuk\ + y2)

(261

y\ y-i = (qui qu.) iq\H ~ qu) + :i

(27)

and

then

where e.i is a small correction term consisting of second

and higher order terms in q and y.
The method described above was used to measure
each of the following differences: "/iooa Tuiob. Vioob
yiiMic. yuKic yiooA yiooA yiooiw. yiixHiu- yuK>or.
and yimmc- yi<HWA- where (he subscripts refer to three
100 pF air capacitors and three 1000 pF air capacitors
designated 100A, 100B, 100C, and 1000A. 1000B.
1000(1. respectively. Four values were obtained for
each of the above differences corresponding to the
following voltage ehanjzes: 12.5 to 25, 25 to 50. 50 to 100.
and 100 to 200 V. rms. at 1592 Hz. The discussion of
measurement uncertainties which follows is appli
cable to each of these values.

Since the differences, qzH qiL and q\n q\L, were

small and of opposite sign in eq (27), most errors ot
the type described in section 3 were negligible. The
only exceptions were errors resulting from drift !
the admittances and transformer ratio. These error*
were partially reduced by the measuring sequence
which follows: qu., q\H, Uit, qiL, qnH, qiu Rather tha/i
attempting to obtain a precise bridge balance at a
specified time, each balance in the above sequence
was made as quickly as possible. Nonuniform timing
was almost inevitable since the detector sensitivity
was better at high voltage. The resulting errors were
kept small by injecting a small current to compensate
the effect of capacitor drift. This was accomplished
with an auxiliary circuit, termed a "drifter circuit."
If no adjustments of e are made, the relative change*
in capacitance with time result in a time-variation in
the amplitude of the current at the detector. The
drifter circuit is used to inject a compensating current
at the detector. The basic part of the circuit is a threeterminal, variable, air capacitor having a maximum
capacitance of 10 pF. The adjusting shaft of the
capacitor is connected to a constant-speed motor
with speed reducer, thus providing a capacitance
which varies linearly with time. Coaxial cables are
used to connect the terminals of the capacitor to the
detector junction and to an inductive voltage divider
whose voltage is supplied by the "active direct-reading
ratio-set" used to obtain Ea, Eg, and e.
The amplitude of the current injected at the de
tector has a first derivative with respect to time which
can be adjusted by adjusting the inductive voltage
divider and a second derivative which is zero. Since
the drift rate of the admittance standards changed
only slightly during the time needed for a set of detec
tor balances, the use of the drifter circuit resulted in
appreciable reductions in the time rate of change of e
required to maintain a detector balance. In addition,
the deviations from constant drift rate, which can
result in systematic errors, were easily measured
A sufficient number of measurements was made so
that the standard deviation of each of the final value*.
etc., was
yiooA ft 00B'
3xl0"10 or less. A systematic error of 2 X lO "
was assigned to each of the final values to account
for incomplete compensation of the admittance and
transformer ratio drift.

4.2. Difference Measurements Involving

Radically Different Admittances
The circuit shown in figure 8 is used to measure the
difference,
_ _ ys + ys + yio
y-y*3

ys + ys + y?
3^

If connectors Q, R, and S are permuted cyclicaU1

with respect to connectors E, E. and C of the junctif
box (fig. 4), six different bridge balances can be madf

250-271

each of the differences to account for systematic

errors which may be described as similar to items 3, 4,
and 5 in table 1.

5. Results

FIGURE 8. Circuit usedfor difference measurements with

Ya - Y. * Y, ^ Y8 * Y, - Y,.

Three of the six balance equations have already been

presented, eqs (7), (8), and (9). The other three
equations are easily obtained by changing the sub
scripts of Y', y, e, and q in eqs (7), (8), and (9). Solv
ing the six equations, we obtain
y.-% = * + + ^*--g,0 + 64

(29)

where e< is a small correction term consisting of sec

ond and higher order terms in y and q.
The method described above was used to measure
the difference,
yiooo- yioo =

V.OOOA + yiOOOB
2

VlOOOC

3
Four different values were obtained for the above
difference, corresponding to the following changes in
voltage: 12.5 to 25, 25 to 50, 50 to 100, and 100 to 200 V,
rms, at 1592 Hz. The discussion of measurement
uncertainties which follows is applicable to each of
these values.
Corrections were obtained for the effects of induced
emfs and external loads (see section 3) by joining
connectors Q, R, and S to the shorting connectors,
E', F', and C (fig. 4). It was not necessary for the
internal loads to be equal or to extrapolate the internal
loads to zero. Errors resulting from admittance and
transformer ratio drift were reduced by obtaining
values of q in the following sequence: 95, q, 97, 9,
9s, g, 9. gio, 9. 9s, 9s. 9 9' 9e, 9s- A sufficient
number of measurements was made so that the stand
ard deviation of each of the differences, y'l000 Yloo
and Tiooo- 7i'oo' was reduced to about 5X10-10. In
addition, an allowance of 5 X 10~10 was assigned to

The values in tables 2, 3, and 4 were calculated

from the results of measurements described in sec
tions 3 and 4. Capacitors 100A, 100B, and 100C are
of cylindrical construction, whereas capacitors 1000A,
1000B, and 1000C are of parallel-plate construction.
Most of the voltage dependence of capacitors 1000A,
1000B, and 1000C is believed to be caused by elec
trode deflections [3]. Most of the voltage dependence
of capacitors 100A, 100B, and 100C is believed to be
associated with the electrode surfaces.
The values in table 4 were obtained either imme
diately after assembly of the capacitors, one month
after assembly, or eight months after assembly. The
values in table 2 were obtained eight months after
assembly. A comparison indicates that the values of
Tiooa. "yiooB> and yiooc were probably smallest imme
diately after assembly. The changes in y are believed
to have been caused by structural changes on the
surfaces of the brass electrodes or by the migration
of thin films onto the electrode surfaces from regions
of the capacitor which were difficult to clean. A con
sideration of typical dimensions leads to the conclu
sion that sizeable effects could result from thin films,
e.g., the ratio of molecular diameters to typical elec
trode separations is considerably larger than the
desired accuracy of a few parts in 10*.
Measurements, similar to those in table 4 and with
comparable accuracies, were also made on capacitors
1000A, 1000B, and 1000C. The largest change during
the eight month period was 4xlO~IH. Capacitors
1000A, 1000B, and 1000C had been in use for a num
ber of years prior to the measurements described.
The electrodes were gold plated. These capacitors
and capacitors 100A, 100B, and 100C were evacuated
with an oil sealed mechanical pump trapped with
dry ice.
TABLE 2. ( ottage dependence of capacitors 100A, ItHIH. ami KMC
Valui-* <! >' ri'|iri"*Mil |rn|Mirliintjl rlianai-. m i a|iai Ham i- lAf.'/O. Valurmi) y" rr|irr.rnl
liana*-, in ili..ipaltiiii failiir. Standard itrvialiim = .1 x I0",#. St.tr-malir rrrm
<;xi .
Changr in viillaui1 alls I ratal
100 1.. 21*)
.VI h. 100
2.i l.. SO
I2.Si.i25

Part, ill I0
+ \M
+ .'(
.1
.11

251-272

Pan. iii 10"

0.')
ii
+ 1
.1

-,

( fanar in vullaar
i'otts 1 I'mjl
1
200
M In 100
25 n. VI
12.5 1" 25

Pan- in I0
+ 8.0
+ 2.'l
* 1.4
M

Part, in III4 II 1
II
- .1
.11

Pan. in I0
-0.4
+ 2
.1
-.2

Pan. ill Ki
ll II
- _'1
.1

Table 3. Voltage dependence of capacitors IOOOA, IOOOB, and

I0OOC
Standard dt-vialmn = 6 X 10 IB. SyMrmatk- rrrnr < 1.2 x 10-.
( Dianur in vnhtiK'r

y'nmr

1 nits lrm.,1
l.i 2I>I
.VI I., Hal
12.3 l 25

Parti, in I0>
+ 70.8
+ 18.2
+ 3.0
+ 2.0

Pari* in 10*
+ HM
+ 4.3
+ 1.2

< lian|- in v.iltajtr

I oils Irmsi
100 In 200
50 I.. KM
In 30
12.3 In 25

Pan. in 10"
+ 84.5
+ 23.2
+ 6.7
+ l.
w

Pan. in 10"
+ 0.1
.0
...1
- .I.

Part, in IO
+ 1.2
+M
+ 0.3

Pan. in 10"
+ 2.3
2.2
+ 1.6
+ 0.8

TABLE 4. Instabilities in the volla/ie dependence of capacitors

KMIA, I0OB. and I00C
Ml >dliu-. i nrrr-imiui I,, a . hange in .nllayr fr..m 100 I,, 200 V. rmv Standard drvialinn
= 2 III1". St.lt-mati< rrn.r < 2 X 10>".
I imr
Months
0
1
8

Part* in 10"
-0.3
- .6
-6.1

Timr

^ia _ Tnaa

Months
0
1
8

Parli. in 10"
~0.2
+ .1
+ .5

tI -y'imc

Ttmr ^iaA

Pan. in 10+ 0.1

+ .8
+ 7.1

Pan. in !0
-0.4
- .2
-1.0
Vim- ~~ T^aaa

Pans in IO
+ 0.4
.0
-.4

Parti in 10"
-0.2

The values of y' in tables 2 and 3 can be used to

determine the functions which relate capacitance to
voltage. If we assume the function
Ck= Co(l +

(31)

where CV is the capacitance at voltage V and Co is the

capacitance at zero voltage, then the values of y' in
each column of tables 2 and 3 will be in the ratio of
23m:2*'":2":l. Values of m were calculated from the
measured values, yiooA. "yiooB* "yioooA. etc., and also
from the measured differences, yiooA- 7ioob. yioooA
yioooB< etc., which had smaller uncertainties of essen
tially different origin (see sec. 4.1). The results of
the two types of calculations differed only slightly.
The calculations yielded m = 3/2 for capacitors 100A,
100B, and 100C and m2 for capacitors 1000A,
1000B, and 1000C.
As a partial check on the overall measuring system,
the value of yiooo = (yioooA + yioooB + yioooc)/3 from 100
to 200 V was determined by two methods: (1) Indi
rectly by measuring the difference, yiooo- fioo, as
described in section 4.2, (2) directly by using capac
itors 1000A, 1000B, and 1000C in bridge 3 of the
three-bridge-circuit (sec. 3). The results differed by
less than 3x lO"10.

Capacitors were chosen from more than two dozec

capacitors on the basis of stability ascertained bi
difference measurements, similar to those describee
in section 4.1. The voltage dependence of capacitor?
100A, 100B, 100C, 1000A, 1000B, and 1000C was founc
to be considerably more stable than that of the other
capacitors. The measurements were limited to lOOpf
capacitors and 1000 pF capacitors. The use of the
drifter circuit (sec. 4.1) eliminated the need for spe
cial temperature control of the capacitors.
The earliest measurements involved three commer
cial 100 pF air capacitors of parallel plate design
The values obtained from difference measurement
were observed to depend not only on time but also nr.
the voltages which had previously been applied tt
the capacitors. In addition, a sudden, nonreversible
change in capacitance of 14 ppm was observed in one
of the capacitors when the applied voltage was in
creased beyond 50 V rms. The cause of the unusual
behavior was found to be metallic whiskers on the
tin-plated housing of the capacitor.
Difference measurements were also made on a set
of ten 100 pF capacitors of cylindrical design. The
construction of these capacitors is described in ref
erence [2]. The values obtained from the difference
measurements, corresponding to a change in voltage
from 80 to 160 volts, were observed to change by a?
much as 3 X 10~8 over a period of a few days. Change?
as large as 3 X 10~8 were also observed when the tem
perature of the capacitors was increased by approxi
mately 10 C.
A different phenomenon was observed when dif
ference measurements were made on a set of three
100 pF capacitors of cylindrical design which were
evacuated during measurement. The time rate ot
change of capacitance was found to be a function at
voltage. The phenomenon is believed to be re'-ted
to outgassing of the epoxy which was used to insulate
and support the electrodes. Measurements of the
time rate of change of capacitance as a function ol
voltage were made as follows: (1) The bridge was bal
anced and the drifter circuit was adjusted so as to
obtain a steady bridge balance, (2) the detector was
shorted, and the voltage applied to one capacitor was
reduced to zero for 30 sec, (3) the original conditions
were restored and the change in bridge balance was
recorded. With 160 V originally applied to the rapacitors. the changes in bridge balance were equivalent
to proportional changes in capacitance of about 1 x 10"".
Difference measurements, corresponding to i
change in voltage from 100 to 200 V, were made or
three commercial 1000 pF capacitors of parallel plate
construction. The changes in the values obtained
from the difference measurements varied from a fe*
parts in 109 to 2 X 10-8 over the period of one week.
The basic design of capacitors 100A. 10OB. and
100C is shown to approximate scale in figure 9. Crit
ical radii differ by approximately 3.2 mm. Line and
detector electrodes are separated from the ground
electrodes by four sets of three glass spacers. The
force of the compressed springs is about 45 N.

252-273

0 1 2 3 4 5 CENTIMETERS

Since the mechanisms which cause instabilities in

the voltage dependence of capacitors may also result
in general instability, the present work has pointed
out certain problem areas which should be considered
in the construction of precision standards of capaci
tance. In addition, the present work has served as a
proving ground for bridge measurements and voltage
ratio measurements at the higher accuracy levels.
The author acknowledges the many suggestions of
R. D. Cutkosky and the assistance rendered by D. N.
Homan and Lai H. Lee in obtaining the numerical
results of this paper.

7. References
ELECTRODE
FIGURE 9. Cross section through the axis of capacitor 100A, 100B,
or 100C.
It was found that after the capacitor was assembled
and connected into a measuring circuit, very small
radial displacements of the line electrode could be
obtained by sharply tapping its base radially. This
procedure was used to adjust the capacitance very
close to a minimum, thus assuring that the radial
components of the electrical forces were small. The
design of the capacitor is such that it can easily be
disassembled for cleaning or for experimental pur
poses. A grounded housing (not shown in the figure)
is used to evacuate the capacitor.

[1] A. M. Thompson and D. C. Lampard. A new theorem in elec

trostatics and its application to calculable standards of ca
pacitance. Nature 1 77, 888 1 1956).
[2| M. C. McGregor, et al.. New apparatu- at the National Bureau
ol Standards lor absolute capacitance measurement. IRE
Trans. Instru. 1-7, December 1958.
|3| N. 1-. Kusters and O. Petersons. The voltage coefficients of pre
cision capacitors. IEEE Trans. Communication and Elec
tronics 82 1 196.il.
|4| R. I). Cutkosky and J. Q. Shields. The precision measurement
ol transformer ratio*. IRE Trans, lnstr.. 1-9 (Sept. 1960).
|5| A. M. Thompson, The precise measurement of small capac
itances, IRE Trans. Instr. 1-7 (Dec. 1958).
|6| A. M. Thompson. A C bridge methods for the measurement of
three-terminal admittances, IEEE Trans. Instr. and Measure
ment (Dec. 1964).
[7] R. D. Cutkosky, Active and passive direct-reading ratio sets for
the comparison of audio-frequency admittances, J. Res. NBS
68C (Engr. and Instr.), No. 4 (Oct.-Dec. 1964).

6. Conclusion
The basic purpose of the present work has been to
develop an accurate system for measuring both the
relative and absolute voltage dependence of admit
tance standards. The causes of voltage dependence
have been pursued only to the extent necessary to
reduce instabilities. Additional measurements will
be required to evaluate the instabilities which remain.

253-274

(Paper 69D4-207)

JOURNAL OF RESEARCH of the National Bureau of StandardsC. Engineering and Instrumentation

Vol. 64C, No. 1, January-March 1960

Capacitor Calibration by Step-Up Methods

Thomas L. Zapf
(October 8, 1959)
Step-calibration methods are used in many physical laboratories for the extension of
measurements to quantities far removed from the magnitude of greatest accuracy at which
absolute determinations are made. The excellent precision of repetitive substitution pro
cedures is exploited by step-up or step-down methods to extend measurements to higher or
lower magnitudes without serious degradation of accuracy. The application of step-up
techniques to the calibration of variable air capacitors is described in this paper as a practical
example of the method.
1. Introduction
One of the important statutory functions of the
National Bureau of Standards is the calibration of
physical standards of measurement used in science
and industry. The chain of measurements con
necting this calibration service to the national
prototype standards of length, mass, and time is
complex, and for electrical measurements involves
meticulous experiments to assign numerical values
to calibration standards and corrections to standard
instruments. These devices, designed for excellent
stability and definitude, serve as comparison stand
ards basic to the calibration services rendered by
the Bureau. Equally important to accurate scien
tific work is the proper use of these standards to
overcome their inherent limitations. Frequently an
appropriate choice of method and the employment
of suitable techniques are as important as the
judicious selection of equipment. The close associ
ation between methods, techniques, and equipment
is particularly evident when, in the course of cali
bration activities, it is necessary to obtain accurate
measurements at magnitudes far removed from that
at which absolute determinations are made. The
extension of range of electrical measurements is
sometimes accomplished by the establishment of
accurately known ratios. For example, ratios very
nearly equal to the squares of integers may be ob
tained through the successive measurement in series
and parallel of resistors having nearly equal values
[l].1 Resistance ratios of approximately 10:1 may
be realized by the successive measurement of 1 1
resistors in arbitrary units and the use of these as
the 10:1 ratio arms of a bridge. A unique 10:1 ratio
apparatus used with a special resistance bridge is
described by Wenner [2].
The building-up to ratios larger than 10:1 is
particularly well exemplified by the procedure
followed in calibrating the standard volt box at the
Bureau [3] in which a group of sections of nominally

Figures in brackets indicate the literature references at the end of this paper.

equal resistance is intercompared. These sections,

connected in series, form the first section of a group
of larger denomination. The buildup to large ratios
is rapid and exact. The standard volt box was
designed specially for self-calibration by this method.
The calibration of resistance decade boxes and
the resistance decades of bridges by the stepsubstitution or step-up method illustrates yet
another technique of obtaining accurate measure
ments over wide ranges [4). A similar process is
used by the Bureau for the calibration of the capaci
tance bridges that are used daily to measure stand
ards of capacitance.
In order to obviate the concern over connection
errors and avoid the detailed consideration of con
nectors, it is customary and convenient to use as
standards of low grounded capacitance such devices
as variable air capacitors and capacitance decade
boxes which may be calibrated accurately for capaci
tance difference from some arbitrary setting. The
calibration of such variable capacitors may be
accomplished quite effectively by the step-up
method employing fixed standards or standards of
capacitance difference.
An excellent description of a step-up method
applied to the calibration of decade capacitors for
both capacitance and dissipation factor has been
described by Ford and Astbury of the British
National Physical Laboratory [5].
2. Equipment
Very little special equipment is needed to calibrate
a variable capacitor by step-up methods. If the
variable air capacitor, X, having a range from 100 to
1,100 pf, is to be calibrated at every 100-pf division
mark, it is necessary to have a fixed air capacitor, S,
of approximately 100 pf that can be connected in
parallel witli the variable capacitor under test in a
precisely repea t able, manner. This can be achieved
if the mating connectors introduce no significant
uncertainties to the capacitance added to the circuit
and if the connectors are designed to couple quickly
and easily to either the variable capacitor or the

Z54-75

jridge that will be used, or if a capacitor can be con

nected or disconnected by a precise switching ar
rangement. The 100-pf capacitor should be adusted close to the nominal value, but it need not be
calibrated. It must, however, be free from signifi
cant drift over the period of a quarter-hour or so
luring which the test is being run.
A calibrated 1,000-pf air capacitance standard, S',
s needed to relate the results of the step-up test to
the national reference standard of capacitance.
The bridge used for this step calibration need not
have great accuracy but must be stable, for it is used
with a sensitive detector for substitution measure
ments. A small variable capacitor, T', is required,
having a least count (smallest readable increment)
one-tenth that of X or smaller. It is advantageous
to choose the smallest possible precision variable
capacitor, V, consistent with other limitations so
that the corrections to V arc negligibly small relative
to the corrections to A'. The total range of V must
be at least a little larger than the range of errors in
the capacitor to be calibrated. The readable
accuracy of this capacitor, if expressed in percent of
total range, need not be very great.
The equipment described is assembled as shown in
figure 1. It is most important that the cables used
to connect components be shielded and rigid, or if
flexible cables arc used, it should be ascertained that
variations in cable capacitance are negligible. The
cables must be fixed in position and must not be
disturbed during the entire calibration. This pre
caution is intended to emphasize the importance of
particular care to one of those sources of systematic
error that could impair good calibration accuracy.
The operator must have a good technical apprecia
tion of the apparatus and quantities measured,
gained tlirough study and experience.
BRIDGE

where x is the correction to the reading (or setting)

Xn. The calibration will consist of the determina
tion of the relatively small correction, x, to each
100-pf division mark, Xn, and since only capacitance
differences are of interest, the observer is free to
choose any one of the division marks as a reference
point. It is frequently convenient to choose as a
reference the first marked point on the dial. In this
case the 100-pf mark is considered as a reference and
a correction of 0.00 is arbitrarily assigned to it.
For this example it will be assumed that the small
variable capacitor, V, has corrections that are
negligibly small.
The capacitor under test is first carefully set to
the 100-pf mark (A^^lOO pf) avoiding backlash
errors by approaching the mark in the direction of
increasing dial readings. The small variable capaci
tor, V, is set to any convenient mark near the center
of its range. The 100-pf capacitor, S, is connected
in parallel with A' and V. Now the bridge must be
balanced using the controls on the bridge itself. If
a balance cannot otherwise be obtained, V may be
used to attain balance. When the bridge is balanced
the reading VA is recorded. The fixed capacitor,
S, is then removed and X set to the 200-pf mark
(^,2=200 pf), again approaching the mark in the
same direction. Without changing any other com
ponent the circuit is rebalanced by changing V
alone, and the reading VB recorded as before. In
the first balance the external bridge arm consisted
of S+Xi+VA, and for the second balance, with the
bridge unchanged, the external arm consisted of
X2-\-VB. These can be equated to yield
S+X + VA=X2+VB

(1)

The cable and connector capacitance, as well as resid

uals within the bridge, contribute equally to both
balances and are therefore deliberately disregarded.
It is convenient to work with small numbers, and
eq (1) can be expressed as
Sn+s+X1+x1+VA=Xn2+x2 + VB

(2)

and since

Figure 1. The variable capacitor under text, X, is calibrated

by a step-up method employing a Axed capacitor, S, and a
small variable capacitor, V.
3. Procedure and Computations
The true value of each of the capacitors involved
in the calibration may be defined as the nominal
value plus a correction; thus, the true capacitance
of the uncalibrated 100-pf capacitor is S=Sn+s,
where S,,= 100 (exactly) and s is the small correction.
Similarly ihe calibrated 1,000-pf standard has the
value S' = S' + s' , and the variable air capacitor to
be calibrated may be represented by X=Xn+x,

S=Xn2-Xnl and ^, = 0

(3)

x2=(VA-VB)2+s

(4)

where the subscript is appended to the difference

(VA VB) to distinguish this set of data from other
sets and to appropriately correlate the difference
with the setting of X in the second balance of each
set.
The quantity x2 is the desired correction to Xu
when Xn=200 pf. The difference (VA-VB)2 is
easily computed from the recorded data.
Without changing AT, S is reconnected and the
bridge rebalanced using the bridge controls and V,
if necessary, to attain exact balance. The reading,
VA, is then recorded. Capacitor 5 is then removed
and X set to the 300-pf mark (A', = 300). The
bridge is rebalanced using V alone and the reading,

255 -76

VB, recorded. When the first balance is equated

to the second balance
S+X2+VA=X3+VB

(5)

Sn+s+Xn2+x2+VA=Xn3+x3 + VB

(6)

or

in parallel with X and V. With X set at 10 0 pf. the

bridge is balanced with the bridge controls and V.
if necessary, and the reading, VA, recorded. S' is
removed, X is set to the 1,100-pf mark, the bridge
rebalanced using V alone, and the reading, ^ s,
recorded. Then

and since

S'+X1+VA=X11+VB

(12.1

S:+s'+Xnl + VA=Xnll+xn+ VB

(13)

S'n=Xnn-XHl

(14)

xtl=(VA-VB)r+s'

(15)

or
Sn = ^-n3 Xn2

(7)
and since

x3=x2+(VA-VB)3+s.

(8)

Substituting eq (4) in eq (8)

Xt=(VA-VB)3+(VA-VB)t+2s.

(9)

Continuing this process step-by-step, in general

for the mth step
*=S (VA-VB)+(m-l)8
2

(10)

*.,=i: (vA-vB)+ios
2

en)

where the subscript T denotes the V difference ob

tained when the capacitor, S', is used.
In this way xu is determined accurately in terms
of a small difference reading of the variable capacitor.
V, and the known correction, ', to the standard
capacitor, S'. The correction, s, to the fixed capaci
tor, S, can now be computed from eq (11)

and finally
10fi=xu- (VA-VB)
2

(16)

or
A tabulation of the differences and the cumulative
sum of the differences is shown in table 1 , which shows
the data and computations for a typical calibration.
It remains to determine the value of s so that the
corrections x2 through xn can be evaluated.
The 1,000-pf standard capacitor, S', accurately
calibrated for insertion capacitance, is now connected

Table 1 . Observations and calculations

All values In picofarads
s

X
100
0
100
0
100
0
100
0
11)11
0
100
0

0
100
0
100
0
11KI
0
1000
0

100
200
200
300
300
400
400
500
500
600
600
700
700
800
800
900
900
1000
1000
1100
100
1100

Va-Vb z(iv-v)

5.00
4.69
4.84
4.86
4.99
4.93
4.92
4.51
4.65
4.64
4.81
4.76
4.87
4.60
4.73
4.42
4.57
4.53
4.67
4.12
5.04
4.12
* = -0.37pl

nt

0.S1

0.S1

-.01

.19

.OS

.35

-.43

-.08

.41

.76

-.58

18

.01

.77

-.7t

.05

.05

.8i

-.86

-04

.17

1.09

-1.01

.08

.31

1.40

-1.15

.15

04

1-44

-1.30

14

.55

1.99

-'44

.55

0.91
10> = -l.Upf

-0.14

0.17
.00

10s=s' + (VA-VB)T-

(Va-Vb)

(17)

2
The quantity 10s is then added algebraically to the
sum of the V differences corresponding to the test
of the 1,100-pf mark, the result being the correction
to this reading. Similarly 9s is added to the sum
of the V differences corresponding to the 1,000-pf
mark, and so on, until only s is added to the Y dif
ferences corresponding to the 200-pf mark. These
small corrections are listed in table 1 under the
heading ns.
The observations can be made rapidly and the
computations are simple, since only small differences
appear. A second complete calibration, preferably
by another observer, enables one to appraise the
precision of the measurements including the stability
and resetabilitv of the capacitor under test, and
serves to reveal measurement and arithmetic errors
that might otherwise remain undetected.
In the procedure described above the fixed incre
ment of calibration was 100 pf. It is well to point
out that other increments can be accommodated
as well. A 50-pf capacitor, if used as a fixed step,
would permit calibration at 50-pf intervals. Al
though the procedure has been described using a
fixed capacitor as a step, a continuously variable
capacitor or decade capacitor would also be satis
factory if it were used in such a manner as to provide
a repeatable difference of capacitance. Care would
be necessary to avoid setting errors caused, for
example, by backlash in the control mechanism,
or by careless setting to the index.
A variable capacitor calibrated for capacitance
difference by this method can be used as a standard
for extending the method to capacitance calibra
tions of still lower magnitudes.

256 -77

BRIDGE

4. Dual Calibration
Reviewing the calibration described above, it is
noticed that for each set of two balances, one bal
ance is obtained with the bridge controls and V, if
necessary. The fact that the change in the bridge
reading is always an amount approximately equal
to S (or S') leads to the consideration of calibrating
two variable capacitors having the same range with
practically no extra work.
If X and U, the variable capacitors to be cali
brated, are connected as shown in figure 2, the pro
cedure is similar to that described above except
that the bridge need not be changed after the initial
setting. The settings of Y and U are listed in
table 2. Care must be taken to apply the proper
sign to the differences and to cumulatively add the
differences for the calibration of U beginning at
the bottom of the table rather than the top.
Lower range capacitors can be calibrated similarly,
but extreme attention must be paid to good mechani
cal rigidity in all parts of the circuit.

Figure 2. Two variable capacitors, X and U, may be cali

brated simultaneously by a step-up method.
5. Discussion
If attention is confined to the calibration of twoterminal variable air capacitors having a capacitance
range from several picofarads to about 1,000 pf, the
procedure outlined in this paper demonstrates the
ability of step-calibration methods to provide ac
curate calibrations of capacitors at levels at which
good accuracy is otherwise difficult to obtain.

Table 2. Observations and calculations (dual calibration)

All values In picofarads
s

0
100
1)
100
0
100
0
100
0
100
II
lll
II
100
0
100
0
100
0
100
0

100
100
200
200
SOU
300
4011
4011
900
500
600
aoo
700
7IKI
Kim
KtK)
two
WMI
1000
1000
110U

1100
1000
1000
WOO
wm
(UK)
sou
TIKI
700
600
mo
sou
500
400
41III
900
300
200
200
100
100

5.00
5.34
5.03
5. 07
.V 111
.V !7
5.11
.vas
4. UK
5.34
5.32
5.63
5.58
5. Kt
5. .18
5.87
5.56
5. 57
5. 54
5. 67
5.13

I0U0
0
won

100
1100
1100

1100
1100
1110

5.37
4. 41.
5. 50

'--<U7 pf

Va-Yb X(Va-

Yb-Va Z(l'Va)

0. 31

0. 31

-o.u

0.17

-.03

IS

-..'

.00

.0ii

.Si

. 41

-.07

.37

.71

-..55

.16

.0!

.73

.69

.04

.OS

.78

-.83

-.US

. te

1.04

-.97

.07

.31

1.35

-1.10

.IS

.03

1.38

-un

14

1.91

-1.38

Si

tu

0.34

t.06

-1.38

0. 67

.04

1.71

-in

.47

.07

1.67

-1. 10

.57

.14

1.60

-0.97

.63

.36

1.36

-.83

.53

.31

1.00

-.69

.31

.in

0.69

-.55

.>9

*3

-41

.0!

.01

-.18

-n

.13

.13

-14

-.01

.91
1.04
KU--/.JS pf

10 = -/.}S p[

The columns r and u arc the corrections to the dial readings of the variable capacitors Xand U. The good precision of the method is noticeable by comparing the
test of X in table 2 with table 1, which represents a test of the same capacitor about an hour earlier.
257-78

Standards of grounded capacitance, often called twoterminal capacitors, are characterized by having one
of the capacitor electrodes connected to the case, in
contrast with standards of direct capacitance (threeterminal capacitors) having both capacitive elec
trodes insulated from the case. The direct capaci
tance, CD, between the two active electrodes, as
shown in figure '.i, is definite to the extent that the
separate terminals and associated leads are shielded
from each other. Adequate shielding that does not
interfere with the direct capacitance is relatively
easily obtained, and excellent accuracy in direct
capacitance measurements is possible to a fraction
of a micropicofarad (10-18 f).

Figure '.i. The direct (three-ttrminal) capacitance, Co, is

made definite by complete shielding. The grounded (twoterminal) capacitance, Co, is indefinite because of variations
in the stray capacity.
The grounded capacitance, Co, shown in figure 3,
is more difficult to define in a precise manner, be
cause the capacitance between the ungrounded
terminal and all grounded objects, Ct, is indistin
guishable from Co unless a separate "zero-balance"
of the measuring apparatus is made with the leads
attached but the capacitor disconnected. Even this
procedure will not insure good precision unless care
is used to connect the capacitor to the measuring
apparatus in identically the same way every time.
The best precision in practical measurements of
grounded capacitance is possible only if the method
of connection to the measurement circuit is well de
fined. An adequately shielded rigid adapter or
connector is necessary as an auxiliary part of the
capacitor and must be used with it for every meas
urement. If the capacitor is to be used as a stand
ard for accurate capacitance measurements, the same
connector must be used with the capacitor when it
is calibrated, and the assembly becomes a standard of,
capacitance added to a circuit, or in other words,
capacitance difference. Thus, good precision of re
peated measurements is simply obtained in any
laboratory if the successive measurements are accom
plished using rigid wiring and the same connectors
every time. The best accuracy in terms of the na
tional reference standards can be obtained only for

magnitudes sufficiently large that negligible errors

result from differences between the connectors used
in the several laboratories involved. Accurate cali
bration of small fixed capacitors can only be accom
plished if mating connectors are submitted.
The accuracy of measurements on fixed standards
of grounded capacitance having electrodes termi
nated in binding posts or unshielded plugs is limited
by the variation in the geometrical design of the
instrument panels, cables, and connectors to which
the standard can be attached. Differences as large
as several tenths of a picofarad are possible with
present commercially available standards and appa
ratus with which they may be used. It is under
standable that differences of several tenths of a
percent are to be expected if a fixed 100-pf standard
of this type is measured in several laboratories or on
different equipment, while if measurements are
performed on 1,000-pf standards the uncertainties
at the connectors would be only several hundredths
of a percent of the quantity measured. In the
step-up procedure described in this paper it is evident
that the precision of repeated measurements, the
freedom from the effects of residuals in test apparatus
obtainable, by substitution methods, and the ac
curacy of measurements at magnitudes closer to
optimum, are combined in a manner favorable to
the accurate calibration of the capacitance differ
ences of variable capacitors. The method is quite
applicable at any frequency although at higher
frequencies residuals in components can be trouble
some and may require special attention. For
example, it may be necessary to apply corrections
for errors introduced by residual inductance in the
cables connecting the apparatus.
6. Conclusion
Step-calibration methods can be employed for the
calibration of variable capacitors. The few neces
sary items of equipment are generally available in
tiny electrical measurements laboratory. Reference
to the national electrical standards is made through
the use of a single fixed capacitance standard that
can be transported to other standardizing labora
tories more easily, and calibrated less expensively
than variable capacitors.
7. References
[1] Lord Rayleigh. Phil. Trans. 173, 661 (1882).
[2] F. Wenner, j. Research XBS 25, 252 (1940). RP 1323.
[3] F. B. Silsbee and F. J. Gross, J. Research XBS 27, 261>
(1041). RP1419.
[4] J. L. Thomas, NBS Circ. 470 (1048).
[5] L. H. Ford and N. F. Astburv, J. Sci. Instr. IS, 122 (103S1.

Boulder, Colo.

258-79

(Paper 64C1-27'

JOURNAL OF RESEARCH of the National Bureau of StandardsC. Engineering and Instrumentation

Vol. 65C, No. 3, July-September 1961

Calibration of Inductance Standards in the Maxwell-Wien

Bridge Circuit*
Thomas L. Zapf
(February 28, 1961)
This paper discusses the errors from residuals in the Maxwell-Wien bridge and the
effect of these on the measurement of inductors in a bridge not having a Wagner ground.
Particular attention is given to the use of substitution methods for accurate measurements
and especially to the "equal-substitution" (comparison) method, which can yield excellent
precision in the calibration of inductance standards.

1. Introduction
Self inductors and mutual inductors can be con
structed to have inductance that is computable from
their geometry and dimensions. Although the in
ductance of certain inductors having measurable
mechanical dimensions and simple geometrical form
can be computed with excellent accuracy, such
inductors are not often used as reference standards
for inductance measurements. In practice, induct
ance measurements are more conveniently made with
reference to noncomputable reactances in the form
of compact and stable reference standards of induct
ance or capacitance. Accurate values of inductance
can be determined by comparing the reactance of
the inductor with the reactance of either standard
capacitors or standard inductors bv use of appro
priate bridge circuits. However, the most precise
measurement of an unknown quantity (in this case
inductance) is made by comparing it with a standard
of like kind and magnitude, the small difference being
measured by a corresponding small change in one
element of the bridge circuit. When comparison
methods are used, it is often possible to reduce the
detrimental effects of residuals in the measuring
circuit to such an extent that the precision of com
parison is much better than the accuracy with which
the calibration standard is known.
The Maxwell-Wien bridge circuit has long been
used for the accurate measurement of inductance.
For measurements of best accuracy with any alter
nating-current bridge circuit, even for comparison
measurements, it is necessary to consider the effects
of residuals, the ways of reducing these effects, and
the handling of corrections to onset the net errors.

standardizing laboratories for measurements of in

ductance. Figure 1 shows a schematic of a wellshielded Maxwell-Wien bridge with Wagner arms for
elimination of errors resulting from current diverted
to ground through leakage, impedances. The in
ductance, L, to be measured is placed in series with
a resistance, rL. If the components of the bridge
arms shown in figure J were pure, the inductance
measured when the bridge is balanced would be
L=CRPRS,

(1)

and to obtain the balance it is necessary that

2. Maxwell-Wien Bridge
A bridge circuit originally developed by J. C.
Maxwell [1] 1 for use with ballistic detectors was
adapted by M. Wien 12] for a-c measurements. For
many years the circuit has been used widely by
"Contribution from the Radio Standards Laboratory, National Bureau of Figure 1.
Standards,
Boulder,
i Figures In
bracketsColo.
indicate the literature references at the end of this paper.
259-183

The Maxwell-Wien bridge circuit with complete

shielding and a Wagner ground.

In practical situations it is not possible to obtain

completely pure components. The residuals in the
components contribute to the systematic errors, and
it is necessary to analyze the circuit sufficiently to
correct for these errors or to determine the maximum
uncertainty in the results if the errors are neglected.
Figure 2 shows the components of figure 1 with t lie
addition of appropriate residual components, but
without the Wagner arms. For example, the series
resistor, rL of figure 1 , in the inductance arm is repre
sented in figure 2 by a pure resistance, rL, in series
with an inductance, lL. The inductance, lL, is a
lumped constant that may be either positive or nega
tive depending upon the relative magnitudes of
inductive and capacitive effects. Although lumped
constant residuals are not independent of frequency,
the lumped constant concept provides a useful
equivalent circuit at low frequencies. The other
resistive components are similarly represented by a
pure resistance and an effective series inductance.
The variable capacitance, C, is presumed to contain
an effective series resistance, r. The resistance
RL in figure 2 is the equivalent series resistance of the
inductor.
3. Equation of Balance

Figure 2.

Ip TP,
RP~

ll-

Ll Ta.
Rs~

and

The impedances of the four bridge arms are

\(R+jo>I)(l+<fW)
" 1 +<Si3C2 +w( \<arC+j) (R +jul) '

The Maxwetl-Wien bridge circuit showing Uu

uals in the arms as lumped conxtants.

(3)

Zp=liP+jUP,

(4)

Zt=liL+rL+ML+lL),

(5)

With negligible error it will be found that

L = CRpRs[ l-afi(TpTs + TRTp+TRTs) -D*c
+ u(Tp+Ts + TB)Dc b^TPTsTRDc]

and

-(RL+rL)rR(6)

The derivation of eq (3) is given in the appendix.

When the bridge is balanced
ZrcZl ZPZS.

RpRs
(tp+ts)-Il,
R

(13

and a convenient simplification results if the snull

terms in brackets are replaced by the svinbol A"
Then the equation of balance becomes

(7)

After the multiplication indicated by eq (7) is per

formed, t lie equation of imaginary components can
be written

L= CRPRs(l +K) - {RL+rL) rR+ RpR.S Oj>+tsW.

R
(141
where

(1 + W*)[fl(L+Zi) + (ItL+rJl-HJp-Rrls]

K= O)1(tPTs+ TRTp+ THTS) D%

+ 0)(Tp+Ts+TR)Dc<-pTpTsTRDc.

= C[RRpRs+w(RRpls+RRJp+NpR^)o,rC
-^(Rlpls+ Rplls + RJIp) - Mpl&rC] .

(8)

Considerable simplification of eq (8) is possible if

the residuals are expressed as time constants and
dissipation factors, as follows. Let
wrC=Dc,
I
TV

(9)
(10)

(15

4. Effect of Stray Capacitance

Before analyzing this equation in detail it is profit
able to consider the effect of stray capacitance acres
the terminals of the inductor. Such capacitanc*
can be introduced by the leads connecting tin
inductor to the bridge (see fig. 3). For this analysts,
whatever capacitance exists within the case of tit1
inductor is considered not a part of C.s, but rathtr
part of the inductor. It should be understood that
this internal shunt capacitance contributes to tbt

260-184

(0)

(b)

FiiiURE 3. The inductor under lent, Lt, is connected to bridge

terminals with leads having stray capacitance, Cs, between
them.
The bridge mcusuros the equivalent inductance, which ditTers from Lt
because of the stray capacitance.
effective series inductance measured at the terminals
of the inductor, thereby causing the effective series
inductance to vary with frequency to some extent.
Figure 3a shows the equivalent series inductance
under test, Lt, in series with the internal resistance
of the inductor, RT. The stray capacitance, Cs,
external to the inductor is in parallel with LT and
Ilr. The bridge measurement does not completely
separate these elements but indicates the effective
inductance, L, as shown in figure 3b. The effective
series inductance, LT, at the terminals of the inductor
is related to the measured value, L, at the ends of
the leads by the approximation
LT~L(l-u>LTCs+!^+o>tRTC%\-

(16)

The derivation of eq (16) is given in the appendix.

The last three terms in parentheses constitute the
{mail correction terms resulting from stray capaci
tance. The first and the last of these vary as the
square of the frequency. The last term is usually
nuch the smaller of the two in practical situations.
The second correction term is not dependent upon
frequency and must never be overlooked as a source
jf error in measurements of the highest accuracy.
For example, in the measurement of a 10 h inductor
saving an internal resistance of 10,000 ohms, at the
;nd of a coaxial cable having a capacitance of 100 pf,
jhe error from neglecting the second correction term
a 0.01 h, equivalent to 0.1 percent of the quantity
measured. Often the first correction term is sigmficant while the other terms are negligible, and the
jffective series inductance then can be written
LT~L(l-<*IrCs).

(17)

However, in this paper both the first and second

jorrection terms will be carried in the analysis to
'ollow.
5. Substitution Methods
The existence of the residuals and the consequent
;orrection terms in eqs (14) and (16) are detrimental
jo good accuracy for residuals are often very difficult
:o measure or estimate to better than an order of
magnitude. To eliminate the effect on measure
ments of some of these residuals, substitution

methods are used. For example, if a standard

inductor of equal nominal value is substituted for the
test inductor and the difference of inductance is
measured by changing C and rL (or R) slightly, then
the residuals in the resistance arms of the bridge have
negligible effect on the measurement. Also, the
effect of the stray capacitance across the inductor
under test or across the standard inductor is much
reduced. The standard inductor must be very
accurately known, however.
If accurate measurements of large inductors are
contemplated and a shorting plug of negligible, or
relatively small, inductance is substituted for the
test inductor, Cs must be reduced to a negligible
value by complete shielding, and a Wagner ground
should be used. The effect of the residual, lL, in the
resistor, rL, is eliminated by this method only if lh is
the same for the two conditions of balance.
The latter method, in which a relatively small
inductance replaces the inductor under test, can be
designated the "zero-substitution" method to dis
tinguish it from the "equal-substitution" method,
in which the test inductor is replaced by a standard
of equal nominal inductance.
A general formula for inductance measurements by
substitution methods with the Maxwell-Wien bridge
may be derived from the two balance equations,
the first with the inductor, LT, in the circuit, and the
second with a standard, Ls, in its place. If the
small correction terms represented by K in eq (14)
are constant for both conditions of balance and if
(\ is changed only slightly (to Cay) when the standard
inductor or shorting plug is connected, then with
certain justifiable approximations
LT=LN+ (C- C)RPRS[1 - u2(LT+L)Cs+K]
uPLtLn (Cs Cs.v)
+ (RT-R%)Cs-(lL-lLN).

(18)

The derivation of eq (18) is given in the appendix.

This equation is useful in showing the effect of un
certainties in the magnitude of circuit components
on the determination of inductance.
If good quality components are used and the bridge
is well designed, the time constants that contribute
to K will be small. However, K cannot be deter
mined with certainty and is somewhat variable
(dependent upon actual circuit constants). The
uncertainty in K sets an ultimate practical limit to
the accuracy of measurement of inductance by the
Maxwell-Wien bridge.
6. Zero-Substitution Method
When the zero-substitution method is employed
the small inductance, LN, need not be accurately
known, but the capacitor, C, and the resistance
product, RPRS, must be known witli better accuracy
than that desired for the inductance, LT. It is
possible to construct very small inductances that are
computable from geometry and dimensions. A short
ing bar between two terminals has a finite inductance

261 -185

that can often be calculated or estimated well enough

to serve as the "zero-reference", LN, for bridge meas
urements of inductors of much larger magnitude. If
the measured inductance is defined as the increase
in inductance when a short ing-link or switch associ
ated with the inductor is manipulated in a specified
manner, then the actual inductance of the short
circuit, Ln, is immaterial. The term (IlIln) ean
be minimized if the resistance, rL, is inductance com
pensated; i.e., if designed so that the effective series
inductance remains constant as the resistance is
varied. The uncertainties in this term may limit
the accuracy of measurement of small inductors. If
Cs is not extremely small, it is evident that the
uncertainty in 6s can have a significant effect on
the accuracy of measurement by this method. In
cq (18) the term in brackets containing C's can
become objectionably large when large values of
inductance are being measured. For example, if the
inductance being measured is approximately 10 h,
if rs=IOO pf10 pf, and if = 10,000 radians per
second, then <o,//',s=0.10 0.01. Thus the correction
term is 10 percent of the measured inductance, and
the measurement uncertainty is 1 percent from this
cause alone, disregarding the probable lack of validity
of the assumptions (when the correction terms are
so large) that were used to derive eq (18). The effect
of capacitance Ca thus places a definite limitation
on the accuracy which can be obtained by the zerosubstitution method with a grounded bridge. In
order to obtain better accuracy when o>2/, is large
it is necessary to use the "equal-substitution method"
(described below) or to resort to separate shielding
on the inductor leads with an ungrounded bridge
using Wagner arms. This latter possibility involves
considerations beyond the scope of this paper.
7. Equal-Substitution Method
If an accurately known standard inductance, LN,
is substituted for the test inductor, LT, the capaci
tance, C, and the resistance product, liplls, need not
be so accurately calibrated. Between the two con
ditions of balance the only circuit components
changed are C and rL, botli by small amounts. The
residuals Dc and lL are essentially unchanged in this
case, and all the other residuals are identical for botli
conditions of balance. For the equal-substitution
method the general eq (18) can be rewritten with
negligible additional error as
Lr=L+(C- Cv)HPNs(l - WLT( s+K)
-<sl*t(Cs-cs) + (irr-n%)rs- (iL-iL.v).

tance of maintaining Cs nearly constant during

the substitution.
The term
ean be reduced to a ininiiuurc
by employing an adjustable resistor, rL, having in
ductance compensation. Uncertainties in this tern
may limit accuracy if the resistances of the test Mi
standard inductors differ greatly, and the inductance
compensation of the resistor, r,., is inadequate. Tht
adjustable resistor is in series with the inductor ami
is connected to the grounded corner of the bridt-f
as shown in figure 2. If, instead, this resistor wity
connected to t lie ungrounded end of the inductawr
arm of the bridge, the capacitance to ground from
the several decades of this resistor would shunt the
inductor, and as the resistance is varied to accom
modate inductors having different internal resistant
the change in stray capacitance would cause au error
that would be difficult to correct. If one end of tbt
resistor is grounded, and if care is taken to k> > j
the resistance as low as possible, the capacitance tu
ground within the resistor merely shunts the resist
ance, contributing to the effective residual in
ductance, lL, which is generally insignificant.
The term (K-T li\)('s deserves special attention
because it is directly dependent upon the effectm
series resistances of the inductors. As an example
of the effect of this term, consider that, in t he meas
urement of a 10 h inductor, /i'r= 10,000 ohms.
/tV=8,000 ohms, and G=100 pf. Then the tern.
(2r-/i^)rs=0.0036 h, nearly 0.04 percent of tht
quantity measured. This correction, being indrpendent of frequency (if skin effect may he neglected;
exists even at low frequencies, at which most of th
other correction terms are negligible.
If the time constants, tp, ts, and tr, are 1 >iser o'
less and Dc is 0.001 or less (these are conservativ
but reasonable estimates for commercially available
components), and the angular frequency is 10,000
radians per second, the magnitude of K is less than
0.0003. For a 1 percent difference in inductors. i(
this term were neglected the resulting error would
be only 3 ppni of the measured inductance.
Thus, the equal-substitution method can be em
ployed without the necessity of using a Wagner
ground providing care is taken to keep the strav
capacitance, Cs, reasonably small and nearly con
stant and to apply corrections to the measured
values to offset errors resulting from the different!
in the internal resistances of the inductors.
8. Effect of Frequency Difference

(is)

In eqs (18) and (19) the stray capacitances, Cs

and ('slt, are not considered to be exactly equal;
however, it is expected that they will be maintained
nearly equal. The importance of this, even in equalsubstitution measurements, is evident from eq (19).
In the comparison of 10 li inductors at an angular
frequency of 10,000 radians per second, if the differ
ence Cs('sx is only 1 pf, then w*Lt (Cs Cs\) is
0.01 h. This error of 0.1 percent shows the impor

It has been assumed that the frequency of the

a-c supply is stable. If the frequency, o>.v, whec
the standard inductor is in the circuit is not equal to
the frequency, u, when the test inductor is connected,
there can be an error resulting from the stray capaci
tance, Cs. A difference of frequency between tbr
test balance and the standard balance will necessitau
the addition of the term

262-186

CsLtLn (to2 u%)

o the right side 01 eq (18), and since oj^oo^ in any

radical case, this can be factored to produce
2uCsLTL!/(o) u)N) .
Since LT~LN for the equal-substitution method,
lie additional term to be added to eq (19) is
2uCsL%{tji coiV).

The resistance product (range) in the bridge is set

so that the difference between the test and the
standard inductors can be accommodated by adjust
ment of the externally connected variable capacitor
and resistor, the other bridge controls being left
unchanged.
The analysis of errors in the Maxwell-Wien bridge
circuit described in this paper was carried out as part
of the investigation of the feasibility of adopting the
equal-substitution method for the rapid and conven
ient measurement of inductance.

If oo=10,000 radians per second, oiN= 10,010

adians per second (a 0.1% difference), and 6*s=100
f, for the measurement of a 10 h inductor by the
qual-substitution method the magnitude of the
orrection term is 0.002 h, an error of 0.02 percent,
t is important to realize that this analysis of the
fleet of frequency variation does not account for
he change of effective series inductance with freuency resulting from eddy currents, skin effect, or
istributcd capacitance within the inductors.

The use of the equal-substitution method for the

precise comparison of inductors at the Electronic
Calibration Center, NBS, Boulder, Colo., was
instigated by Chester Peterson, NBS, Washington,
D.C. The author is indebted to Mr. Peterson for
helpful comments and suggestions pertaining to this
paper.

9. Calibration of Inductors

10. Appendix: Derivation of Equations

In recent years standard inductors having good

tability have become available in a wide range of
lominal values. These have made accurate measirements by the equal-substitution- method feasible
,nd convenient. Several complete sets of standards
re maintained in the laboratories of NBS botli at
Vashington and at Boulder, and are used regularly
or calibrating similar standards submitted for certiication. The values assigned to NBS working
tandards ultimately depend upon a computable
nductor or capacitor.
The equal-substitution method can be utilized
rith a variety of a-c bridge circuits. The essential
equirements are that the bridge have good shortime stability, adequate resolution, and means for
xternally equalizing the storage factor (Q) of the
nductors being compared. An advantage of the
qual-substitution method is the lessened need for
.ccuracy of adjustment of the bridge components
ecause the bridge is used merely to measure small
lifferences. A number of commercially available
ridges embodv the Maxwell-Wien bridge circuit to
vhich the analysis given in this paper is primarily
levoted. It is almost invariably necessary, however,
o improve their resolution by adding a calibrated
rariable capacitor in parallel with those built into
he bridge.
The resistor, rL, is partially inductance com>ensated and is connected in series with the cable
onnecting the inductors to the bridge. It has been
letermined that the existing variations of inductance
ire negligible relative to the inductance being
neasured by this method. This resistor may be
egarded as serving the purpose of externally
iqualizing the Q of the standard and test inductors.
It is seldom necessary to apply a correction for the
erm (R2TR%)CS in eq (19). On rare occasions
vhen this term is significant, a crude measurement of
?r and RK is adequate.

Equation (3) is derived from the equation of the

impedance of Zc and ZR in parallel.

315-411 O - 68 - 18

1 = 1
Zrc R+jwl

uC
wrCj

1
<*C(wrC+j)
R+jur o)V6a+l
_ 1 +(Jtrt(? +wC(wrC+j) (R +>>/)
(+i0(l+orV(7)
Therefore
K

1 +W+wC(wrC,+j) (R +jwl)

Equation (8) is obtained by substituting eqs (3),

(4), (5), and (6) into eq (7), giving
(l+<**&)(R+jl)[(RL+rL)+ML+lL)]

(Rp+jo>lp) (Rs+juls) ,
from which the equation of imaginary components is
(1

[R(L+lL) + (RL+rL) l-RslP-RPls]

=ClRRpRs+u(RRPls+RRslP+RpRsl)urC
-^{Rlpls+Rplls+Rsll^-wHlpl^rC].

Equation (16) describes the effect of stray capac

itance. Figure 3a represents an inductor, LT, having
an internal resistance, RT, connected to the bridge
terminals by a cable having shunt capacitance, Ca.

263 187

The impedance of this circuit, Z, is derived as follows

Z RT+juLT
(Rr+juLr) ( 1 -JLTCs-jwlirCs)
Z=( 1 o>2L tC's +jwtf / 'a) ( 1 wi rCs-jalt rCs)

the two conditions of balance; hence, the assump

tion that K is constant is justifiable. In the above
equations since
CRPRS~LT
and
( sRpRs ~ Ly
and
(J?*+n)(ff+itar)

7/ r+^r(l -ffLjCs) -li\Ca\

CRpRaRlCalLr^RlCs and
CyRrRRRyCsfllL.y~R'sCiS
In the equivalent circuit shown in figure 3b
Therefore,
LT-Ly=[(C-Cy)RpRs-RpRs^(.(,LrCs
and from the two preceding equations it is evident
that
,

-CyLyCsyWnCs-RyCsyW+IQ -

LT(\-u?L7f!s)-mCs
The factor (CLTCsCyLyCsy) can be expanded a?
follows:

and, with negligible error,

(CLTCs- CyLyCsy) = (C~ Cy) (LTCS+ LCaK)

^CyLrCs-CLyCss

i= (l-2w1rC's+'/^D '
It should be noted that the discarded term,
u>*LtCs, in the denominator is always much smaller
than the term, 2u}LTCs, which is usually much
smaller than 1 in practical inductance measurements
at low audiofrequencies. Since the terms in paren
theses, other than 1, are small, this equation can be
solved for LT and simplified by neglecting higher
order terms.

If Cs and CSy can be reduced to zero, this factor

becomes zero, and the equation is greatly simplify:
However, if ( a is not zero, but if Cs approximately
equals CSy,
CLTCS - CNLy< 'SN = {('-(- -s) (LT + LS)CS
+CyLTCs-rL,(s,
and
LT-Ly ={(C- Cy)RPRs[l - <S{LT+Ly) Cs]
-RpRSJ(('xLTCS-CLyCsy)

Equation (18) is obtained from the following two

equations which are derived from eq (14) modified
according to eq (16)
LT=CRpRs (l-oi'LTCs+R^^0+K)

-(ffi+rJra+^Tp {rP+rs)-lL
and

and with negligible error the substitution CyRFR;

= Ly and CRPRs=Lr can be made in the snial
correction terms. Tlie second term in braces can be
reduced to u2LTLy{CsCsy). If the terms unde:
consideration are not relatively small, the approvjmations are not valid.
Neglecting second order terms, the above equation
can be simplified, giving eq (18):
LT=Ly+ (C-Cy)RPRs[l-o\LT+Ly)Cs-rK]

Ly=CsRPRs (i-^LsCsy+^f^ (1+JO

- <JLTLN{CS- CSy) + (R2t-R%) Cs-iU-!^

(RL.\+rL.v)TK+^ "f/s (tp+ts)Ilx-

In eq (15) it will be noted that at the higher fre

quencies the first term in K is the largest term, and
that at lower frequencies all terms in K are small
and often negligible. The first term in K is in
dependent of Dc, which may vary slightly between

11. References
[1] J. C. Maxwell, A treatise on electricity and magnetiss
1st ed., 2, 377-379 (1873), or 3d ed.,"Art. 778.
[2] Max Wien, Ann. der Phvs., 44, 689-712 (1891).
[3] H. L. Curtis, Electrical Measurements, 113-1 17, McGrsr
Hill Book Co., Inc., New York, N.Y. (1937).

264-188

(Paper 65C3-S

JOURNAL OF RESEARCH of the National Bureau of Standards C. Engineering and Instrumentation

Vol. 70C, No. 4, October-December 1 966
Some Techniques for Measuring Small Mutual Inductances
D. N. Homan*
Institute for Basic Standards, National bureau of Standards, Washington, D.C. 20234
(July 6, 1966)
A method of measuring small mutual inductances is presented. The smallest inductor measured
was 0.1 fiH. The circuit is a transformer-ratio-arm bridge with multiple balances and is described in
detail. Uncertainties are of the order of one part in 107.
Several ideas for the design and construction of suitable mutual inductance standards are presented.
Key Words: Auxiliary generator, bridge, coaxial choke, mutual inductor, stray magnetic field,
transformer.
1. Introduction
A recent theoretical study by Page [1] 1 suggests
the desirability of making accurate measurements of
mutual inductance at the 0.1 /xH level and of stepping
up to larger values. An experimental bridge has been
constructed for the purpose of investigating some tech
niques which seemed appropriate in this range.
Some of those techniques which appear to be particu
larly useful are described in this paper. The bridge
is a transformer-ratio-arm bridge, having a ratio of
10 to 1, and has been designed to compare mutual
inductors through five orders of magnitude, from 0.1
/u.H to 1 mH at a frequency of 1592 Hz. Sensitivity
of a part in 107 is achieved at the 0.1 /xH to 1 /xH level
provided the inductors carry 3A in their primary
windings.
2. Bridge Circuit
A simplified version of the bridge is shown in figure 1.
Mutual impedances Z\ and Z2, which are both mainly
inductive, are defined by the equations
Z\ =j(l>M1 + Ry

method described by Cutkosky and Shields [2]. The

bridge balance is accomplished by the adjustment of
e3, which is an adjustable voltage source, small relative
to ei and t%.
The detector consists of a tunable amplifier and a
tuned high-impedance preamplifier, coupled to the
bridge through an impedance matching transformer.
The transformer is tuned with a variable capacitor
to give maximum output voltage.
Accurate measurement of small inductances re
quires that all sources of stray magnetic field be well
controlled. To that end, inductors and transformers
are toroidal, bridge components are enclosed in copper
eddy-current shields, components are interconnected
by coaxial cables, and shield current of the coaxial
leads is restricted to be nearly equal and opposite
to current in their inner conductors. Figure 2 illus
trates shielding and the use of the technique of thread
ing a coaxial lead through a high permeability core
[3] forming a coaxial choke to equalize current in
the inner conductor and outer shield of the coaxial
cable.
Following construction of the bridge, tests were
made to establish the effectiveness of the design in
keeping stray magnetic fields small. Such tests were
made by using two coils of wire, one connected to a

Z2=joiM2 + R>.
If both inductors have the same current, t, in their
primary windings,2 the comparison of Zx to Zi is
achieved by measuring the ratio of the induced voltages
d and e2.

Since Z,=^r and Z2 = ~r, then -rr =

l
l
Z-i f?2
The ratio of ei/e2 is compared with the 10 to 1 voltage
ratio of an inductive voltage divider, which may be
calibrated by a modification of the capacitance-ratio
*('rt--i-fil dililr*"...: Hiuli rreirorni'} lm|M-ddnrf Standard* Srrlixn. NHS. H..uldrr.
(...I.. HJI.U|_'.
1 Figures in bracket* indicate the literature reference* at the end of this paper.
' Ground capacitance current between the primary windings is negligible; it is discussed
in *4-<-!i,m \ "t tin* p;i|-r.
265-221

T
FIGURE 1. Simplified inductance bridge circuit.

I" Figure 3. Active coaxial choke.

FIGURE 2. Inductance showing shielding, junction points (A, B,

C, D, E, F) of the mutual inductor, and use of high-permeability
core to equalize coaxial current.

detector and used to find sources of stray field, and

the other connected to a generator and used to find
places in the circuit that were sensitive to stray field.
For example, the detector coil was moved around
until a source of stray field was found, and the amount
of detector off-balance was noted. The coil connected
to the generator was driven so that its field gave about
the same off-balance at a comparable distance from
the detector coil. Then the driven coil was used as a
probe to determine the effect of such a field on the
main bridge balance when brought near various
bridge components.
The largest sources of stray field were the coaxial
leads in the loop that includes the inductor primary
windings since the current is largest in that part of
the circuit. It was determined that the coaxial choke,
mentioned above, was not sufficient to equalize cur
rent in the inner conductor and outer shield of the
coaxial cable. A greatly improved coaxial choke
was realized by making it an active device using two
cores in a manner described by Cibbings [4]. As it
is used in this application, it may be called an "active
coaxial choke." The principle is shown in figure 3.
If ep = Zpi, eq = Gep, and er = ep + eQ, it can be shown
that Zt = Zp(1+G). Thus, the active choke reduces
net current much more than a coaxial choke com
posed of a single magnetic core. The gain of the
amplifier and the number of turns of cable through
the device was adjusted to reduce net current by an
amount which made the error negligible.
Small stray magnetic fields due to slight nonsymmetries in the coaxial cables remained. Using the
probe technique discussed above, three modifications
were made. First, the copper can serving as an eddycurrent shield for the inductive divider was found to
be unsatisfactory because the lid was electrically
attached to the can at only one point. The shield was
improved by bolting the lid to the can firmly with
screws spaced about 2 in apart around its circum-

ference. Second, an uncertainty in the value of mutual

nductance arises from the coupling between the leads
connected to the inductor primary windings and the
leads connected to the inductor secondary windings,
if the leads move relative to each other. Subsequent
inductors were designed to have the primary winding
terminals on the opposite side of the shield can from
the secondary terminals. Third, the smallest induc
tors (for which the ratio of mutual inductance of the
leads to mutual inductance of the inductor is largest
were modified as follows: The eddy-current shield
was extended over the secondary leads by passing
these leads through a three-inch-long copper cylinder
having Vs-in wall thickness (about two skin depths at
the frequency used).
The standards to be compared, as illustrated it
figure 2, are defined as follows: Let i'A be the current
into the inner terminal of connector A, i"t be the cur
rent out of the outer terminal, and eA be the voltage
from the inner to the outer terminal. Currents and
voltages are similarly defined at connectors B, C, D.
E, and F. Then the mutual impedances of interest art
defined by ZAb = with the conditions i = ii'. f>
= 0, and iA = iA = 0; Zot' ~ with the conditions
h:
h: = h:< et = 0, and ii,= ip = 0. The above conditionare similar to those used in the theoretical analysis
by Cutkosky of four-terminal-pair networks [5].

3. Auxiliary Balances
The techniques used to assure the realization a
the conditions defining the inductors involve the nsf
of several auxiliary generators and detectors. Figure
4 shows the complete system. Each of the four auxiJ
iary generators consists of a decade capacitance net
work in parallel with a decade conductance network,
driven by a 40-turn winding on the magnetic core *
the bridge supply transformer. Figure 4 shows t
such 40-turn windings, the sign of the voltage of on*
being opposite to that of the other so that each capaci
tance decade or conductance decade can be switchet

266-222

AUXILIARY GENERATORS
FIGURE 4. Complete system for measurement of small mutual inductances.

independently to reverse the sign of the voltage driving

it. Stray capacitance from the 40-turn winding to the
shield raises the potential of the shield and would
result in current in the shield. This difficulty is over
come if an adjustable capacitor, Co, is attached to the
20-turn winding, the voltage of which is opposite in
sign to that of the 40-turn winding. A capacitance
bridge is formed by placing a detector between the
shield and the bottom of the 40-turn winding, and the
bridge can be balanced by adjusting Co- A similar
bridge is formed with Cc as one component. Perma
nent adjustments of Cr. and C('; were made following
the completion of this component of the system.

inductors to the inductive voltage divider, T, and

capacitance in the inductive voltage divider itself.
Each of the auxiliary generators is adjusted until un
plugging the inner conductor of the coaxial connection
at the inductor secondary shows no change in the null
at detector No. 1. Such a condition can be expressed
as i'A = 0 and io = 0. The impedances of the auxiliary
generators are high relative to the inductor secondary
windings so that convergence to balance is more
quickly achieved. Without the auxiliary generators,
the loading in the circuit is such that the voltage eA
at connector A of a 100 /aH inductor was changed by
approximately one part in 105 when the load was dis
connected from the circuit by unplugging the lead to
The load on the secondary winding of each inductor the secondary winding. Thus the auxiliary generators
is supplied by an auxiliary generator. The loading need to be adjusted to one part in 103 to obtain accuracy
consists of capacitance in the cable connecting the of one part in 10".
267-223

At the 1 mH level, an error arises from a capacitance

change at the secondary terminal when a coaxial cable
is connected to the terminal. When such a connection
is made, the inner connector of the coaxial cable slips
over the inner member of the connector on the inductor,
thus changing the capacitance from the inner connector
of the inductor to its shield about 0.3 pF. The im
pedance of the secondary winding of the 1 mH inductor
used in this work is about 20 ft. A calculation shows
that one could expect an error of about 6 in 10* due to
such a capacitance change. This error can be made
reproducible by using the same coaxial cable each
time the 1 mH inductor is measured.
The inductive voltage divider T is a two-stage device.
Two stage devices have been analyzed by Cibbings
[4] and by Cutkosky [6]. The auxiliary generator
(gen. 4 in fig. 4) incorporated in the transformer sup
plies the excitation current in the ratio winding. With
the bridge balanced, a detector winding on the second
core assures that no excitation current is supplied
by the inductor secondary windings. Thus this induc
tive voltage divider behaves like a device having infinite
input impedance in that it draws no current. An
additional feature of the two-stage design is a highly
accurate 10 to 1 ratio [4].
The conditions eB = 0 and e = 0 are met by adjust
ment of the auxiliary generator 1 in the part of the
circuit containing the primary windings of the mutual
inductors such that detector No. 3 indicates null.
Shunt admittances in conjunction with the impedance
of the coaxial cable from the detector junction to points
B or E may cause an error. A calculation based on an
assumed equivalent circuit and approximate values
for the impedance and admittance indicates that such
an error is in the order of 1 in 108 or less for typical
length of coaxial cable used at this point. An experi
mental verification of the above calculation was made
by observing the difference in the main bridge balance
caused by changing the length of coaxial cable between
B and E.
Tests of interdependence of the detector balances
were made by bringing the bridge to balance, and then
changing one auxiliary generator voltage while observ
ing changes in the null conditions at the detector junc
tions. Such tests and experience in bringing all the
detectors to a simultaneous null indicate some depend
ence of one balance on another. However, the time
involved in obtaining convergence of all balances is
not unreasonably long.

The condition i^ = i'A' can be assured by an auxilian

device illustrated in figure 5. One can adjust e< i
null Da) to compensate for net current which can resn:
from capacitance from the primary to the secondan
winding of an inductor. However, it is possible i
construct inductors in such a way that the change i:
the main bridge balance (detector 1) when e* is swrtcl*-:
in and out is small enough so that the corresponds,
error is less than 1 part in 10". The 1 mH mutu*
inductor was altered by inserting a permanent shor
between the low-voltage side of the primary winding
to the low-voltage side of the secondary. Inductorsmaller than 1 mH were not fixed with a short as de
scribed above since the error corresponding to the
change in the main bridge balance when e is switchr^
in and out in their case was smaller than 1 in 108 wits
out the short.

FIGURE 5. Method of assuring zero net current at inductor secondaryterminal.

Figure 6. Basic bridge and, in dotted lines, circuit used to oh-'

balance.

4. Main Bridge Balance

The circuit used to obtain the small adjustable
voltage, 3 is shown in figure 6. That portion of figure
6 shown within dotted lines represents a componerii
part of an "a-c direct-reading-ratio set" which has
been described by Cutkosky [6]. This is an active
circuit with high input impedance and low outpu'
impedance. Hence Af3 is not strongly loaded, ani
reasonably large loads can be driven by the circuit
The in-phase and quadrature components of e3 are
read directly from dials of two inductive voltage divid
ers, IVD#1 and #2 in figure 6. The mutual induct :
Mi has nearly the same current in its primary winding
as the inductor Mi. The voltage e3 is related, throup
current i, to e\ and e%. The relationship is controlled
by the choice of the value of M%, which provide* ;
scale factor. The value of Ms has been calculates
such that one step of the last dial is one part in \lr

268-224

The range of e$ is 5 in 104 and the least count is 1 in

10*. The current in M3, therefore, must be no more
than 2 in 10s different from that in Mi for accuracy
of 1 in 10*. A calculation shows that ground capaci
tance between Mi and M3 and in M3 must not exceed
1000 pF for this accuracy. The inductor M3 is con
structed to have ground capacitance well below 1000
pF. Components of the apparatus shown within
dotted lines in figure 6 were initially adjusted to an
accuracy of a few parts in 10s. Past experience with
the type components used in the apparatus indicates
that at the time of these measurements the accuracy
of the dial readings was better than 1 part in 103 of
the dial reading. Thus if the inductors being com
pared (i.e., Mi and M2) are adjusted so that their values
differ by less than 1 in 105, the measurements can be
relied upon to 1 in 10*.

coaxial connectors in such a way as to create a mini

mum loop. The inductor is mounted in a copper
shield of Vs-in thickness (about two skin depths at
1592 Hz). The coaxial connector of one of the pri
mary leads is connected to the copper shield at one
point inside the shield. The primary leads and the
secondary leads are brought out on opposite sides of
the toroid as shown in figure 7 to keep magnetic coup
ling to a minimum outside the toroid.
If two windings have toroidal symmetry about a
common axis, and if one of these windings is totally
within the other, the mutual inductor thus formed has
been observed to have the following property: The
phase angle of the inductor is practically independent
of the thickness of the wire used for the outside wind
ing; the phase angle depends on the thickness of the
wire used on the inside winding, a larger phase angle
being produced by thicker wire. The outside winding
is, therefore, better suited for the primary, since the
5. Measurement Procedure
primary should have low resistance for minimum heat
A single detector is used for all balances; it is dissipation.
A few inductors have been constructed using toroidal
plugged alternately into positions D\, Di, and D3 (fig.
4). Balancing starts by bringing Di to null by the dials forms of Bakelite; they were potted in epoxy resin
IVD#1 and IVD#2 (fig. 6). Next, leaving the detec and enclosed in Vs-in thick copper shield cans. These
tor at Di, the secondaries of the mutual inductors Mi inductors are useful for testing some of the features
and Mi are unplugged in turn. With the secondary of the inductance bridge, but have temperature coeffi
Mi unplugged, a null at Di is obtained by turning the cients that are too large to make them useful as stand
dials of gen. 3 (fig. 4); similarly, the secondary of Mi ards in this work.
A 0.1 fiH inductor was made by painting silver
is unplugged, and the balance reached with gen. 2.
The detector is then plugged into the Di position and windings onto a toroidal form of fused silica. The
a null is achieved by turning dials of gen. 4. Finally, silica form, with the silver secondary winding, was
the detector is plugged into the D3 position and brought baked and then a primary of copper wire was wound
around the form but mechanically isolated from it.
to null by gen. 1.
All balances described above are then repeated A critical problem is that of bringing the leads from
the secondary out through the primary (see fig. 8).
several times until convergence is reached.
A cylindrical fused-silica rod was fused to the toroidal
form as in figure 9 so that the leads could be brought
6. Inductors
out on a solid mount. An attempt to use silver painted
leads was made but abandoned in favor of using #36
Some of the requirements of the inductors are copper wire. These lead wires were tightly twisted
that they be stable, astatic, have small phase angle,
be close to nominal value, and be capable of carrying
substantial current in their primary windings.
To make the inductors astatic, both the primary and
secondary have toroidal symmetry about the same
axis; the primary winding is outside the secondary
SECONDARY
winding and, therefore, has a larger cross section.
LEADS
The leads, as shown in figure 7, are brought out to the

[ft
- J V.

FIGURE 7. Lead geometry and shielding of mutual inductor.

FIGURE 8. Winding geometry of mutual inductors having fused

silica toroidal form.

269-225

Figure 9. Fused silica toroidal form and lead mount.

and attached to the fused-silica rod with alkyd resin

paint, and electrical contact to the secondary was
made with silver paint. The gage of the wire was
chosen to be as small as possible to keep the loop
area small, but large enough such that lead resistance
was below the resistance of the painted secondary.
A possible source of uncertainty due to relative
motion between primary and secondary is the magnetic
coupling from the "single turn" produced by the pro
gression of the turns about the toroid. This is an
effective loop around the toroid in which an emf can
be induced by a component of magnetic field parallel
to the axis of the toroid. To reduce this uncertainty
all the skew of the turns was concentrated at the inner
surface of the toroid as shown in (fig. 10). After a thin
insulating coating of alkyd resin paint was applied,
a return loop was painted on top of the skewed wires
forming a loop in the opposite direction to the loop
formed by the skew.
A nonuniformity .of the secondary winding painted
on the toroidal form could create a net loop which
would cause the inductors in effect to be lopsided.
Thus relative motion between the primary and sec
ondary windings could change the inductance. A
calculation for the 0.1 /xH inductor indicates that such
a net loop of 1 mm2 would dictate that for stability
of one part in 108, the relative motion of primary to
secondary must be 10~5 mm.
It appears that improved techniques in painting
the secondary winding onto a fused-silica toroidal
form would result in inductors of good stability.

(Paper 70C4-233)

FIGURE 10. Illustration of painted secondary winding shoui'j

skew of turns concentrated in inner toroidal surface.
7. Conclusion
At the smallest level of inductance measured here,
bridge sensitivity is one part in 107. Sensitivity im
proves at larger values of inductance such that 100 jiH
can be compared with 1 mH with sensitivity of a fn
parts in 108. It appears likely that the bridge tech
niques described here can be relied on to yield errorthat are less than one part in 107. An analysis of ail
errors in such a step-up from 0.1 fM to 1 mH m\i~\
await the development of a more reliable set of mutua.
inductance standards.

The author thanks R. D. Cutkosky for suggesting

the scheme of measurement used here.
8. References
|1| C. H. Page. A new type of computable inductor. J. Re*. NBS67B
(Math, and Math. Phys.) 1. (Jan. -Mar. 1963).
[2| R. D. Cutkosky and J. 0- Shields, The precision measuremc-:
of transformer ratios, IRE Trans. Inst. 1-9,2(1960).
[3] A. M. Thompson, The precise measurement of small capac*
tances, IRE Trans. Inst. 1-7, 3 and 4 (1958).
|4] D. L. H. Gibbings, A Circuit for Reducing the Exciting Currrr;
of Inductive Devices, The Proceedings of the Institution li
Electrical Engineers, London, England, 108, Part B, 390961
[5] R. D. Cutkosky, Four-terminal-pair networks as precision adm;:
tance and impedance standards, IEEE Trans. Communication*
and Electronics, pp. 19-22 (Jan. 1964).
[6] R. D. Cutkosky, Active and passive direct reading ratio sett fa
the comparison of audio frequency admittances, J. Res. NB?
68C (Engr. and Instr.) (Oct. Dec. 1964).

270-226

IRE Transactions on Instrumentation

Volume 1-7, Numbers 3 & 4, December 1958

New Apparatus at the National Bureau of Standards

for Absolute Capacitance Measurement
M. C. McGregor, J. F. Hersh, R. D. Cutkosky,
F. K. Harris, and F. R. Kotter
This paper describes a transformer bridge constructed at the
National Bureau of Standards for measuring the direct capacitance
of 3- terminal capacitors ranging in values up to 1 nj and having
a least count of 10" pF. The transformers and network components
were designed specifically for operation at 1 kHz, however, with
relatively minor modifications, satisfactory operation should be
possible over the audio- frequency range to at least 10 kHz. The
construction of a cylindrical cross capacitor as a computable
standard is also described.

F. K. Harris
August 28, 1967

JOURNAL OF RESEARCH of the National Bureau of StandardsC. Engineering and Instrumentation

Vol. 68C, No. 4, October-December 1964
Active and Passive Direct-Reading Ratio Sets for
the Comparison of Audio-Frequency Admittances
Robert D. Cutkosky
(July 17, 1964)
Design considerations and constructional details for two audio-frequency direct-reading
ratio sets are presented. These devices allow the comparison of admittances with accuracies
up to one part in 10'. The first direct-reading ratio set (DRRS) is based upon operational
amplifier circuitry, and utilizes a two-stage amplifier system. With this technique, only a
moderate voltage amplification per stage is required, and the circuitry is therefore not
difficult to stabilize to prevent oscillation. The second DRRS is purely passive and makes
use of two-stage voltage transformers to reduce the detrimental effects of transformer ex
citation current upon the linearity and ratio stability of the device. Equivalent circuits
representing the behavior of critical parts of the two direct-reading ratio sets are presented
and analyzed. A convenient procedure for calibrating a DRRS is treated mathematically.

271-1

IEEE Transactions on Communication and Electronics

Volume CE-70, January 1964
Four-Terminal-Pair Networks as Precision
Admittance and Impedance Standards
R. D. Cutkosky
Practical accuracy limitations of four- terminal impedance
and three- terminal admittance measurements are discussed. It
is shown that there exists a middle impedance range in which
neither technique results in good measurement precision. A
theoretical analysis of measurement systems using the two
techniques simultaneously is made.
NBS Technical Note 57
May 1960
VARIABLE CAPACITOR CALIBRATION WITH AN
INDUCTIVE VOLTAGE DIVIDER BRIDGE
Thomas L. Zapf

ABSTRACT
The use of an inductive voltage divider bridge for the calibra
tion of three -terminal and two -terminal variable air capacitors is
discussed.

JOURNAL OF RESEARCH of the National Bureau of StandardsB. Mathematics and Mathematical Physics
Vol. 67B, No. 1, January-March 1963

A New Type of

Computable

Inductor

Chester H. Page
(November 21, 1962)
The mutual inductance analog of the generalized Thompson-Lampard theorem (for
cross capacitances) is developed. An infinitely long cage of five parallel wires can yield an
absolute inductance of
10-' In 3 + v5
henries per meter. End-effects of order 1//J occur in a finite cage, but can be reduced to
order l/l* by using eight wires.
The eight-wire cage has the advantage of overdetermined relations among the induct
ances to be measured, allowing an estimate of experimental error in the calibration of a
standard. Errors due to faulty cage geometry are shown to be of the order of 1 in 10".
272-2

Instruments

Transfer

and

AC-DC

Standards

Papers
Ac-dc transfer instruments for current and voltage measurements, F. L.
Hermach
Thermal converters for audio-frequency voltage measurements of high
accuracy, F. L. Hermach and E. S. Williams
A system for accurate direct and alternating voltage measurements, F. L.
Hermach, J. E. Griffin, and E. S. Williams
A comparator for thermal ac-dc transfer standards, R. S. Turgel
Calibration of peak a-c to d-c comparators, D. Flach and L. A. Marzetta
A differential thermocouple voltmeter, J. E. Griffin and F. L. Hermach..

275
281
290
300
310
318

Abstracts
Practical aspects of the use of ac-dc transfer instruments, E. S. Williams.
Calibration of volt-ampere converters, E. S. Williams
Standard electrodynamic wattmeter and ac-dc transfer instrument, J. H.
Park and A. B. Lewis
Precise comparison method of testing alternating-current watthour
meters, A. W. Spinks and T. L. Zapf
Voltage ratio detector for millivolt signals, J. R. Houghton
Notes on the care and use of electrical instruments, F. D. Weaver

324
324
325
325
326
326

273

AC-DC

Transfer

Instruments

Voltage

for

Current

and

Measurements*

FRANCIS L. HERMACHf

Introduction
energy (which is governed by the heating produced by
THE basic electrical units are maintained by groups a current and by the electromagnetic force produced by
of standard cells and resistors. With these stand currents acting on each other) is a function of the square
ards and a dc potentiometer, measurements of of this value.
The three general types of ac-dc transfer instruments
direct voltage and current are readily made with an ac
curacy of 0.01 per cent or better. Corresponding meas which are at present suitable for rms measurements of
urements of alternating voltage and current at power better than 0.1 per cent accuracy are dependent upon
and audio frequencies depend, at the present time, upon these same laws of the interchange of energy. They are:
ac-dc transfer instruments which have nearly equal ac 1) electrodynamic instruments, which depend upon the
and dc response. Such instruments may be calibrated force between current-carrying conductors; 2) electro
on direct current and then used on alternating current, static instruments, which depend upon the force be
or alternatively, may be used directly to make ac-dc tween charged conductors; and 3) electrothermic instru
difference measurements, as in the certification of other ments, which depend upon some effect produced by the
heating of a current-carrying conductor. The first and
instruments.
In addition to the instantaneous values as deter third respond essentially to current and the second to
mined by an oscilloscope or point-by-point methods, voltage, but series or shunt resistors make all three types
there are at least seven other quantities that may be suitable for both measurements, while other circuit ar
determined for a periodic ac wave in which all values rangements make power measurements almost equally
separated by the periodic time, T, are equal. These are: feasible.
1) the rms or effective value, I, defined by
Response Equations

where i is the instantaneous value; 2) and 3) the positive

and negative average values, i.e., the average of all the
positive or all of the negative values during one cycle,
defined by
Im+ = lrf'{ ' *' 1 + i)di and /o- = irf*{ 1 i 1 ~ *
where |t| is the magnitude of ; 4) the rectified fullwave average, defined by

The instantaneous torque of an electrodynamic in

strument in which two sets of conductors, one fixed and
one rotatable about an axis, carry the same current,
, is1

where M is the mutual inductance and 0 some welldefined angle between the two sets of conductors.
In its simplest form such as instrument has an oppos
ing torque, UO, so that, with direct current applied, the
rest position, 6/, is defined by
US, = BxP where Bt =

5) and 6) the positive and negative crest or peak values;

and 7) the crest-to-crest or peak-to-peak value.
Instruments for measuring each of these values of an
ac voltage or current wave are commercially available
(with widely varying accuracies). Ordinarily, however,
only one of these values is really required in a measure
ment. Unfortunately, it is difficult to deduce accurately
one value from the measurement of another, except for
a few waveforms (which, of course, must be known). In
most cases the rms value is the one really desired, since
the rate of transformation of electrical to other forms of
* Manuscript received by the PGI, August 14, 1958.
t National Bureau of Standards, Washington, D. C.

dlT\
de J_i,

The differential equation governing the angular deflec

tion is1
PB" + AO' + U6 = Btl
where P and A are the inertia! and damping constants,
respectively. If the mechanical inertia is sufficiently
great that with alternating current applied the periodic
fluctuations of 0 are negligible compared with the aver
age value, termwise integration over an integral number
of cycles gives

1 F. K. Harris, "Electrical Measurements," John Wiley and Sons,

Inc., New York, N. Y., pp. 47 and 413; 1952.
275-235

B
U6f

i'dl = BJ\

Thus if U and B each have the same value on direct as

on alternating current, the instrument has the same
response, 8/.
The instantaneous torque of the usual two-element
electrostatic instrument with an applied voltage, d, is
_ v> dC
2 dd
where C is the capacitance between the elements. Sim
ilar considerations lead again to identical expressions
for the rest position with direct or alternating voltage
applied, viz.,
U6f = B,V2 where B2 =

1 dCl
2 <W_U-*/

With reasonable restrictions, the temperature rise, 8,

of a thin homogeneous conductor carrying a current t
in the x direction is governed by the differential equa
tion
d*6

H8 i*R
1 d6
K + K ~ D dt

where D is the thermal diffusivity and K, K, and H are

the electrical resistance, thermal conductance, and rate
of surface heat loss per degree rise of a unit length of
the material. Again, if the "thermal inertia" is suf
ficiently great that periodic fluctuations in temperature
are negligible, the average temperatures are equal with
equal direct and rms alternating currents.
In each instrument the equality holds even if the
"constants" such as U, B, R, D, H, and K in the equa
tions are dependent upon the response, 8. This impor
tant advantage stems from the property that each in
strument combines in a single measuring element a
function proportional to the square of the instantaneous
current or voltage, a restraining function, and an inertial function or "flywheel effect" to enable it to integrate
so that the time-average value of the response is propor
tional to the square of the rms current or voltage. For
this reason high accuracy may be more easily attained
with these instruments than with other squaring de
vices or circuits, which must synthesize exact and equal
square-law responses in two quadrants (or in one quad
rant with an accurate rectifier) and provide a separate
integrator. However, each of these three types of instru
ments has a lower limit of frequency below which the
average value of the response is incorrect if the con
stants are dependent upon 8.
Transfer Instruments
It is easier in an actual instrument to insure that the
constants in the equations are indeed equal on direct

and alternating current than it is to measure them sep

arately and compute the response of the instrument.
Thus with few exceptions2 absolute instruments, such
as the electrodynamic current balance,* are used solely
for dc measurements, and most ac instruments of high
accuracy are calibrated in terms of dc standards. In the
usual instruments, this calibration is preserved on a
scale, and the instrument is tested or standardized pe
riodically to guard against changes in the factors which
affect the response. However, if the instrument is cali
brated at the time of each use, long-time stability, ex
tremely low temperature coefficients, and other nor
mally desirable characteristics become of secondary im
portance, and the instrument may be designed and con
structed to have high resolution and the best possible
frequency characteristic. Indeed, in a standardizing
laboratory in which the standard is used primarily to
test other instruments, a dc calibration may not be nec
essary because most instruments which are submitted
to such laboratories can also be used on direct as well as
alternating current. The scale calibration of such instru
ments may be verified on direct current with a potenti
ometer (taking the mean of the values required to ob
tain the same deflection for the two directions of applied
current or voltage). The ac-dc difference or frequency
influence may then be determined by connecting the
instrument under test and the transfer instrument so
that they respond to the same quantity (current, volt
age, or power) and successively switching both to alter
nating-, direct-, reversed -direct, and alternating current.
In each case the measured quantity is adjusted to ob
tain the same deflection of the test instrument and the
differences in deflection of the standard instrument are
observed. From these differences and the scale factor of
the standard instrument (per cent change per division
the ac-dc difference of the test instrument may be de
termined.
This procedure has been used for over 40 years in the
certification of instruments at the National Bureau oi
Standards, because it separates several sources of errors
and can be carried out with great accuracy. Since the
ac-dc differences of a well-designed instrument are
small within its working frequency range and are rela
tively permanent, they need not ordinarily be redeter
mined, so that the periodic tests of the instrument need
be made only with direct current.
For either a direct ac measurement or an ac-dc dif
ference determination, however, the ac-dc difference oi
the standard instrument must be known to the full ac
curacy of the measurement. The determination oi this
difference has been recognized as a major problem in
H. B. Brooks, F. M. Defandorf, and F. B. Silsbee, "An absolu:electrometer for the measurement of high alternating voltages '
/. Res. NBS, vol. 20, pp. 253-316; 1938.
* R. L. Driscoll and R. D. Cutkosky, 'Measurement of current
with the NBS current balance," J. Res. NBS, vol. 60, pp. 297-305.
1958.

276-236

most national standardizing laboratories. In general,

there are three rather distinct steps in a complete de
termination. These are: 1) choice of two different types
of instruments that are inherently suitable as ac-dc
transfer instruments, and a theoretical study of all
known effects that can cause ac-dc differences in each
type; 2) construction of instruments of each type in
such a way that each of these effects by computations
and tests can be evaluated over the required ranges;
and 3) comparison of the actual transfer performance
of the two instruments to guard against unknown
sources of error. The tests of step 2) will often take three
forms. These are: a) tests of components of the instru
ment for the effects of known factors (such as magnetic
susceptibility, for example; b) intercomparisons of
instruments of the same type but of different ranges, in
which some of the effects (such as residual reactance)
may be expected to be very different; and c) repeated
intercomparisons, in which each known source of error
in one instrument is accentuated by definite amounts
(such as by the deliberate introduction of known react
ance). Normally the comparisons of step 3) can be made
over only part of the useful ranges of frequency, voltage,
etc., and the performance must also be judged by the re
sults of steps 1) and 2). Normally also one of the two in
struments will have better-known characteristics,
wider or more suitable ranges, or better-behaved ac-dc
differences than the other, and thereafter is used as the
principal transfer instrument for that laboratory, but
both are important.
The accuracy of ac measurements is often limited by
the stability of the source rather than the instrument.
For the calibration of instruments with a transfer instru
ment, stabilized sources free from short-time fluctua
tions are required, but long-time stability and good load
regulation are of less importance.0 Repeated sets of
readings make it possible to determine the ac-dc differ
ence to a precision considerably better than the fluctua
tions in the source, and the procedure described tends
to eliminate drifts in both the test and standard instru
ments. If both instruments inherently respond to the
same values (rms for example), low-order harmonics up
to several per cent ordinarily cause the test to be in error
by much less than 0.1 per cent.
Electrodynamic Instruments
The electromagnetic force between two conductors
with reasonable geometry and currents is so small that
to obtain satisfactory torque the effect is magnified by
coiling the conductors. The torque-weight ratios thus
obtained are still considerably less than those enjoyed
by dc instruments with permanent magnets, but rea-

sonably rugged and portable pointer-and-scale instru

ments of the 0.1 per cent accuracy class, based on
Weston's designs, have been available for a number of
decades. For increased resolution and accuracy, recently
developed commercial instruments, following Silsbee's
composite-coil design,* now use a taut-suspension sys
tem with the moving coil of a second instrument rigidly
fastened to it to provide an electromagnetic restoring
torque.7-9 This torque can be precisely determined by
measuring the current through the second instrument
with a resistor and dc potentiometer.
As the torque equations indicate, an electrodynamic
instrument responds to the square of the current, the
scale being marked to indicate current directly. As an
unshunted ammeter (limited to about 0.1 ampere be
cause of the springs or ligaments which must carry the
current to the moving coil), it will be useful as a transfer
instrument over the range of frequencies for which the
average torque, T=f(I), is the same as with direct cur
rent. The chief factors which affect this in an actual
instrument are: 1) capacitances between turns and be
tween coils, which alter the currents; 2) eddy currents
in neighboring metals, which alter the field that links
with the moving coil ; and 3) electrostatic torque, caused
by differences in potential between the fixed and mov
ing coils. In general, these set an upper limit of about 1
kc for such instruments. The frequency ranges of am
meters, which use shunts across the moving coils, and
voltmeters, which use series resistors to limit the cur
rent, are more sharply limited by the inductances of the
coils. The manufacturer must insure that the ratio of the
impedances of the two parts of the divided circuit of the
ammeter is equal to the ratio of resistances, to the full
accuracy of the measurement over the frequency range
of interest. Similarly he must insure that the magnitude
of the impedance of the voltmeter (defined as the ratio
of the applied voltage to the current through the coils)
is equal to its dc resistance. By connecting resistors in
series with the coils of the ammeter and capacitors
across portions of the voltmeter, the upper frequency
limit for 0.1 per cent error in commercial instruments
can be raised from 100 cps to about 1000 cps, and for
0.25 per cent error to about 2500 cps.10
Because the mutual inductance between the coils
changes with scale position (the operating torque de
pends upon this) complete compensation cannot be at
tained at all scale positions. Mutual inductance intro-

' F. B. Silsbee, "Composite coil electrodynamic instruments,"

J. Res. NBS, vol. 8, pp. 217-264; 1932.
' G. F. Shotter and H. D. Hawkes, "A precision ac/dc comparator
for power and voltage measurements," Proc. IEE, vol. 93, pp. 314324; 1946.
' J. Sorge, "A new precision instrument for ac power measure
ments," VDE Fackberichte, vol. 17, pp. 27-30; 1953.
R. F. Estoppey, "Inductronic Electrodynamometer," Confer
* A. H. M. Arnold, "Alternating-current instrument testing ence on Electronic Standards and Measurements, paper 35; August,
equipment," Proc. IEE, vol. 101, pp. 121-133; 1954.
1958.
F. L. Hermach, "Power supplies for 60-cycle tests of instru
10 J. H. Miller, "Frequency compensation of ac instruments,"
Trans. A IEE, vol. 70, pp. 217-221; 1951.
ments and meters," Proc. ISA, vol. 11, paper 56-21-3; 1956.
277 237

duces other errors as well, so that the electrodynamic

ac-dc transfer instruments designed and constructed at
the U.S. National Bureau of Standards11 and the Na
tional Physical Laboratory of South Africa12 are nor
mally operated only over a narrow range of deflections
about the position of zero mutual inductance, by the
use of a continuously adjustable external multiplier for
the first instrument and a torsion-head for the second.
In the NBS instrument (see Fig. 1), two sets of coils are
astatically arranged to eliminate the effect of uniform
external fields, and a strip suspension and light-beam
pointer with a scale 2 meters from the instrument insure
definite readings and permit a scale factor at 0.01 per
cent/mm at currents from 0.1 to 20 amperes. The instru
ment is normally used as a wattmeter or ammeter, and
a similar but older companion instrument as a voltmeter
(10 to 600 volts).1* The NPLSA instrument is quite sim
ilar, but may be connected as a voltmeter as well (0.05
to 5 amperes and 25 to 500 volts). The ac-dc difference
of each of these instruments is believed to be known to
better than 0.01 per cent at power frequencies and to
perhaps 0.1 per cent up to 3 kc (above 50 volts), verified
by careful study and by comparison with electrothermic
and electrostatic instruments.
Electrostatic Instruments
The electrostatic force between two conductors at
reasonable spacings is very low at normal line voltages.
The resultant torque can be multiplied, as in the Kelvin
electrostatic voltmeter, by interleaving the fixed and
moving conductors (plates), but unfortunately only on a
one-to-one basis, so that torque-weight ratio remains
low. Thus electrostatic instruments, as distinct from
electrodynamic instruments where the fixed coil may be
designed to have more ampere-turns than the moving
coil, seem rather delicate except at voltages above about
1 kv, and are rarely available commercially for precision
measurements. However, taut-suspension, light-beampointer constructions, using improved optical systems
and modern materials, have begun to appear commer
cially in Europe and may presage a renaissance of this
type of instrument.
Electrostatic instruments inherently respond to volt
age but may be adapted for current measurements bymeasuring the voltage across a resistor carrying the
current. However, the voltage must be rather high (50
to 100 volts) for adequate accuracy and the shunt must
then dissipate considerable power at high currents, or
an instrument transformer must be used (either a cur
rent transformer with a resistor in its secondary, or a
voltage transformer between a resistor and the instru
ment).

Fig. 1 NBS electrodynamic ac-dc transfer instrument

(external multiplier and scale not shown).
Electrostatic instruments also have a number oi
sources of ac-dc differences. Among these are: 1) the
series impedance of the suspension or springs, 2) surface
charges on insulation in the electric field, which can
modify the field with direct voltage applied, and 3) im
perfectly conducting films on the plates which can cause
the field strength between the plates to differ on direct
and alternating voltage.14 None of these limit the fre
quency range as sharply as does the inductance ot
electrodynamic instruments, and electrostatic voltme
ters may be accurate to perhaps 1 mc. However, when
an electrostatic instrument is used as an ammeter the
impedance of the 4-terminal resistor must be equal to
its dc resistance. (If a transformer is used, it is no longer
an ac-dc transfer instrument.) This is difficult to achieve
above about 20 kc.
Electrostatic voltmeters with ranges from 50 to 160
volts have been the principal ac instruments for over
40 years at the National Physical Laboratory in Eng
land14 and somewhat modified instruments of the same
basic design are similarly used at the National Stand
ards Laboratory in Australia. The ac-dc differences oi
the NPLE instruments are known to 0.01 per cent a:
power frequencies and to better than 0.05 per cent up to
100 kc (verified by calculations of the known sources oc
error and by comparison with electrodynamic and elec
trothermic instruments and with other electrostatic

11 J. H. Park and A. B. Lewis, "Standard electrodynamic watt

meter and ac-dc transfer instrument," J. Res. NBS, vol. 25, pp. 545579; 1940.
" J. W. Whittaker, "A precision electrodynamometer standard and
ac/dc transfer instrument," Proc. IEE, vol. 101, pp. 11-20; 1954.
14 R. S. J. Spilsbury and A. Felton, "The electrostatic voltmeteu F. K. Harris, "A suppressed zero electrodynamic voltmeter," as a dc/ac transfer instrument," /. IEE, vol. 89, pp. 129-1.57.
J. Res. NBS, vol. 3, pp. 445-457; 1929.
1942.
278-238

instruments having noble-metal conductors). Special

transformers and resistors have been constructed and
tested to permit the measurement of current and of
voltage below 20 volts, and resistance voltage dividers
have been evaluated for voltage measurements to 1000
volts (at audio frequencies).*
Electrothermic Instruments
Easily measurable effects are produced by the heating
of short straight metallic conductors or very small beads
of semiconductors, having very low electrical timeconstants. These effects can thus be independent of fre
quency up to 100 mc or more. In addition, the effects
can be measured electrically, rather than mechanically,
with high precision and freedom from mechanical
resonances. These are the chief advantages of electrothermic instruments.
Because their output is electrical, and because of their
low reactance, thermocouple instruments and bolometer
bridges are the only two forms now generally used. In
commercial thermocouple instruments the temperature
rise of a conductor (heater) is measured with a thermo
couple and millivoltmeter. The combination of the
heater and thermocouple is called a thermal converter.
Many commercial bolometer bridges now contain bead
thermistors (temperature-sensitive resistors) and are
unbalanced by the temperature rise caused by the heat
ing of the applied current. The unbalance voltage is
measured or balance is restored by additional currents
of a different frequency, which are measured. Thermo
couple instruments of the J to 2 per cent class have long
been commercially available for current and voltage
measurements, while most commercial bolometers are
used for power measurements (absorbing all of the
measured power). In either case undesired thermal ef
fects, such as high ambient temperature coefficients,
accelerated aging of the heater at the elevated tempera
ture, and transfer of heat to other parts of the measur
ing element have limited the sustained accuracy obtain
able. However, by using such instruments solely as ac-dc
transfer instruments, calibrating them before and after
each reading, these effects can largely be eliminated.
Studies at the National Bureau of Standards" have
shown that properly designed thermal converters may
be used as ac-dc transfer instruments at currents from 1
ma to 50 amperes with an accuracy approaching 0.01
per cent at power and audio frequencies and very prob
ably up to 200 kc. Their low reactance also makes possi
ble transfer voltmeters from 0.2 to 600 volts (at 133
ohms per volt) with wire-wound series resistors that
have ac-dc differences less than 0.03 per cent at audio
frequencies. Voltmeter elements with film resistors in
coaxial lines terminated by thermal converters have
been constructed and intercompared to 0.05 per cent to
" F L. Hermach, "Thermal converters as ac-dc transfer stand
ards for current and voltage measurements at audio frequencies,"
J. Res. NBS, vol. 48, pp. 121-138; February, 1952.
279
31S-4U O - 68 - 19
** ' '

10 mc." The calculated ac-dc differences of the convert

ers and of the voltmeter elements were verified by intercomparisons of different ranges and by comparisons
(over limited ranges) with electrodynamic and electro
static instruments. A Lindeck deflection potentiometer
is used at NBS with these converters and voltmeter ele
ments to provide high resolution (0.005 per cent/mm at
rated output emf) and rapid reading.
Similarly it has been shown that a properly selected
and mounted pair of indirectly heated thermistors may
be used in a bridge as an ac-dc transfer instrument for
an audio-frequency potentiometer to better than 0.02
per cent at frequencies from 0.2 to 20,000 cps." Ther
mistors are also used for ac measurements at the Physikalisch-Technische Bundesanstalt in Germany.1* Their
rather long thermal time constants (about 10 seconds)
provide good low-frequency response but make highly
stable sources necessary.
The small ac-dc differences of thermal converters and
thermistor bridges at audio frequencies are caused al
most solely by: 1) thermoelectric effects in the heater,
which affect the temperature rise on direct current, but
because of thermal inertia do not affect it on alternating
current; 2) integration errors at low frequencies, which
are inversely proportional to the square of the fre
quency; and 3) small reactance effects in the voltmeter
elements, such as capacitance currents between parts of
the multiplier, and from the multiplier to its surround
ing shield. Fortunately, it is possible to obtain thermal
converters with heaters of manganin or other alloys of
low thermoelectric effects and of sufficient length to
make the ac-dc differences 0.01 per cent or less from 20
cps to 200 kc.
A single 10-ma thermal converter is used with multirange shunts and series resistors for ac measurements
in the U.S.S.R., both in a central standardizing lab
oratory and in convenient consoles at branch labora
tories." Series and shunt resistors are also used in the
laboratories of the Germany Authority for Weights and
Measures, but with two converters in a differential
circuit." As a further modification of the transfer princi
ple, the heater of the thermal converter alone (without
the shunt or series resistors) can be switched to direct
current for the dc calibration." Only a small dc source is
" F. L. Hermach, "Electrothermic instruments for the measure
ment of alternating current and voltage," Proc. NPL Symposium on
Precision Electrical Measurements, HMS Stationary Office, London,
Eng., paper No. IS; 1955.
" F. C. Widdis, "The indirectly heated thermistor as a precise
ac-dc transfer device," Proc. IEE, vol. 103, pt. B, pp. 693-703; 1956.
" H. J. Schrader, "A-C potentiometer measurements with in
directly heated NTC resistors," EUctrotech. Z., vol. 73-A, pp. 547549; 1952.
11 K. P. Shirokov, "An installation for calibrating ammeters and
voltmeters at high frequencies," Proc. Russ. Inst. Meteorology, vol.
24, p. 24, 1954.
" W. Rump, "On the exact absolute measurement of ac voltage)
and a potentiometer for testing ac instruments," Elektrotechnik, vol
5, pp. 64-67; 1951.
11 F. L. Hermach and E. S. Williams, "Multirange audio-fr*
quency thermocouple instruments of high accuracy," J. Res. NB.
vol. 52, pp. 227-234; 1954.
-239

required, making possible a convenient portable acces

sory to a dc potentiometer, which, in a recently im
proved model (see Fig. 2) provides ranges of 7.5 ma to
20 amperes and 0.5 to 600 volts with an accuracy of
0.05 per cent at frequencies from 5 to 50,000 cps.
Other Instruments
A transfer instrument is an essential part of any ac
curate ac potentiometer. Such potentiometers have been
used to some extent in Europe but rarely in the United
States, except for special measurements. They require
phase as well as magnitude balance, and with the usual
tuned detector, measure only the fundamental com
ponent rather than the rms value. An electrodynamic
milliammeter is most commonly used with a standard
cell to standardize the current through the resistance
network, but a thermal converter or thermistor bridge
could serve equally well and over a wider range of fre
quencies.15,17 Recent studies25 have shown that some
converters have excellent long-time stability if they are
maintained at a constant ambient temperature.
A method of determining the equality of a dc voltage
and the crest-to-crest value of an ac voltage to better
than 0.005 per cent has been developed at the National
Standards Laboratory of Australia21 as well as a method
of generating an ac voltage wave with no harmonic
greater than 0.001 per cent of the fundamental. The
equipment has been used to verify the ac-dc perform
ance of the NSL electrostatic voltmeter to better than

Fig. 2Volt-ampere converter for ac current

and voltage measurements.

0.01 per cent at 50 cps and may be considered to be a

crest-value ac-dc transfer instrument.
High-gain amplifiers with a rectifier in a feedback
loop, and driven rectifiers based on the ring modulator,
may be used with a potentiometer to measure the aver
age value of an ac voltage wave to 0.1 per cent or better,
provided the galvanometer or other detector properly
integrates the rectified output. Their accuracy must also
** J. J. Hill, "A precision thermo-electric wattmeter for power and be verified by the procedure outlined for ac-dc transfer
audio frequencies, Proc. IEE, vol. 105, pt. B, pp. 61-68; January,
instruments. At present this is most easily done by com
1958.
W. E. Smith and VV. K. Clothier, "Determination of the dc/ac parison with rms instruments in circuits with voltages
transfer error of an electrostatic voltmeter," Proc. IEE, vol. 101,
of known waveforms.
pp. 465-469; 1954.
Reprinted from IRE TRANSACTIONS
ON INSTRUMENTATION
Volume 1-7, Numbers 3 & 4, December, 1958

280-240

Thermal

Converters

for

Measurements

Audio-Frequency

of

High

Voltage

Accuracy

F. L. HERMACH, fellow, ieee, and E. S. WILLIAMS

AbstractThe ac-dc differences of a reference group of thermo

elements have been evaluated at audio frequencies to a few parts per
million (ppm) at currents from 5 to 20 mA. A technique for comparing
the ac-dc differences of two thermoelements with an uncertainty of
about 2 ppm has been developed. Two 5 mA thermoelements are
used with a plug-in set of resistors of computable reactances to form
thermal voltage converters for voltage measurements. With this same
technique adjacent ranges of these converters can be compared to
step up from 0.5 to 500 V to better than 10 ppm.
I. Introduction
IN PRINCIPLE, the current balances which have
been developed to determine the ampere in absolute
measure with direct current could also be used with
alternating current. However, the most stable electrical
standards are still the dc standard cell, and 1 ohm re
sistors, which are not usually designed for ac use. Thus,
ac measurements of current and voltage have, in prac
tice, been based on rms ac-dc transfer standards or
comparators. For many decades these were electrodynamic or electrostatic instruments, designed to have the
same torque constant with either alternating or direct
current (or voltage) applied, making it easier to obtain
a wider frequency range than with instruments like the
current balance which must have a computable torque
constant. More recently, electrothermic instruments
such as thermocouple transfer standards and bolometer
or thermistor bridges have been used. All of these instru
ments have been highly developed in the national labo
ratories of several countries to make ac measurements
at audio frequencies to about 100 ppm in combination
with dc standards. The combination possesses a number
of advantages over ac potentiometers (which must also
include a transfer instrument), and over electronic and
other squaring circuits (which must have exact squarelaw characteristics in two quadrants).
Recent advances in operational amplifiers and induc
tive voltage dividers have made possible wide-range ac
power sources with linearity and day-to-day stability
better than 100 ppm. This implies a need for better
accuracy in basic ac-dc transfer standards.
At this accuracy, conventional forms of electrodynamic transfer instruments are limited to power fre
quencies by the effects of inductive and capacitive re-

actances in their coils, and electrostatic instruments

have a serious minimum-voltage limitation because of
their low torque-weight ratio. Both would require new
means of sensing deflections to obtain precisions of 10
ppm or better. This could be done by measuring the
reactance, which is a function of the deflection of the
instrument. However, a major difficulty would remain
that of vibrations from external forces or from the
cyclic variations of the ac driving torque. Experience
has shown that it is often difficult to control spurious
torques caused by minor mechanical resonances. It is
possible that floating the moving system in a liquid of
the same density would minimize these difficulties.
However, it has seemed to the authors that improved
thermal converters (thermocouple instruments) offer
better promise of obtaining an ac-dc transfer accuracy
of 10 ppm or better, and considerable effort has been
concentrated on them over the past five years. This
paper may be considered as a progress report describing
the present group of reference thermoelements and co
axial thermal voltage converters, and the techniques
and equipment which have been developed to compare
them to a few ppm.
II. Thermal Converters
A thermoelement (TE), defined as the simplest type
of thermal converter [l], consists of a heater and ther
mocouple. In its usual form, the heater is a short,
straight wire suspended between two supporting leadin wires in an evacuated glass bulb. The hot junction
of a thermocouple is fastened to the midpoint of the
heater, and is electrically insulated from it with a small
bead. The thermocouple EMF (about 5 to 10 mV for
a conventional TE at rated current)1 is then a measure
of the heater current. For voltage measurements, a re
sistor is connected in series with the heater to form a
thermal voltage converter [l].
At the National Bureau of Standards, fourteen TEs
form the present reference group for ac-dc difference.
Twelve of them (made by four different manufacturers)
are of conventional design but with heaters of Evanohm
or Karma (modified nickel-chromium alloys), which
have low thermoelectric effects.1 They were chosen from

Manuscript received June 23, 1966. This paper was presented at

1 The international system (SI ) of units with its symbols is used in
the 1966 Conference on Precision Electromagnetic Measurements, this paper. Because of their frequent occurrence, the word thermoele
Boulder, Colo. This work was supported in part by the Army Metrol ment is abbreviated as TE and thermal voltage converter as TVC in
ogy and Calibration Center, Frankford Arsenal, and the Metrology this paper.
Engineering Center, Bureau of Naval Weapons, Pomona, Calif.
* Evanohm and Karmn are registered trademarks of the Wilbur
The authors are with the Electrical Instruments Section, Elec B. Driver Company and the Driver-Harris Company, respectively.
tricity Division, National Bureau of Standards, Gaithersburg, Md.
(See Acknowledgments.)
281-260

Pertinent data on the TVCs, and the ac-dc differences

larger groups on the basis of low dc reversal differences
and high bead resistances, and are all mounted in one assigned to each range are given in Table II. The TVCs
aluminum enclosure lined with foamed plastic for are shown in Fig. 2.
thermal insulation. Each is supported by its heater
IV. EMF Comparator
leads, which are soldered directly to binding posts
The ac-dc differences of a transfer instrument may be
mounted on the low-loss phenolic top plate of the en
closure, and its thermocouple output leads, which are defined as
soldered to two-pin connectors. The other two thermal
S = (X. - Xd)/Xd
current converters are of a radically different design
having many thermocouples (40 or more) attached to a where Xa and X* are the magnitudes of the ac and dc
bifilar (doubled-back) heater [2]. They were obtained quantities that are required to give the same response
from their inventor, F. J. Wilkins of the National Physi (such as deflection or output EMF) of the instrument.
(Normally, the average for the two directions of the dc
cal Laboratory in England.
The fourteen TEs are identified in Table I, along quantity gives the best measure of the dc reference.;
with certain pertinent characteristics, as well as the ac- This is a useful definition, for the ac quantity for a
dc differences assigned to them at audio frequencies as given response is then simply X. = Xd(i + 6). As is well
known, the instrument may be used as an ac-dc transfer
a result of the tests to be described.
standard simply by observing the response with the ac
TABLE I
quantity applied, then measuring (with external stan
Reference Thermoelements
dards) the average for the two dc quantities required to
obtain the same response, thus avoiding many of the
Rated
Dc Approx. Bead Ad-dc differlimitations of an ordinary instrument.
Identi
heater resis
re
ence I ppmj
At the National Bureau of Standards, TEs are evalu
fication rent 1 EMF versal! length tanceJ
mm
ppm
Gn 20 Hz 2 kHz|| ated in pairs with the EMF comparator shown in Fig. 3,
mA
mV
to determine the differences in their ac-dc transfer
NPL 14* 20
25
< 10
>1
0
0
performance, i.e., 5i Sj as defined previously. TVCs are
25
<10
NPL 15' 20
>1
0
0
A20E3
20
5
40
5
>1
0
0
connected in parallel and evaluated in the same way.
A20E6
20
5
60
5
>1
0
-1
With the desired current or voltage applied, Ri and R>
130
A10E18 10
8
5
0.3 +6
+1
A10E19 10
8
160
5
0.5
0
-1
are
adjusted in a preliminary balance to obtain a near
W10E1
10
7
200
8
0.6 + 1
+3
null on the detector D\. Then, in regular succession,
WI0F.2
10
7
<10
8
0.8 _2
0
B10K4S 10
160
8
4
>1
-8
-2
alternating-, direct-, reversed direct-, and alternatingA5E8
7
10
5
0.7 + 1
5
+1
currents are applied to the TEs at nearly equal time
A5K9
6
5
40
0.6
0
5
-1
_}
10
K5K9
4
5
0.5 -3
S
intervals and without changing 7?i or R%. In each case,
K5K13
4
260
5
0.5 -2
5
0
the current is adjusted to obtain the same output EMF,
B5K77
110
5
8
6
>1
+6
Ei of TEi, as indicated by an auxiliary Lindeck poten
* These can be used at much higher currents.
tiometer (P with its detector Dt), and the resulting de
t Al/I for the same EMF.
1 At 2 kHz and rated /.
flections of Di are observed. Then, as shown in Appendii
|| Same value assigned at 20 kHz,
I, the relative ac-dc difference is
III. Thermal Voltage Converters (TVCs)
Each of a new set of TVCs developed at NBS for this
work consists of one or more cylindrical metal-film resistors in a coaxial metal cylinder. The resistors can
be connected in series with one of two S mA TEs
mounted in separate cylinders so that the output EMF
is then a measure of the input voltage. Like an earlier
set developed at NBS [3], the effect of the reactance
on the frequency response can be computed at least
roughly from the simple geometry. Adjacent ranges of
these TVCs may be compared with greater certainty
than those of the older design, which had integrally
mounted TEs.
As shown in Fig. 1, a single two-watt resistor is used
for each range below 200 V. The higher-voltage resistor
units have two or more resistors in larger cylinders, with
two adjustable inner shields to control capacitance currents, as shown in Fig. 1, and have ports and baffles to
permit the controlled flow of cooling air from the Iaboratory compressed-air supply.

*i * = s(A> Dt)
where Da and Dd are the averages of Dx with alternating
and direct currents applied, respectively. The sensitivity factor s can be determined in one of several ways, is
indicated in the Appendix, such as by changing the current in the heater of TEj by a small known amount and
observing the resulting deflection of D\.
The sequence of readings, nearly equally spaced in
time, greatly reduces errors from drifts, temperature
changes, etc., in either TE and from either constant or
slowly changing thermal EM Fs in the comparator. The
comparator accommodates any reasonable range of
thermocouple EMFs, which need not be equal, and the
small currents (up to 20 uA) in the thermocouple circuii
do not affect the measured values.
With this circuit, small fluctuations in the supply
(heater) current do not appreciably change the deflection if the two TEs have reasonably equal proportional
changes in EMF (AE/E), with a given change in heater
current and reasonably equal time constants. Although

282-261

TABLE II
NBS "Model F" Thermal Voltage Converters

D
I

"1

S,
/H
m
E

s2
R
r

HH
m

Resistors
Sr (ppm)J
Total
ka
*Ht
No. (each)
2 kHz 20 kHz 50 kHz
0.5-A* none
0.1
+1
+1
+ 15
0.5-B* none
0.1
-4
-2
+ 11
1
1
0.1
0.2
-2
0
+ 11
2
1
0.3
0.4
-3
-2
+ 3
3
1
0.5
0.6
-1
0
+ 3
5
1
0.9
1.0
-1
0
- 1
10
1
1.9
2.0
0
+1
- 1
20
1
3.9
4.0
0
+1
- 1
30
1
6
6.1
+1
+1
- 2
50
1 10
10
+1
0
- 4
100
1 20
20
+1
0
- 7
200
4 40
40
+1
0
- 7
300
6 90
60
+1
0
3
500
8 200
100
+1
-6
+ 13
TEs A and B as TVCs, without series resistors,
t Including 0.5 V TE. Resistors connected in series-parallel for
higher ranges.
t Sr is the impedance error only (see text) computed from Table
III with either TF. of pair #1. A similar table with the #2 pair differs
from it by not more than I ppm at 2 and 20 kHz except at 0.5 V
range.
Range V

R
s

V
r
E

Fig. 1. Essentials of thermal voltage converters (construction de

tails not shown).
A Thermoelement unit, with S mA 100 a TE, for 0.5 V, in brass
cylinder, 4 cm long and 5.1 cm diameter, with brass end
plate D.
B Low-voltage resistor unit (1 to 100 V) with one 2-watt metal
film resistor R in brass cylinder 11 cm long and 5.1 cm di
ameter.
C High-voltage resistor unit (200-500 V) with multiple 2-watt
metal-film resistors, in brass cylinder 18 cm long and 6.4 cm
diameter with brass end plate D. Air inlet at I, outlet holes
at other end (not shown).
Si, Sj Inner brass shields 5.5 cm long, 5.1 cm diameter, connected
as shown, with no end plates. 5a can be moved parallel to
axis and locked in place.
P Non-conducting air-baffle plate with holes.
E Coaxial connector GR874.
F Two-pin output connector AN 10SL.
it would be difficult to calculate the resultant "stabi
lization factor" of the circuit, in practice, fluctuations are
much less than those observed when the output EMF
of each TE is balanced against a constant voltage
source, as in earlier comparators [4], Thus, with this
circuit and technique, TEs may be intercompared with
an imprecision which is much less than the fluctuations
of the supplies used.
R\ consists of ten 100 fl resistors mounted on an enclosed-contact silver-alloy switch whose thermal EM Fa
are much less than 0.1 piV when the wiper is stationary.
Rt is a 100 J2, 10-turn helical resistor of manganin,
with a manganin wiper to minimize thermal EMFs.
Detector D\ consists of a commercially available pri
mary galvanometer, incorporating photocells in an ad
justable negative-feedback circuit, which serves as a
current amplifier to a secondary galvanometer. The
scale factor of the combination is about 10 pA/mm,
and the input resistance (which is dependent on the
negative feedback) ranges from 200 to 3000 JJ. The
primary galvanometer is liquid filled to reduce distur
bances from mechanical shocks and vibrations and, with
its period of two seconds, is an excellent mechanical lowpass filter. A similar pair of galvanometers is used for Z)j.
Both secondary light beams appear on one scale so that
the observer may quickly read the deflection of D\ if it
is stationary and if Dt is in the desired range. (Residual

283-262

Fig. 2. Thermal voltage converters.

Fig. 3. Essentials of EMF comparator.

deflections of D2 are less critical if the TEs are well

matched.)
All of the components of the comparator except the
detectors are mounted in a metal box behind a metal
panel to provide electrostatic shielding, and the box is
lined with foamed plastic to reduce temperature fluctua
tions. The metal cases of the primary galvanometers are
connected to this box by the shields of two-conductor
cables.
With Ki and Ri at their midpoints and each of the
thermocouple inputs to the EMF comparator shortcircuited, peak-to-peak fluctuations in the secondary
galvanometer of D\ in periods of 10 seconds amount to 1
mm at the gain setting normally used. With 7 mV at
each thermocouple input, this corresponds to a value of
81 82 of 2 ppm. This is only about four times the calcu
lated Johnson-noise value for this circuit.
A comparison of two TEs (or TVCs) usually consists
of four or more determinations of Si 82 by the method
described. Three times the standard error (standard de
viation of the average) computed from a number of
typical series of such measurements is about 2 ppm.
This is a reasonable measure of the uncertainty in the
comparator. Comparisons of various pairs taken from
three TEs, with the TEs interchanged in the comparator
and with the EMF of first one TE and then the other
held constant in each sequence, have disclosed no system
atic error as large as 1 ppm.' These measurements were
made with commercially available dc and ac voltage
sources, having peak-to-peak instabilities, observed with
a TVC in 10-second intervals, of about 10 and 20 ppm,
respectively, at 2 kHz.
A TVC can be used for ac measurements with appro
priate dc standards by balancing its output EMF
against an adjustable stable voltage source such as the
potentiometer P. To reduce errors from drifts, a similar
sequence of ac and dc readings can be used, with the
setting of P fixed, and the direct voltage Vj preset and
accurately known. Then
V. = V[l + i + s(Da - Dt)]
where s, the sensitivity factor, is determined as before.
V. Comparisons of TEs
The fourteen reference TEs were compared, using
sinusoidal currents at frequencies of 20 Hz, and 2 and
20 kHz. All of these measurements (as well as the com
parisons of TVCs) were made in a shielded room to
reduce electromagnetic interference, which is particu
larly troublesome with TEs because of their high sensi
tivity and very wide frequency range. The step down
from one current to the next was made by connecting
J If the net dc reversal difference of the two TEs is more than ,?0()
ppm, an error is introduced by nonlinearities in D\. This can be mini
mized by taking determinations with first one and then the other TE
EMF held constant.

the heaters of two low-current TEs, a and b, in parallel

and their thermocouples in series aiding, and consider
ing the combination as a single TE connected in the TEj
position of Fig. 3. 4 As shown in Appendix II, if the
TEs are reasonably well matched, the ac-dc differeno;
of the parallel combination is simply 8p=(8a + 8) 2
to better than 1 ppm. From the test, &, &i = Ki where
81 is the ac-dc difference of the high range TE in position
2, and K\ = s(Da Dd). The two low-current TEs are
then connected in series and compared to determine
Ki = ic- Si. Thus, S0 = Xl+S,-r-XJ/2 and S^^K^o,
-Kill.
The network of comparisons at 2 kHz and the results
are shown in Fig. 4. Here the circles identify individual
TEs and the rectangles represent a pair in parallel. The
arrows indicate comparisons, and the numbers on them
are the values of the relative ac-dc differences in ppm
i.e., the ac-dc difference of the TE at the point of the
arrow is greater or less than that at the tail by the
amount shown. These tests have extended over a period
of three years during the development of this project;
in many cases, each number is the average of two or
more tests, each consisting of four or more sequences
of ac-dc readings already described.5
As indicated by Wilkins, the ac-dc difference of the
two NPL TEs should, theoretically, be less than 1 ppm
In tests at NBS at 20 raA, the two agreed to better thar.
0.2 ppm with a 3<r imprecision of 0.3 ppm.' Thus, it b
reasonable as well as convenient to start with an assign
ment of zero ppm to the ac-dc differences of each. Then,
following principles given by Youden [5] m weighting
the various paths in a network, it is possible to calculate
the ac-dc differences of the other TEs. The resulting
values are given to the nearest ppm in each circle and
in Table I.
The average for all of the conventional TEs is only
0.3 ppm. Thus, these simple TEs agree with the more
complicated multijunction converters very closely, and
essentially the same values would have been obtained
for each TE on the alternative basis of assigning zero to
the average of the entire group.
A complete but less redundant chain of measure
ments was carried out at 20 Hz to evaluate low-fre
quency errors of each TE. The precision was less at tha:
frequency because of much larger instability in the ac
source. The results of the values assigned on the basis
of zero error for the NPL TEs are given in Table I.
Not enough TEs were intercompared at 20 kHz to
assign independent values to each, because ac-dc dif* To minimize certain aberrations discussed in Section IX, resis
tors of at least live times the heater resistance were added to each of
the heaters before paralleling them.
1 Because measurements were made over an extended time in a
developing program, the network is not symmetrical.
* The EMF was higher and a higher detector sensitivity could be
used than with other TEs since the two TEs were well matched, s
that an extraordinarily high precision was attained.

284-263

Fig. 4. Intercomparisons of NBS TEs, and assigned ac-dc

differences in ppm at 2 kHz.

ferences from all known sources [3] should be the same

as at 2 kHz within 1 ppm. No measured values at the
two frequencies differed by more than 2 ppm. Thus, the
same values were assigned to each TE at the higher fre
quency.

where the subscripts a and d refer to alternating and

direct current (or voltage). Since 7M and I* produce
equal output EMFs, /M = 7^(1 +5.). If we let Z,
= i?((l + 8,), we get, if second-order terms are negligible,
5,-=S,+Se. Thus, 5, can be determined if 5, can be de
duced from the comparisons of adjacent ranges, since
8e can be evaluated in terms of the reference group of
TEs.
Two complete sets of adjacent-range comparisons
from 0.5 to 500 V were carried out at frequencies of 2,
20, and 50 kHz, with two different pairs of 5 mA TEs.
These TEs had 80Ni20Cr heaters bufwere otherwise
similar to those in the reference group. Measured values
of 8 as current converters were 1 or 2 ppm, except at
20 Hz where they ranged from 6 to 9 ppm.7 The
results of the adjacent-range tests given in Table III
showed excellent precision. In addition, the inter-range
differences were less than 2 ppm through 20 kHz, except
for the lowest and highest ranges. The differences of a
few ppm observed on the lowest ranges were rather
puzzling and will be discussed in Section IX. They
were essentially independent of frequency through 20
kHz but were negligible above the 1 volt range.
TABLE III
Comparisons of Thermal Voltage Converters

VI. Comparisons of TVCs

The TVCs were designed so that adjacent ranges
could be compared to step up and down the voltage
scale. Their nearly identical 5 mA TEs were each first
compared with one of the reference group of TEs as
current converters, i.e., with the heaters connected in
series.They were then paralleled through a" T" connector
and compared with each other as voltage converters
(basic range 0.5 V). The lowest resistor (see Table 2)
was then connected in series with one TE, say TE,
to make a one-volt TVC. This was connected in parallel
with the other, TEi, and compared at 0.5 volt. A twovolt TVC was then formed with the next resistor and
TE, and compared with the one-volt TVC at 1 V. This
process was continued until the highest range was
reached. If the ac-dc difference of each combination is
independent of the voltage level, the ac-dc difference of
any range may be found in terms of one as a reference.
This requirement is easily met if the TEs have no signifi
cant errors, for the reactance errors of each resistor unit
should be independent of voltage. The validity of this
assumption itself can easily be verified by making each
comparison at two voltage levels.
If we define Zi Vi/h for a TVC, where V< is the
magnitude of the applied voltage and /0 is the magnitude
of the current through the TE, then, from the definition
for ac-dc difference, applied to a TVC,

TVC 1

TVC 2

Voltage
V

A
B
1-B
1-A
2-B
3-A
5-B
10-A
20-B
30-A
50-B
100-A
200-B
300-A
500-B

B
A
A
B
1-A
2-B
3-A
5-B
10-A
20-B
30-A
50-B
100-A
200-B
300-A

0.5
0.5
0.5
0.5
1
2
3
5
10
20
30
50
100
200
300

2 kHz
#1*
-3 -3
+2
+2
-1
+2
0
+1 +1
0
+1
0
0
0 +1
0
0

Si It in ppm
20 kHz
50 kHz

-4
0
+2
-2
+2
0
+1
0
0
-1
0
0
0
-6

n n
+2
-3 -4
-2
0
+1
-1 -8
+1
0
0 -4
0
0
0
+1
-1
0
0 -2
0 -3
-1
0
0 +4
-5 + 16

n
-4
+3
-4
-1
0
-2
-1
0
0
-2
+4
+6
+8

* Test flWith one pair of 5 mA, 100 ohm TEs.

t Test flWith another pair of 5 mA, 100 ohm TEs. These tests
made three months after those with pair 1.
Except for this, the results of the comparisons are in
excellent agreement with computed values (see Section
VIII), which show that 8, should be less than 1 ppm
through the 100 V range. It is extremely unlikely that
any single source or combination of sources of ac-dc
difference should produce the same 8, in each of these
ranges. Thus, the authors believe that, from the con
cordance of theory and experiment, it is safe to assign
zero ppm to the average of the values of 8, from the 2 V

V' IO V,
'it
a. =

' These TEs, which were the only ones available at the time with
the correct heater resistance, will be replaced by others.
285-264

through the 100 V ranges in establishing the values given

in Table II from the comparisons of Table III.
The comparisons at SO kHz (which were less precise)
were made to verify that ac-dc differences were not
sharply dependent on frequency at 20 kHz. They indi
cate that excellent accuracy should be expected for most
ranges at even higher frequencies.
A complete set of measurements at 75 percent of
rated voltage was also made at 2 kHz with TE pair
#1 in order to investigate voltage coefficients. These
results agreed with the rated-voltage measurements to
1 ppm, with no systematic difference.
A completely independent verification of the accuracy
of these reference TVCs has come from some careful
comparisons by Flach and Marzetta [8] of a "time-gate"
peak ac-to-dc voltage comparator (designed by Mar
zetta) with an rms comparator which incorporated a
TVC that had been calibrated with the reference set.
In a long series of tests at 400 and 1000 Hz, where the
error in the comparison introduced by distortion of the
ac source was less than 10 ppm, the two comparators
agreed on the average to 4 and 8 ppm, respectively, at
the two frequencies.
VII. Thermoelectric Errors
Hermach's treatment of the ac-dc errors arising from
Thomson effects in the heater of a conventional TE [4]
has been extended by Widdis [6] to include the effects
of Peltier heating, and to correct for the heat abstracted
by the thermocouple. Widdis' expression for the Thom
son ac-dc difference, it=a*B/\lpk, is the same as that
found by Hermach for high-range TEs, where a, p, and
k are, respectively, the Thomson coefficient, electrical
resistivity, and thermal conductivity of the heater
alloy, and 6 is the midpoint temperature rise (100C to
150C in a conventional TE). The error can thus be
computed reasonably well if a can be determined. The
Thomson heating of copper is reasonably well known
[7], and is not much affected by moderate temperature
elevations or by the cold working of ordinary handling.
Thus, this metal can serve as a practical reference. The
authors have calculated the Seebeck and the absolute
Thomson coefficients of a number of pertinent alloys
from careful measurements of the EMF vs. temperature
of thermocouples formed from each of these alloys and
copper. The results are shown in Table IV, along with
the calculated ac-dc difference &, (using Widdis' for
mula) in a TE having midpoint heater temperature rise
of 150C (giving a 10 mV output with a typical NiCrconstantan thermocouple). With Evanohm or Karma,
5( should not exceed 1 ppm. At an output of 5 mV it
should be less than 1 ppm, since both 6 and a are
smaller.
Additional measurements indicated that the thermo
electric coefficients may be sensitive to cold working
and may be changed by exposure to high temperature.
For Evanohm and Karma significant changes were ob286

TABLE IV
Thermoelectric Data and Errors
Metal
or Alloy

a. MV/C*
at 400K

5
vrc

Copper
Evanohm
Karma
Manganin
80Ni20Cr
60Nil6Cr24Fe

-2.0
-1.5
-1.0
-3.0J
-3.9
-2.3

+0.3t
+ 1.8
-1-8*
+4.4
+0.8

Calculated Error in a TE
Ac-dc
Dc
reversal** difference
S PPni
A///, ppm
200
ISO
400
500
300

<1

* a (Thomson coefficient) is proportional to the absolute tem

perature.
f S is Seebeck coefficient vs. copper.
t This was quite dependent on neat treatment.
|| From Thomson heating, with =150C and p* = 25 X
10-i V/C at 500K, the weighted average temperature of the heatr
With the bead off center (see text).
served at 800C, but not at 400C. Thus, care must be
taken in controlling the temperature of the wire nea:
welded joints and when out-gassing during the construc
tion of TEs. Here the use of TEs in the reference group
by four different manufacturers having differing tech
niques provided powerful assurance against unknoar.
errors.
As indicated by Wilkins, the thermoelectric errors are
greatly reduced in the NPL TEs because of the bifilar
heater and the low temperature rise (about 10C at 20
mA in the present TEs). The resulting error should be
much less than 1 ppm, even with nickel-chromiurr
heaters.
Rather unfortunately, commercial TEs have lead-in
wires (see Fig. 1), through the glass bulb to the heater,
of a copper-coated nickel alloy to match the therniil
expansion coefficient of the glass. The combination has
a rather uncertain but fairly large Seebeck coefficienvs. the heater alloys used. With direct current applied
Peltier heating and cooling of the junctions of these leadin wires and the heater can cause a temperature differ
ence (+ at one junction and at the other). Most of thi.heat is conducted by the lead-in to the thick glass pre?.or disk which forms the base of the glass bulb. This c;:.
be considered as a "heat sink" so that the first-order
junction temperature change is A8=TIS/K, where T
is the absolute temperature, I is the current, .S is the
Seebeck coefficient, and K is the thermal conductance
of the lead-in. Widdis showed that this leads to an ac-dc
difference term 6,p = cxM/Vk where a is the Thomsw
coefficient and Vk is the heater voltage. There mav. in
addition, be a second-order term at the midpoint be
cause the Peltier heat itself depends on the junction
temperature. This can be calculated by carrying out the
second-order approximation mentioned by Hermacb r
Appendix 2 of reference [4] to give, with reasonable
accuracy, hv = -f-Ac?y Ttfl, where Ta is the heat-sink tem
perature. Of course, A0 can be calculated only verroughly, but this is sufficient. For a typical 10 mA TF
265

it is less than 0.01C, and the Peltier errors, Slp and 5,

(independent of frequency), are less than 1 ppm. This
has been verified by comparisons with Evanohm-heater
TEs which had copper lead-in wires butt-welded to the
nickel alloy wires at the glass press. However, Sp in
particular could be significant in higher-range TEs.

IX. Other Errors

Peltier effects at the heater and lead-in junctions can
also cause a second-order change in the effective dc re
sistance of a TE. The calculation is difficult and the re
sults uncertain, but the change should be much less than
one ppm for the 5 to 20 mA TEs in the NBS reference
group. Thus, this cannot account for the somewhat
larger changes in effective dc resistance implied by the
VIII. Low- and High-Frequency Errors
voltage comparisons of the TEs alone (without series
As the frequency is lowered, the temperature of a resistors) in the TVCs. Even larger and more puzzling
TE begins to follow the cyclic fluctuations of the square changes have been observed in some other TEs. There is
of the heater current. If the TE is not a linear device, good evidence that they may be caused by rectification
the average temperature rise may no longer be the same at oxide films, which may persist in spite of the welded
as at higher frequencies. Hermach's solution of the re joints, but there is as yet no direct proof of this.
sulting nonlinear differential equation for the tempera
Dc reversal errors (A/// for the same EMF) provide a
ture rise has been verified experimentally [4], at least good clue in studying these secondary effects. As shown
for high-range TEs having large heaters in which the in Table IV, AI/I calculated by Widdis' formulas
heat abstracted by the thermocouple is negligible. How should not exceed, perhaps, 200 ppm with Evanohm or
ever, for the low range TEs we are concerned with now, Karma, and 500 ppm with nickel-chromium heaters,
these formulas may not apply. The solutions for such even if the thermocouple is mounted off the midpoint by
cases would be even more difficult, and have not been 5 percent of the length of the heater (which is easily
attempted, particularly since the thermal mass of the noticeable). Larger differences are, thus, a sign of other
bead should then also be considered. In the simpler difficulties.
case, Hermach showed that the ac-dc difference is
As Hermach showed [4J, the current from the heater
bi = hdd'l/tiiHA, where d is the diffusivity of the heater
to ground through the bead of an ungrounded TE can
alloy, 21 is the heater length, u = 2wf, where / is the fre cause heating of the bead and, thus, an error. This is
quency, and h is a factor expressing the nonlinearities. difficult to compute because power dissipated in the
The same relationship might be expected for the more bead is somewhat more effective in raising the thermo
complicated case, but with a different numerical factor. couple temperature than that in the heater, and the cur
Thus, the agreement to better than 10 ppm among the rents in the two halves of the heater are no longer identi
TEs at 20 Hz, shown in Table I, with their varying cal. However, rough calculations indicate that with one
lengths, gives strong verification to the absence of sig volt across the bead of TEi in the circuit of Fig. 3, the
nificant low-frequency errors down to this frequency. bead resistance of a 10 mA TE must be greater than 500
These reference TEs have heaters less than 1 by 10-' megohms if the error is to be less than 1 ppm. For the
cm in diameter, with resistivities of 120 jifi cm. The bead materials commonly used, the resistance is a sharp
computed errors from skin effect are much less than 1 function of temperature and is much less at audio fre
ppm even at SO kHz. The ac and dc conductances and quencies than with direct current. At NBS, bead re
the capacitance between binding posts on the phenolic sistances are normally measured with a 2 kHz ac bridge
top of the box in which the TEs are mounted are much with rated direct current through the heater of a TE.
less than the limiting values of 0.01 jitA/V and 100 pF, The results have been verified (within a factor of 2) by
respectively, which might cause 1 ppm errors at audio ac-dc comparisons of pairs of TEs with added resistance
frequencies.
between the two heaters to significantly increase the
Reactance and skin-effect errors in the resistor units bead voltage.
of the TVCs, computed according to formulas developed
by the authors [3], are also less than 1 ppm through 50
X. Conclusions
kHz up through the 100 volt range, and the comparisons
verify this. They cannot easily be calculated for the
A reference group of two very different kinds of TEs
higher ranges, which have several resistors and adjust has been established. The authors believe that it is very
able inner shields. Additional reactance errors can arise unlikely that the average ac-dc difference of the group
from the lumped capacitance C of the connector be differs from zero by more than 2 ppm at audio fre
tween the resistor and the TE. This shunts the heater, quencies. This is based on the agreement to better than
which at these frequencies can be considered as a lumped 0.5 ppm of the averages of the two kinds of TEs, and on
resistor R in series with an equivalent lumped in allowance of 1 ppm for the uncertainty of the average
ductance L. It can be shown that the resultant effect of in the network of comparisons. The results are in excel
this capacitance on the magnitude of the heater current lent agreement with calculations, which show that the
approaches a112 a/3 as the series resistance (voltage ac-dc differences should not exceed 1 ppm. With the
range) is increased, where a=u>CR and 8 = wL/R precise techniques that have been developed, pairs of
(a and /31). The computed error of less than 1 ppm at TEs of the group may be compared with a 3a uncer
50 kHz has been verified by deliberately doubling the tainty of about 2 ppm without any evidence to date
capacitance.
(in over three years of measurements) of significant
287-266

systematic errors. This is somewhat remarkable, since

fluctuations in the best available wide-range ac and dc
sources greatly exceed this, and most of these TEs have
rather large residual thermal EMFs and have tempera
ture coefficients of EMF exceeding 1000 ppm/C.
Adjacent ranges of the TVCs may also be compared
with the same precision to step up and down the voltage
scale. Their ac-dc differences can be determined to bet
ter than 10 ppm when used with TEs of known ac-dc
differences, again without evidence of systematic error.
Alternating voltage measurements can be made to 20
ppm or better by using the TVCs or ac-dc transfer
standards with appropriate dc standards.
This accuracy and precision in the ac-dc transfer,
more than ten times better than heretofore available,
are good indications of the virtue of simplicity in refer
ence standards. Properly designed and constructed TEs
are rather simple devices that correspond to the desired
idealized mathematical models remarkably closely, thus
providing the agreement between theory and experi
ment which is so desirable in a development such as this.
The agreement to better than 1 ppm between the con
ventional TEs and the new very different multifunction
NPL TEs (which should theoretically have smaller er
rors), in addition to providing semi-independent verifi
cation of the accuracy of both, indicates that even
higher accuracy may be possible in the future with these
new designs. Conventional single-junction TEs could
be improved by the use of bifilar heaters and closely
spaced lead-in wires with good thermal conductance
between the two halves of the heater and between the
two lead-ins. In this way, other heater alloys could be
used.
Well-designed and well-constructed ac-dc transfer
standards can be remarkably stable over most of their
frequency range. Most factors which may change the
response, such as temperature and dimensional changes,
have little or no effect on the ac-dc difference. There is
already some evidence that long-time stabilities of a few
ppm can be attained in properly designed and con
structed TEs and TVCs. However, we have noted an
apparently abrupt change of +5 ppm in one TE, ac
companied by a change of over 100 ppm in its dc reversal
difference. Thus, TEs should be compared frequently if
the highest accuracy is desired.
Comparisons of adjacent ranges, with a step-up pro
cedure such as the one described, can be used to guard
against unexpected changes in TVCs. This has already
been verified in the design and construction of two some
what similar sets of TVCs by W. Scott, K. Ballard, and
D. Bailey of the NBS Electronic Calibration Center,
Boulder, Colo. These units incorporate large wirewound resistors of controllable reactances and low tem
perature coefficients, so that they do not require trouble
some forced-air cooling. They will be described else
where.
The step-up procedure is also feasible with simpler
sets of TVCs incorporating fewer plug-in resistor units
and two TEs of different current ranges. Two such sets
have been built but have yet to be completely evaluated.

Appendix I
EMF Comparator
If the thermocouple resistance Rn of TEi (Fig. 3) is
much less than Ri+Rt, the EMF E\ is substantially the
same with alternating and direct currents applied during
the test (the procedure involves holding the voltage at
the output terminals of TEi constant). Then, from the
definition of ac-dc difference, (/ Id) /Id = S\. Similarly,
for TEi (,I* Id')/Id' = &i, where 7/ is the direct current
that would be required to produce the same EMF ot
TEj as /. The change D.Dd in detector Di which
occurs when 7 is changed from 7. to Id (Id being ad
justed to keep Ei constant) results from the change in
Ei. Da Dd may then be interpreted as a measure of the
departure of Id from the value Id which would have
produced the same Et as would I. Thus, combining the
above two equations and neglecting second-order terms,
we have (// Id) IId = Si St.
The sensitivity of the detector to small current dif
ferences in TEi can be determined by shunting its
heater 2? with a resistor
to cause a small change
in its heater current A7, noting the change in deflection
AD of Di and computing s=AI/IAD (AI/I = RA/R4 if
the heater current of TEi is unchanged). Then, if j is
constant over the range of deflections involved, we have
Si St = s(Dt Dd), as in the text. Similarly, in the com
parison of two voltage converters, a resistor R.<CRm car.
be added in series with TVd to give proportional change
in applied voltage A 7/ V = R./Rm, where R is the TVC
resistance. Then s =AV/ VAD.
These two calibration methods must leave the current
in TEi unchanged. The change AI can often be intro
duced and measured more conveniently in some way
(as by changing the supply voltage) that changes the
current in both TEs by a small known amount. We have
found it convenient to disconnect Ei, set the voltage of
P to the same value, short D%, observe the resulting AD
when I is changed, and compute i as before.
If only a few TEs are ordinarily used in the "stan
dard" position (TEj in Fig. 3), it is convenient to make
use of the approximate output-input relation (at a fixed
frequency) for a TE, AE/E = nAI/I, where n is depen
dent on 7 but not A7 if A77, and is close to 2. For each
such TE, n can be determined in advance at several
values of E and plotted vs. E, since it is quite stable.
Then the sensitivity of the detector in the circuit can be
determined by inserting a small known voltage A7f,
(with a Lindeck potentiometer, not shown in Fig. 3) in
series with and opposing E%, observing the change AD
in Di and computing the factor k=AD/AEr. Then, from
this and the previous equations Si St = (Id'Id)/h
-(Ed Ed)/niEt = {Da Dd)/kmEt. Et is easily mea
sured with potentiometer P.
If Ra is not negligible, E\ is not constant during the
test, since the currents in 7?i and Rt depend on both :
and Et. The corresponding equations for determining
Si Si by means of deflections of Di then contain addi
tional correction terms involving the resistances of the
several parts of the EMF comparator, and are too com

288-267

plex to be useful for routine measurements. They show,

however, that if Ra<(R\+Ri)/\00, the simplified equa
tions already derived are valid to a few percent of the
computed 5i 5j. They show that Rtt is not critical. For
our comparator, Ri+Rt= 1100 and Rt, < 10 S2 for the
usual TEs.
For each NPL converter, 7?i1000 2. It is always
placed in the TEj position, except when the two NPL
converters are intercom pared, requiring the use of the
more complicated formula.
Appendix II
REs in Parallel
By using the definitions for and 6 of a TE (see
Appendix I) we have, closely enough for small changes
in E,
E, Ed
I I*
=
nEt
Id

8.

In the ac-dc test, nominally equal resistors are con

nected in series with TE and TE&, and the two branches
are connected in parallel. The sum of the two EMFs,
+&, is held constant. Then, with second subscripts
to identity the two TEs, we have
Eaa + Eah F-da Edb = 0
and
(Ida - Id*\

(Idb Idb\

If the reactances of the two paths are equal

Iaa Ida Iah Idb la ~ Id
=
=
= 5,
Ida
Idb
Id
where Ia and Id are the sums of the alternating and
direct currents in the two branches.
If Ea = mE and t=(l m)E,
,
, Via ~ Idl
[nam + Wb(l m)\ |^
J = tiatnSa + (1 m)8b.

If the reactances of the two paths are not equal, addi

tional terms will be introduced which are difficult to
calculate. However, they are zero if aw = nt(l m),
i.e., if the TEs are matched in response. Then 5,
= (8.+W/2. Even if the reactances are unequal, the
added terms are essentially eliminated if a second test
is made with the two resistors interchanged and the
results averaged.
Acknowledgment
The authors express their appreciation and thanks to
F. J. Wilkins for the two multijunction TEs, to L. Julie
of the Julie Research Laboratories for suggesting the
basic idea of the EMF comparator, and to the following
for their cooperation in the manufacture of other TEs
used in this group: the Levy brothers of the American
Thermoelectric Company, L. R. Graham of Graham
Associates, F. Gay of the Best Products Company,
M. Rosenfield of the former Field Electrical Instrument
Company, and L. W. Pignolet of Weston Instruments
Incorporated.
The use of these particular TEs with heaters of Evanohm and Karma should not in any way be construed as
constituting an exclusive endorsement by NBS of these
products or alloys. Other TEs, made to the same specifi
cations with the same or other alloys of equally low
thermoelectric effects, should be equally suitable.
References
[1] "American standard requirements for electrical indicating instru
ments." New York: American Standards Association, C.39.1,
1964.
[2] F. J. Wilkins et al., "Multijunction thermal converter," Proc.
IEE (London), vol. 112, pp. 794-806, April 1965.
[3] F. L. Hermach and E. S. Williams, "Thermal voltage converters
for accurate voltage measurements to 30 megacycles per second,"
Trans. AIEE (Communication and Electronics), vol. 79, pt. I, pp.
200-206, July 1960.
[4] F. L. Hermach, "Thermal converters as ac-dc transfer standards
for current and voltage measurements at audio frequencies,"
/. Res. NBS, vol. 48, pp. 121-138, February 1952.
[5] W. J. Youden, "Measurement agreement comparisons," Proc.
1962 Standards Laboratories Conf. Washington, D. C: NBS,
Misc. Pub. 248, August 16, 1963, pp. 147-152.
[61 F. C. Widdis, "The theory of Peltier and Thomson effects in
thermal ac-dc transfer devices," Proc. IEE (London), Monograph
497M, January 1962.
[7] Landolt-Bornstein, "Zahlenwerte und Funktionen aus Physik,
Chemie, Astronomic, Geophysik und Technii," 6. Auf., Band 1 1/6.
(6th ed vol. 2, pt. 6). Berlin: Springer-Verlag, 1959.
[8] D. Flach and L. A. Marzetta, "Calibration of peak a-c to d-c
comparators," Proc. 1965 ISA Conf., pt. 1, vol. 20, paper 14.

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT

DECEMBER

1966

289-268

A System for Accurate Direct and Alternating

Voltage Measurements
F. L. HERMACH, fellow, ieee, J. E. GRIFFIN, member, ieee, and E. S. WILLIAMS

AbstractA system for calibrating and using a group of dc and

ac standards has been developed at NBS to meet increasing needs
for greater accuracy in the measurement of direct and rms alternating
voltages at audio frequencies. With a group of saturated cells, a
universal ratio set as a potentiometer, a simplified self-calibrating
volt box, and a new differential-thermocouple ac-dc comparator, an
accuracy of 10 ppm can be obtained for direct voltage measurements
and 40 ppm for rms alternating voltage measurements between 20
and 20 000 c/s, in terms of the direct volt and the ac-dc transfer
standards maintained at NBS.
Introduction
THERE HAVE BEEN increasing needs for mea
suring direct voltages to about 20 ppm (parts per
million) and rms alternating voltages at audio fre
quencies to perhaps 50 ppm. A system for calibrating
and using a group of dc and ac standards has been de
veloped at the National Bureau of Standards in an at
tempt to meet these needs. It consists of 1) a group of
saturated standard cells; 2) a six-dial universal ratio set
(URS) with a control box so that it may be used as an
accurate potentiometer; 3) a simplified volt box of the
Silsbee type and a new convenient method for calibrat
ing it with the URS; 4) a new differential-thermocouple
ac-dc comparator; and 5) an inductive voltage divider
for low alternating voltage measurements and for cali
brating the URS. A dc picoammeter for checking cer
tain vital insulation resistances and suitable dc and ac
detectors and other apparatus normally available in a
standards laboratory are also required. With the system
of calibrations, the sustained accuracy in use depends
only on the standard cells, the very stable inductive
divider, and the ac-dc difference of the comparator. As
a result, the authors believe that after corrections based
on these calibrations are applied, direct voltages from
0.5 to 1000 volts can be measured with an uncertainty
(a "limit of error" which will only rarely be exceeded)
of 10 ppm, and that alternating voltages from 0.5 to
500 volts at frequencies from 20 to 20 000 c/s can be
measured with an uncertainty of 40 ppm in terms of the
dc standards and the standards of ac-dc difference main
tained at NBS.
Manuscript received May 3, 196S. This work was supported in
part by the Metrology Engineering Center, Bureau of Naval Weap
ons, Pomona, Calif., and By the Army Metrology and Calibration
Center, Frankford Arsenal, Philadelphia, Pa.
The authors are with the Electrical Instruments Section, Na
tional Bureau of Standards, Washington, D. C.

Universal Ratio Set as a Potentiometer

Potentiometers are normally calibrated with a URS.
which is in essence a resistance divider1 with a resolution
of 0.5 ppm of the input for a six-dial instrument. Thus,
its use as a potentiometer to obtain the highest accuracy
has seemed natural, and has indeed been recommended/
However, a number of very important precautions mus:
be observed before the accuracy realized begins to ap
proach the precision. We have found it necessary to de
sign and construct a control box to adapt the URS for
voltage measurements with reasonable convenience and
to minimize the various errors.
A URS can be used as a single range 0 to 2.1 volt
potentiometer if it is supplied with an adjustable stable
direct current, as shown in Fig. 1. To make the dials
direct reading, the current / is adjusted by first setting
the dials numerically equal to the EMF of the standard
cell (SC) and then adjusting R\ to null the galvanom
eter G with the cell connected. This eliminates a sepa
rate standard cell dial with its additional resistors and
leads to greater accuracy. It is not as inconvenient as it
may at first seem. The current drain of the usual 2100-2
URS is only 1 mA with direct-reading dials, so that wit':
six 14-ampere-hour mercury cells in series parallel, the
drift is less than 3 ppm/hour. Thus, the standard cei.
balance need not be made very frequently.
The control box shown in Fig. 2 contains all of the
components (except the standard cell) necessary for
voltage measurements to 2.1 volts with the URS. Is
this way the URS is unchanged and is available for re
sistance ratio measurements by simply disconnecting
a few wires. As shown in the figure, the potentiometer is
ungrounded and two inputs are provided for voltage
measurements, both of which may be reversed. Thisisa
necessity (often overlooked) for many types of mea
surements, such as the reversed dc calibration of watt
meters and attenuators.
The control box is of metal and is lined with foamec
plastic insulation for thermal stability. Each com pone-:
is insulated with bushings or subpanels of tetrafluoroethylene (TFE), and all wire is similarly insulated. Tf-f
insulation resistance must be greater than 2X10* Qtt
1 J. L. Thomas, "Precision resistors and their measurement.'
NBS Circular 470, 1948.
* Proposed American Standard C39.6, "Automatic digital volt
meters and ratio meters," ASA, September 1963.

290-215

controlled air bath and may be intercompared by means

of the URS or with a Lindeck potentiometer built into
the standard cell box.'
The commercially available six-dial URS used with
this control box is adjustable to 0.001 12 (corresponding
to 1 fiV at 1 mA). For convenience, its dials have been
marked in volts as well as ohms. Its principal resistors,
which have temperature coeflfcients matched to better
than 5 ppm/C, are mounted in a thermally shielded
inner enclosure, and are connected to the circuit by en
closed silver-alloy switches of low thermal EMFs. A
taught-band light-beam galvanometer, with a sensitiv
ity of 500 mm/nA and a near-critical damping range of
500 to 2000 12, is used as a detector because of its excel
Fig. 1. URS as a single-range potentiometer.
lent rejection of alternating currents at frequencies
above a few c/s. It is supported on TFE resting on a
metal plate, and is connected with a two-conductor
shielded cable having TFE insulation. To eliminate ex
ternal stray currents, this plate, the metal control box,
the case of the URS, and the cable shield are all
grounded. Because of the high insulation resistance, the
potentiometer circuits may be grounded at the point of
measurement rather than at the potentiometer termi
nals. It is undesirable to ground a potentiometer itself
unless the circuit being measured is very well insulated
because of small error-voltages which may result from
ground currents in the potentiometer leads.
When the dials of the URS are set to zero, a residual
voltage of a few
normally remains because of switch
and lead resistances. This is stable to better than 1/xV,
however, and is easily eliminated by the zero-adjust cir
cuit shown in Figs. 1 and 2. Resistor R2 is adjusted for a
galvanometer null with the URS dials wet at 0 and the
Fig. 2. Diagram of control box.
input switch in the "short" position.
The URS was calibrated by the Resistance and Re
insure that leakage-current errors are less than 1 ppm. actance Section of NBS, with dc resistors and a direct
With TFE insulation this requirement is easily met and reading ratio set. However, it is constructed with re
maintained in laboratory service. With all leads discon sistors of low electrical time constant. It can be checked
nected, the resistance of each circuit of the control box conveniently at low audio frequencies by direct com
to the panel and between each pair of input posts has re parison with a suitable calibrated inductive voltage
mained well above 10" 12 even at 60 percent RH.
divider.4 We have found the simple circuit of Fig. 3,
Resistor Ri in the control circuit must be adjustable described in Appendix I, satisfactory for this. The re
and stable to 0.001 12. Commercially available enclosed sults of tests at 100 and 400 c/s agreed with the normal
silver-alloy switches and a bifilar 0 to 0. 1 12 continuously dc tests to 1 ppm of input.
adjustable resistor make this possible. Resistors with
As an additional overall check the URS with its con
temperature coefficients less than 10 ppm/C are used trol box was connected so that its output voltage could
in the first four dials. The six-mercury cells are con
be measured with a Brooks-Harris wide-range potenti
nected in three parallel groups of two cells each, with a ometer. After corrections to both potentiometers were
1-kfl series resistor of low temperature coefficient to applied, the residual differences at 25 dial settings of the
reduce circulating currents. Insulated copper wire
(about 45 12) is wound around each cell and connected
in series with it to compensate (at the cell current of
1 A. \Y. Spinks and F. L. Hermach, "Portable potentiometer and
1/3 raA) for the temperature coefficient of EMF (about ihermostatted container for standard cells," Ret. Set. Instrum , vol.
26, p. 770, 1955
+60 pv/"C). The thermofree galvanometer key is
* The possibility of comparing resistance ratio devices with in
ductive
dividers at low frequency (which was first suggested
provided with a resistor of low thermal EMF across the to us byvoltage
F. K. Harris) has occurred to many people. The Resistance
normally closed contact, as shown in Fig. 2, for proper and Reactance Section of NBS has made such comparisons using
methods for injecting the necessary small quadrature volt
damping and a stable zero reading. The saturated different
age into the detector circuit. Their conclusions substantiate those
standard cells are housed in a separate temperature reached in our investigation.
291 216

Fig. 3. Circuit for calibrating URS with inductive voltage divider

(IVD). Ci0 to 1100 pF, 3-terminal adjustable air capacitor;
Cismall low-loss capacitor connected directly to URS termi
nals; and Dtuned audio-frequency null detector. Note: Each
dial of URS is set at midscale except decade under test.
URS did not exceed 2 /zV, even though some time had
elapsed since the calibration of the Brooks-Harris po
tentiometer.
A URS is normally calibrated by determining the in
crease in resistance (with A and G as the potential
terminals) for each setting of each dial from the resist
ance at the zero setting in terms of an arbitrary unit.
This can be expressed as R R<, = K(D-\-c), where
R Ro is this increase for a setting D of the dials (in
ohms or volts), c is the total correction, R0 is the re
sistance with all of the dials at zero, and it is a constant.
When the standard cell balance is made in the use of
the URS as a potentiometer, we have, from Fig. 1,
E.+ V0 = IR. = IK(D.+c.)+IR0, where R, is the re
sistance between the branch points of A and G, at the
combined dial setting D,.1 Similarly, when an unknown
voltage VT is measured, we have at balance Vx+V<,
= IK(Dx+cx)+IR0. With V, and all dials at zero, V0
is adjusted to make /7?o=V0. Then, since I is un
changed,
Vx = E.

(D, + O
(D, + c.)

It is convenient to set D. numerically equal toE, c,.

Then, very simply, Vx = Dx + cx.
Voltage Divider
A guarded volt box designed by Silsbee is based on the
principle that the resistance of any section is nominally
equal to the sum of resistances of one or more adjacent
lower sections. It can thus be calibrated with a directreading ratio-set and auxiliary resistors by a series of
substitution measurements.* Silsbee's volt box (133
J2/V) was originally designed for calibrating other volt
* D and c are expressed in terms of the voltage markings (mil)
when the URS is used as a potentiometer.
' B. L. Dunfee, "Method for calibrating a standard volt box,"
J. Res. NBS, vol. 67c, p. 1, 1963.

boxes by direct comparison, so that it has more ranges

than necessary for voltage measurements. In addition,
with modern insulation, its guard circuit is no longer a
necessity, and with stable, modern, high-resistivity al
loys, higher-valued resistors can be used with less selfheating. A simpler, much smaller and lighter 12-range
volt box embodying the same principle was designee
and constructed for this system.7 It is rated at 500 volt
at 1 mA (1000 fi/V) but it can be used with the 2 volt?
URS for measurements to 1000 volts with negligible
self-heating errors. Its oil-filled sealed resistors are
mour> ;d directly on the binding posts, which are insulateo from the top panel with TFE bushings. The
measured temperature coefficients of the resistors are all
less than 3 ppm/C, and the computed temperature
coefficient of each ratio is less than 2 ppm/C. The wood
box and an inner metal shield provide sufficient thermal
and electrostatic shielding for laboratory environments
If the insulation resistance from the circuit to the metal
shield (brought to a separate binding post) is greater
than 2X10" fl, leakage current errors are less than
1 ppm even if all of the resistance is lumped at the worst
possible place. In the three years that this divider has
been in use, this resistance has been greater than 10" G
even at 60 percent RH.
Because of the relatively few ranges of this divider {1
to 500 volts in the series 2, 3, 5, 10, etc.), a direct "stepup" calibration may be made very accurately and con
veniently with the URS, as shown in Fig. 4 and as de
scribed in Appendix II. Adjacent resistors of nominally
equal value are intercompared with the URS connected
as shown, and then with the leads at a and b inter
changed. The average of the two URS readings is essen
tially independent of URS errors and lead resistance
The leads at g and b are each moved up 1 range on the
volt box, posts 1 and 2 are short-circuited, and the pro
cedure is repeated. This is continued in accordance with
the schedule shown in Table I, until the top post is
reached. A complete calibration including computations
can be made rather quickly. As shown in the table, the
results agreed with the more accurate and more versatile
direct-reading-ratio-set method to 1 ppm.
In use, one pair of X posts of the URS control box
is connected to the 0-1 posts of the volt box, and the
unknown direct voltage Vic is applied to the 0 and ap
propriate range posts. Then
Vic = N'V.(l + n + w)
where, in addition to the terms already defined. A" is
the nominal ratio of the volt box, N'(l +w) is its actiui
ratio, and w = Ri/Rv, where Rt is the resistance of the
two leads from the volt box to the points at which volt
age is to be measured, and R, is the nominal resistance
of the volt box.
' The layout, construction, and initial tests of this box w
capably carried out by J. L. Yarne.

292-Z17

VOLT BOX

Fig. 4. Calibration of a volt box with the URS. Note: The values
shown on the volt box are the voltage ratings and also the re
sistance in kli between the marked tap and the "0" post. Twice
rated voltage may be applied without significant heating. Leads
aA and bB are air-insulated.
TABLE I
Calibration of Volt Box with a URS

5
0-1
0-1
0 2
0-5
0-10
O-10
0-20
O-50
0-100
0-100
0-200

Section
X

NY
Link

1-2
2-3
1-2
3-5
2-3

5-10

10-20
20-30
10-20
30-50
20-30

50-100

100-200
200-300 100-200
300-500 200-300

1
1
2
510
10
20
50
100
100
200

URS Reading
Rai
Rai
1055 +
.5510
.5540
.5680
.5675
.5600
.5585
.5400
.5490
.5545
.5560
.5540

1055 +
.5600
.5570
.5430
.5440
.5510
.5530
.5715
.5630
.5570
.5560
.5575

Voltage
Range
/v'-zav

0
0
+ 4.0
- 5.5
-17.0
-17.0
-21.0
- 9.0
- 2.5
- 2.5
- 1.0

+ 8.5
+ 3.0
-19.5
-28.0
-25.5
-22
+ 9.0
+ 4.5
0
- 2.5
+ 2.5

2
3
5
10
20
30
50
100
200
300
500

Ra
+ 8.5
+ 3.0
-23.5
-22.5
- 8.5
- 5.0
+30.0
+ 13.5
+ 2.5
0.0
+ 3.5

Volt Box
Correction
URS DRRS

Nt'x

+ 8.5
+ 3.0
- 39.0
-140
-255
-220
+ 180
+225
0
-250
+500

+ 8.5
+ 11.5
- 27.5
-168
-423
-643
-460
-235
-235
-485
+ 15

+ 4.0
+ 4.0
- 5.5
-17.0
-21.0
-21.5
- 9.0
- 2.5
- 1.0
- 1.5
0.0

+ 3.4
+ 3.9
- 5.9
-16.6
-21.0
-21.4
- 9.0
- 2.4
- 1.1
- 1.4
- 0.6

URS readings were interpolated from galvanometer deflections to j of last dial unit. All calculations are shown in ppm and are carried to
nearest 0.5 ppm (where significant) for comparison with DRRS method. Corrections in URS column are calculated from (ZNt'x)/N'. Cor
rections in DRRS column were determined by method.'
Ac-Dc Comparator
A new type of transfer standard, a differentialthermocouple voltage comparator (DTVC), was de
signed, constructed, and calibrated for ac measurements.
This contains two identical voltage converters, as shown
in Fig. 5, each consisting of a chain of resistors (to ob
tain various ranges from 0.5 to 500 V) in series with
the heater of a 10-mA thermoelement.8 The two thermo
couple outputs are permanently connected in opposition
and to a galvanometer G. In use, the two voltage con
verters are first switched in parallel (by means of Si) to
a direct voltage source (not shown in the figure) which
is nominally equal to the alternating voltage to be mea-

sured, and the outputs of the thermoelements are ad

justed (with the shunt R) to obtain a null on the
galvanometer. One converter is then switched to the ac
source, and a null on the galvanometer is again obtained
by adjusting the dc source. The alternating and direct
voltages at the terminals of the instrument are then
equal, and the direct voltage can be measured with the
dc standards already described (volt box and URS
potentiometer). To eliminate dc reversal errors in the
thermoelements, the direct voltage is reversed, and the
sequence of measurements is repeated. The results are
then averaged. As shown in Appendix III, the unknown
alternating voltage is simply V'c= VdcO+8) where Vie
is the corrected average direct voltage and 8 is the ac-dc
difference of the DTVC in proportional parts.
With the dual converters, errors from drifts and other
uncertainties, which tend to limit the accuracy of the

The hot junction of the thermocouple is electrically insulated

from the heater wire by a small bead. The lower pole of switch 5i
makes contact before the tipper one to keep the bead voltage low.
293 218

single thermoelement system, are greatly reduced. The

dual principle also makes possible simultaneous com
parisons of alternating and direct voltages with high
precision. These factors contribute to high accuracy in
measurements.
A dual hot-wire comparator was developed many
years ago.8 It consisted of two current-carrying parallel
wires with a small mirror cemented as a bridge at their
midpoints, to measure (with a reflected beam of light)
the differential sag caused by the heating of the wires.
However, transient unbalances because of temperature
fluctuations made very precise measurements difficult.
In any differential instrument, close matching of the
dual channels is necessary to achieve a precision of 10
ppm. This might seem to be impractical with thermo
elements because of their large temperature coefficients
approaching 0.1 percent/C. However, the temperature
coefficients are nearly equal. Furthermore, the two
thermoelements in the DTVC are mounted in two
closely spaced holes in a thermally well-shielded copper
block. Thus, the errors resulting from this source are
considerably less than 5 ppm in a typical measurement
under normal laboratory conditions.
Wire-wound audio-frequency resistance cards were
used for the two identically arranged series-resistor
chains to minimize self-heating errors. Repeated tests
of the 5 and 500 volt ranges of the completed instrument
with both voltage-converters in parallel have shown
that after a 10-minute warmup net, self-heating changes
do not exceed 10 ppm.10
A complete diagram of the DTVC is shown in Fig. 6.
To reduce reactance errors, each voltage range has a
separate resistor, and all unused resistors are connected
to their respective shields through the range switch. The
metal shields, which surround the components, prevent
interaction between circuits. Simple L-C circuits with
by-pass capacitors mounted on the thermocouple box
provide considerable attenuation at radio frequencies.
The difference in the lead resistance from the ac and dc
input connectors to the common point at the ac-dc
switch (Si) on the instrument is less than 200 nil. En
closed silver alloy switches with contact resistances
stable to better than 500 jifi are used. Thus, the total
error from unequal lead resistance is less than 15 ppm
on the 0.5 volt range and correspondingly less on higher
ranges.
The ac-dc differences of the alternating voltage con
verter of the DTVC were carefully measured with a
newly developed set of coaxial thermal voltage con
verters at NBS. This set of coaxial converters has
ranges from 0.5 to 500 volts in the sequence 0.5, 1, 2,
3, 5, 10, etc. By stepwise intercomparisons of adjacent
ranges, the ac-dc difference of each range of this set

has been determined relative to any one range to bet

ter than 10 ppm at audio frequencies. The 0.5 volt
range has been compared repeatedly with a group of
thermoelements, which at present serve as the NBS
standard of ac-dc difference. From these intercompari
sons, the ac-dc differences of the DTVC could be de
termined in terms of the average of the N BS group.
The results at 20 000 c/s are shown in Table II. The
measured differences were less than 10 ppm at lower
frequencies.
Since there are two converters in the DTVC, there
is considerable merit in comparing the ac-dc difference*
of each. Changes in the thermoelements or in the re
sistors which could slightly alter the ac-dc difference of
one converter would be very unlikely indeed to affect
both converters equally. It is easy to compare the ac-dc
differences of the two converters by noting the change?
in the galvanometer deflections with switch Si to the
right and with direct, alternating, and reversed direct
voltages successively applied to the dc input terminals of
the instrument. The two thermoelements are matched
well enough so that the applied voltages need be equal
only to 0.05 percent. Calibrated ac and dc sources may
be used or may be adjusted to give equal readings of a
millivolt potentiometer connected to one of the thermo
couple outputs. These checks can be made on each volt
age range and provide precise and very valuable veri
fications of the constancy of the ac-dc differences.11
In using the DTVC, it is sometimes more convenient
to adopt a deflection method rather than to obtain a
null at each balance. In this technique the scale factor
of the galvanometer is first determined, with the proper
direct voltage applied, by momentarily shunting the
heater circuit (switch S2 in Fig. 5) and noting the change
in deflection. Then, in a measurement at this voltage
level, the galvanometer deflection is read with the
double-pole selector switch Si, in each position, with the
nominal value of direct voltage applied. The direct volt
age is reversed and the two deflections are again read,
without resetting the dials of the URS or rebalancing
the galvanometer. Then, as shown in Appendix III.
the unknown alternating voltage is simply l"
= Fdc [l +s(Du; Z?dc) +5] where, in addition to the
terms previously defined, Dac and Dic are the average*
of the two galvanometer deflections with the selector
switch in the ac and dc positions, respectively, and i :the scale factor in proportional parts per division of the
detector.
A photograph of the complete system is shown in Fig
7. Each of the components is used for other measure
ments in other areas, so that console mounting seemed
impractical. The resistances of the two coaxial cable*
from the ac source (not shown) and the volt box to the
DTVC must be equal to 10 ppm of the resistance of the
DTVC range in use. Other leads are not critical.

E. F. Northrup, "A new instrument for the measurement of al

ternating current," Trans. AIEE, vol. 24, pp. 741-757, 1905.
11 Techniques for intercomparing the ac-dc differences of adjacent
10 After being energized, many thermoelements exhibit a drift up
to 0.1 percent, with a 2- to 4-minute time constant. No effort was ranges to step up and down the alternating voltage scale are no*
being investigated.
made to match these drifts in this instrument.
294-219

Fig. 7. Direct and alternating voltage measuring system. Note:

The differential-thermocouple voltage comparator is at the left,
the universal ratio set and control box at the center, the volt
at the rear and the "standard cellar" at the right.
l
Fig. 5. Simplified diagram of the DTVC. J?m = series multiplier,
100 tl/V, and Rh = resistance of TE heater 40 11 + plus additional
resistance to make SO R, 0.5 V range.

Kig. 6. Differential thermocouple voltage comparator. Note: All

range resistors not in use are connected to the shields through ad
ditional switch contacts as shown. The shields are shown as
dotted lines.
TABLE II
Ac-Dc Differences ok DTVC
Range V
0.5
1
2
3
5
10
20
30
50
100
200
300
500
315-411 O - if - 20

Ac-Dc Difference ppm at 20 c/s

-30
-10
+ 10
+ 10
+ 10
0
0
+ 10
+ 10
0
0
+ 10
+30
2 95

Accuracy
It is difficult to assess the accuracy of a complex
measuring system such as this, in which several stan
dards extend and subdivide the direct voltage scale and
transfer to alternating voltage with high precision. Ex
cept near 1 direct volt, no voltage standards were avail
able with stability and accuracy adequate to directly
determine the precision and accuracy of the complete
system.
As indicated in the text, the calibrations of the ratio
standards were checked by more conventional tech
niques to verify that unknown systematic errors in the
proposed new methods were not significant. For addi
tional verification, direct voltages from 0.5 to 1.5 volts
were measured simultaneously with the URS and the
Brooks-Harris potentiometer. In several repeated tests
the largest discrepancy (after calibration corrections
were applied) was 3 ppm. Similarly alternating and
direct voltages were compared at the same time by
means of the DTVC and the coaxial thermal voltage
converters. The DTVC was operated in its normal way,
and a sequence of readings was used to greatly reduce
errors from drifts in the coaxial converters. The largest
discrepancy was 10 ppm. The observed agreements
might lead one to believe that direct and alternating
voltage measurements could be made to these accuracies
if corrections were applied, but this is not necessarily so,
since these are only partial tests with standards which
are not completely independent. In the use of this sys
tem, it is indeed assumed that corrections for all known
sources of errors, as determined from the calibrations of
the several standards, will be applied, when significant,
in accordance with thd formulas given. However, even
after this has been done, there remain many residual
errors, which cannot be evaluated precisely. The essen
tial features of this system were developed in 1960, and
some components have been in use over three years. In
this time, enough experience in measuring direct and
alternating voltages has been gained to estimate the
upper limits of these residual errors (some of which are
proportional to measured voltage in a given range and
some of which are proportional to the range) and to ob220

1963
tain the estimates of uncertainties given in the introduction.
Conclusions
The system described has met the desired goals in
accuracy and in voltage and frequency ranges. Its con
tinued accuracy in service can be based on a few highly
stable standards by means of relatively direct and sim
ple procedures of high precision. Many of the com
ponents of the system are readily available in a stan
dards laboratory. Others can probably be constructed
without too much difficulty. Most of them can be used
for other measurements in addition to those described
here.
The DTV'C is much more stable than earlier singleunit ac-dc transfer standards, making more precise ac
measurements possible. Its ac-dc differences were as
signed with confidence only after several comparisons
with the NBS ac-dc standards verified both the preci
sion of the comparison process and the ac-dc stability
of the comparator. It is not to be expected that the same
certainty can be attached to the results of a single ac-dc
test of a similar instrument. However, ac-dc intercomparisons of adjacent ranges in the user's laboratory may
provide some of the additional assurance that would be
needed in such cases.
Appendix I
Ac Calibration of a URS
The circuit shown in Fig. 3 is easily set up in the
laboratory if an ungrounded source or a shielded trans
former is available. In this way a guarded detector is
not required, and the shield of the 3-terminal capacitor
may be grounded as shown. As is well-known, an induc
tive voltage divider can tolerate rather large shunt
capacitances. The change in ratio with 5000 PF con
nected to one side of the divider used in these tests w-as
undetectable (<10~7 of input). At low frequencies
(100 c/s is recommended), short unshielded leads are
satisfactory; their resistances are not critical, since only
changes in the effective resistance Rao corresponding
to changes in URS dial settings are measured. An input
voltage of 10 volts is recommended. At this level, the
resolution and noise level of the tuned detector should be
less than 10 fiW.
The capacitors C\ and Ci are necessary to achieve a
sufficiently precise balance but need not be known accu
rately, since their effect on the magnitude of the URS
ratio is small. This can be seen most easily by neglect
ing (for this evaluation only) the small lead resistances
Ra and Ri,. Then at balance V, = VpZa/(Za-\-Zb) w;here
Za = Ra /(l+jofi) and ZB = Rn/(l+jai) and ai=o>CiRA,
<xi=<i3CiRb- Then, if V, = MVr(cos y j sin 7), where M
is ratio of the magnitude of the output to the input
voltage of the IVD, we have

where
a2
\RA + rbJ

\Ra + rbJ

Equating real parts, using the series expansions of the

trigonometric functions, and noting that 7, cti, and a:
are small compared to unity, gives to the desired ap
proximation,
Ra
Ra + Rb
Thus, if the phase defect angle of the inductive volt
age divider (IVD) is less than 100 microradians, and ai
and as are each less than 0.01, the phase-defect-angle
correction terms are less than 1 ppm.
The calibration consists of measuring the ratios for
each setting of each dial of the URS. It is convenient
to set all of the other dials at their midpoints so that
the capacitance balance may be made with capacitors of
reasonable size. For the first calibration, the measure
ments should be repeated after the frequency of the
source has been doubled. If the results agree to 1 ppm
of input, there is strong evidence that the residual re
actances within the URS are negligible at both fre
quencies. If not, lower frequencies must be used.
The corrections to a URS are usually expressed in
terms of the change in the resistance Ra (between the
branch points at A and G) corresponding to changes ui
dial settings, with one hundredth of the average of the
20 resistors of the main (X100) dial as the unit of re
sistance. If the total resistance is constant, the correc
tion for a given setting of any one dial is independent 0;
other dial settings. This can be expressed as

where Rk, Rko is the increase of the resistance Ra at

the setting Dkn of the kth dial from the value Rm at the
zero setting Z?*o of that dial, ct is the corresponding cor
rection, and K is a constant. Similarly, Rv 7?10 = KDy.
with zero correction by definition, where Ry/ and Rv
are the resistances at the 20 and 0 settings of the first
(X100) dial.
From Fig. 3, at balance, and neglecting the react
ances,
Vag = V. - IRa = VTM - IR.= IRA.
Combining these equations, if the fcth dial is varied with
the other dials fixed and if / is independent of dial set
ting, gives,
Rkn RkO
Ru Rr

Dkn + Ckn
Dy

Mu - Af,

Thus,

VPRA(\ +jai)
= MVp(cos y j sin 7)
(Ra + Rb)(1 + jc)

ck. = Dul
296-221

) - Dk

TABLE III
Partial Calibration of a URS

1st
0
10
12
20
9
0
10
12
20

Dial Settings
2nd

Others

Frequency
c/s

s.sss

400

0
4
10
5

5. 555
5.555

Corrected
1VD Reading
-Mn
0263238
4999996
5947346
9736760
4289492
4478957
4763156
0263240
4999997
5947345
9736760

400
ioot

Correction, mSl
ac
dct

mIf.la AIS k

0.0
-0.6
-1.2
0.0
0.0
-1.2
-2.6
0.0
-0.6
-1 .6
0.0

4736758
5684108
9473522*
0189465
0473664
4736757
5684105
9473520*

0
-1
-1
0
0
-1
2
0
-1
-1
0

* Equals Aft/ + Ml
t Tests at 100 c/s were made on another day.
t From dc tests.
In this way the correction for the setting of each dial
can be determined in terms of the corresponding ratios
(M) of the IVD." Note that the lead resistances do not
enter into the formula.
A portion of the data taken in one of several tests of
the URS, and the computations to determine the correc
tions are given in Table III. These corrections in ap
propriate units apply to the URS when used for resist
ance as well as voltage measurements.
The constancy of the total resistance of the URS
must also be checked to 1 ppm. This can be done easily
by adding a stable 2000 0 resistor in series with the B
terminal of the URS, connecting the detector to the B
terminal, and noting the changes required in the IVD
as each dial of the URS except the first is varied through
its range. The variation should not exceed 0.5 ppm of
input of the IVD."
Appendix II
Calibration of the Volt Box with the URS
Since the resistance of any section of the volt box is
nominally equal to the total resistance of one or more
lower voltage sections, the box can be calibrated by
measuring these ratios systematically with the URS.
Figure 4 shows the connections for the test of the first
ratio. For the second ratio the leads to the volt box are
connected to posts 0, 2 and 3 with a low resistance link
(<1 mJ2) between 1 and 2. Connections for the other
ratios are shown in Table I. Since quite a number of 1/1
ratios must be measured, an error in the true midpoint
setting of the URS is compounded in the calculations. It
can be essentially eliminated, however, by measusing
each ratio a second time (before going on to the next),
with the leads aA and bB interchanged at the volt box

and taking the average of the two URS readings. Only

the lower dials, which have negligible corrections in a
reasonably good URS, will differ.
The voltage applied should produce rated cur
rent (1 mA) in the high-voltage sections of the voltbox.
The power dissipation in the URS is only 2 mW, so
that its heating is entirely negligible. To minimize errors
from thermal EMFs, the dials of the URS are adjusted
to give no change in galvanometer deflection when the
supply voltage is reversed (thus balancing to a "false
zero"). The galvanometer and URS control box are
convenient for this test, with the A and G posts of the
control box connected to the BA and BA' posts of the
URS, respectively, the control switch in the shorting
position, and F0 = 0.
From Fig. 4, at balance
Rx t Rt
Rt + Ra

where RA\ and Rm are the values of Ra and Rb at

balance, and the internal leads from the branch points
of the volt box to its terminals are included in R, and
R,. The resistances of the internal leads are negligible
(less than 1 mft) for this box. Ra and R>, are the re
sistances of leads aA and bB.
With the outer leads interchanged at the volt box,
at balance,
A\ + Rt
R* 4" Ra

Rbi
Ra2

Adding unity to each side of each equation and divid

ing the first equation by the second gives
Rx + R*
R, + Ra

u For convenience, the "nominal" value of M can be computed in

advance.
u Correction for larger changes can be applied when the URS is
used but are too complicated to be useful.

Rbi
Rm

(Rbi + Rai)Rai
Rai(Rbi + Rai)

Subtracting unity from each side of the equation,

noting that Rai + Rbi = Raj + Rbi, for a good URS, and

297 222

solving for Rt gives

fc-*L1+-^(1+*)JThe lead correction term is thus reduced to the second
order of smallness, and only the difference in the read
ings of the URS is significant.
Let K = K'(l+x) and R. = R.'(l+s), where the
primes refer to the nominal values and x and 5 are small
compared with unity. Then, since Rx' R,', we have, by
neglecting second-order terms
s+
In this volt box the potentiometer is always con
nected to the 0-1 section. If leakage currents are negligi
ble, the voltage-ratio of a range of the volt box con
taining m sections is
V = Z xt
where Nk is the ratio of the resistance of the kth section
of the box to the first section.
Let N = N'(l+n) and Nt = Nk'(l+xk) where the
primed quantities represent the nominal values and x
is small compared with unity.
From this

Z AY*t
^rr =

further adjustment. The direct voltage Vt is measurec

with external standards and then reversed. The pro
cedure is repeated and a voltage Vr is measured. If V.
is constant, the output EMF of TEi is the same when
each null is obtained, so that, directly from the defini
tion of 8,
Vt+ V,
F. =

(1 + )

In the deflection method the deflections D\ and Dtz<.

read with Si to the right and to the left, respectively
With the direct voltage reversed, deflections Z)jandP,
are read in the same way. For each reading the direct
voltage is held constant at the nominal value IV Th
Vi V
V V
= *(>, - Di) and
V.

= s(Dt - Ds:

where Vd and V, are now the direct voltages in the

direct and reversed position, respectively, that would
have to be applied in place of the alternating voltage to
produce the observed deflections Dt and Dt, and s is the
scale sensitivity factor." Then

where 2AD = D, - Di + D , - Dt = 2 (D - DAc) .

The alternating voltage required to produce the same
average deflection is, by the definition of ac-dc differ
ence,
Fd + V,

The calculations are carried out in accordance with

the schedule of Table I.
The results are useful only if the same current passes
through all of the resistors in use; i.e., if leakage and
corona currents from the resistors to the metal shield
of the box are negligible. The total leakage resistance is
easily measured by connecting a 1-kV direct voltage
source in series with a well-insulated dc picoammeter to
the 500-V post and the shield of the volt box. The leak
age and corona current indicated by the instrument
should be less than 5 nA.
Appendix III
Formulas for Using the DTVC
The ac-dc difference of a thermal voltage converter
such as the ac-dc converter of the DTVC (Fig. 5) is
defined as i = (F,c Vic)/Vic, where VK is the applied
alternating voltage required to produce the same out
put EMF of the thermoelement (TEi) as the average
of the two directions of direct voltage I7dc applied to this
same converter. In the null method of measuring an
alternating voltage Vu, the thermocouple shunt is
adjusted with the switch 5i to the right and the direct
voltage is adjusted with Si to the left until a null of the
detector is obtained in both switch positions without

V.e =

(1 + ) = F(l + + sAD),

neglecting second-order terms.

The scale factor 5 is readily determined by momen
tarily shunting the 0.5 volt range, consisting of the
heater of TEi, with a resistor R, = 100 KQ as shown is
Fig. 5. This decreases the current through TEi slightly
and is equivalent to decreasing the input voltage to that
converter by a proportional amount, = 5Xl0~4il'.
0.5)/Fr, where VT is the voltage range in use. From
this, s = p/D where D. is the resulting change (con
sidered positive) in galvanometer deflection.1' The scale
factor need not be known accurately, since only smil'
differences between the nominal and actual applied
voltages are involved. In practice it can be measured it
several points and graphed as a function of V/Vr- I'
should, of course, be rechecked at one point occaslor.
ally.
More complete formulas for the deflection method oi
using the DTVC can be developed to take into account

14 The change in voltage that is required to produce a I divMi"

deflection of the galvanometer.
" With the galvanometer now used, s is approximately 1X19^
/mm (i.e., 10 ppm/mm) at rated voltage on all ranges. A more tcstive galvanometer can be used if necessary.
298 223

small errors arising because the detector has a linear

instead of a quadratic scale, because the EMF vs.
heater current characteristic of the two thermoelements
may be different for the two directions of direct current,
and because of other small aberrations. However, if the
deflections and the dc reversal differences do not exceed
0.1 percent of the applied voltage, the resulting correc
tions are all less than 1 ppm.
For both the null and deflection methods, the direct
voltage must be constant to perhaps 0.05 percent during
each sequence of measurements, because the input
output characteristics of the two thermoelements are

not perfectly matched. Since the load on the dc source

changes when the selector switch is operated, the inter
nal resistance of this source must not exceed about 50
mfl/V or the source must be readjusted. It would be
desirable to add another set of resistors on the ac range
switch and another pole on the selector switch to keep
the load constant.
Author's Note: Other systems for accurately measuring dc volt
ages, with a different type of divider that is used also for the volt
box calibration, have been disclosed. See for example:
DC Parameter Measurement by the Rationmetric Method, Loebe Julie,
Data Systems Engineering March and June, 1964.
Making Measurements to standards Accuracy, Loebe Julie, Electronic
Industries, September 1964.

Reprinted from IEEE TRANSACTIONS

ON INSTRUMENTATION AND MEASUREMENT
Volume IM-14, Number 4, December, 1965

299-224

INSTRUMENT SOCIETY of AMERICA

prepkint#i Zllt Annual ISA Conference lad Exhibit October 24-27, 1966, New York Number 12.3-1-66
For mMM 19 mubbak wriU ISA. *J0 Wm I

A COMPARATOR FOR THERMAL AC-DC TRANSFER STANDARDS

by Raymond S. Turgel, Physicist
National Bureau of Standards
Galthersburg, Maryland
ABSTRACT
Thermal transfer standards play an Important
role In precision a-c measurements. They are
calibrated by intercomparlson with standards of
known ac-dc difference. A comparator Is described
that simplifies such routine calibrations. A
sequence of null balances In the measuring circuit
operates a simple analog computer which Indicates
the result of the Intercomparlson directly In
parts per million of ac-dc difference.
INTRODUCTION
Accurate a-c measurements In the audio fre
quency range depend on transfer standards which
relate these measurements to the basic d-c units
of current and voltage.
In recent years thermoelements and thermal
voltage converters have been widely used as
transfer standards especially at higher audio
frequencies. They are convenient to use and are
capable of accuracies of 100 ppm or better. ^'^'^
A thermoelement generally consists of a
heater in thermal contact with a thermocouple, but
usually electrically Insulated from it. The
thermocouple output is a function of the heating
effect of the heater current and to a first ap
proximation is the same for direct current as for
the equivalent rms alternating current. Actual
characteristics deviate from those of the simpli
fied model, and for certain geometric configura
tions limiting values for the departure from
ideal conditions can be calculated, based on the
thermoelectric properties of the materials used
in construction.^!*
Since absolute calibration of transfer
standards is not feasible, thermoelements must be
lntercompared to establish their inherent ac-dc
difference. By selecting thermoelements of ap
propriate design and corroborating the calculated
performance by experimental cross checks, NBS has
established a group of thermoelements of known
ac-dc difference with which other similar trans
fer standards can be compared. '
When high accuracy is required for the
intercomparlson a number of factors have to be
taken into account. The output from the most
commonly used thermoelements is obtained from a
single thermocouple junction and Is typically
10 mV. Multiple junction thermoelements with

higher outputs have been constructed in England

and the USSR , but are not generally available.
To achieve a resolution of a few ppm the system
must be capable of sensing a voltage change of
only a few nanovolts. This low voltage puts rather
severe demands on the detector and associated
circuitry which have to be designed to prevent
noise and stray thermal emfs from masking the
signal. A further complication is the temperature
coefficient of the thermoelement itself, which
often exceeds 0.1l/*C. Small changes in ambient
conditions will therefore produce noticeable
changes In output emf. The output is also affected
by asymmetry in the construction with the result
that the thermoelement response to d-c is dependent
on the sign of the heater current. In some cases
the observed change on reversal is considerably
larger than can be explained from the geometry of
the devices. Although manufacturers are making
efforts to reduce this reversal effect it cannot
be Ignored when making precision measurements.
Fluctuations in the heater current will
generally not produce identical changes in the
outputs of two thermoelements being compared.
This unequal response has to be considered in the
circuit design since even with the stable power
supplies now available some random fluctuations
occur.
The measuring system therefore has to com
pensate for:
1. Ambient temperature drifts
2. D-C reversal effects
3. Power supply fluctuations
Since in a properly designed system the ambi
ent temperature drifts are slow compared to the
time it takes to make a measurement, the drift
rate can be assumed to be constant. It is there
fore possible to reduce the error from temperature
changes by making a series of measurements ar
ranged symmetrically in time. This same sequence
can be used to average the output changes due to
d-c reversal. Thus a sequence of comparison
measurements using first a-c, then d-c forward,
d-c reverse, and again a-c, made at uniformly
spaced intervals will provide satisfactory
compensation.
The effect of power supply fluctuations can
be reduced by a ratio circuit' which has proven

Superior numbers refer to similarly-numbered references at the end of this paper.

300-1

satisfactory in this respect. An adaptation of

this circuit Is discussed In detail In this paper.

small, considerably less accuracy is required if

that difference can be measured directly instead of
being calculated from the two values. It Is there
fore desirable to arrange the circuit so that this
difference, or one equivalent to it, can be deter
mined experimentally. In practice an intermediate
step Is required, because the "standard" thermo
element cannot simultaneously be excited with a-c
and d-c to produce Egac and Es<jc.

THEORY
The ac-dc difference Is defined as
D - (I ac - I.dc )/I.dc when E ac - E dc
.
where I
circuit
rms a-c
heater.

(1)

The basic circuit is shown in Figure 1. The

outputs of the two thermoelements, Ej and Eg, are
connected In series with polarities so that the
sum of their emfs is connected across the voltage
divider R. The divider can therefore always be
adjusted to null the voltage across the detector,
and proportional changes in ET and Eg will not in
fluence the null position.

denotes the heater current, and E the open

thermocouple emf. The subscripts Indicate
or d-c excitation of the thermoelement
The relation
E - constant x I

(2)

Is postulated for the thermoelements, where n Is

approximately equal to 2, but in practice may vary
from 1.8 to 2.1.
Experimentally the comparator determines the
quantity DT - Dj where Dj is the ac-dc difference
of the "test" thermoelement, D the ac-dc differ
ence of the "standard" thermoelement. (Similar
ly, other quantities are identified by the sub
script T and S.)
As shown in the appendix
(DT - Ds) - (1/V) (ESac - Esdc)/Esdc

(3)

If Ef remains constant and Eg varies the cir

cuit will no longer be in balance. In a previous
method5 this unbalance was measured with a deflec
ting galvanometer. To calculate the ac-dc differ
ence from the deflection measurements the sensi
tivity of the galvanometer had to be determined
each time. In the present comparator a simpler
procedure is followed. The detector is used only
for null balance and therefore its sensitivity and
linearity do not have to be taken into account.
The ratio R is adjusted to an approximate balance
and final null balance is achieved by injecting an
emf E, .

This equation is valid only if the following

conditions are maintained during actual measure
ments:
1. The output voltage of the "test" thermo
element must be the same when using alternating
or direct current in the heater.
2. The heater currents in both "test" and
"standard" thermoelements must be identical. For
voltage transfer standards the equivalent condi
tion requires that the voltages applied to the
two thermal voltage converters (thermoelements in
series with their range resistors) are equal to
each other for both a-c and d-c heater currents.
The exponent n of equation (2) has to be
determined separately for the operating range of
the "standard" thermocouple. In practice n can
be found using the relation ^/E - n(AI/I) , which
is equivalent to equation (2), and measuring E/E
for a small change In current Al/I. An accuracy
of a few percent is sufficient for this measure
ment. Experience has shown that under normal
condition n does not change with time.
From equation (3) it follows that the intercomparison is reduced to the experimental deter
mination of the voltage ratio
(E Sac

If a-c and d-c currents are applied to both

thermoelement heaters in a t tme-symme t r ica 1 sequence
as discussed before, then four values of
will be
obtained, one for each null balance of the sequence.
From these, using appropriate calibration factors,
the ac-dc intercomparison value can be calculated.
E Aac - E Sac - E Tac
(4)
EAdc " ESdc " ETdc
where each of the E^ values here are the averages
of two readings.
With the condition given that ETac - ET<jc It
follows that
E Sac - E Sdc - E Aac - E Adc
combining equation
denominator of (3)
factor since It is
ference, we obtain

(5)

(3) and (5) and noting that the

can be considered a calibration
independent of the ac-dc dif
the intercomparison value

DT - Ds - (calibration factor) x (E^ - Z^) (6)

DESIGN CONSIDERATIONS

ESdc>/ESdc

under the specified conditions. Because the dif

ference of the two voltages In the numerator is
301-2

The design objective set for this project was

to produce a system that would reduce the effort
Involved In making routine lntercomparlsons of

thermoelements. An approach leading to full auto

mation could economically not be Justified, but
preliminary studies showed that significant im
provements over existing methods could be achieved
without greatly Increasing the complexity of the
equipment. Within these limitations a set of
specifications was formulated to incorporate the
features deemed most desirable. These included:
1. A read-out giving the lntercomparison
value in parts per million of ac-dc difference
without intermediate calculations;
2. The system to be self calibrating, with
the built-in reference to require infrequent, if
any, adjustment;
3. The system to be compatible with existing
external a-c and d-c power supplies which may not
be remotely programmable;
4. The system to be transportable in the
sense that it can be moved easily within the
laboratory.
To satisfy the first requirement a simple
analog computer Is incorporated Into the system.
Its function is to convert the ratio-circuit
balancing adjustment into an electrical signal,
store this signal, and then use it to compute the
final result. The second requirement, the self
calibrating feature, is part of the computer
circuit. As described in the next section, by
means of an Internal comparison a calibration
factor is obtained which is used to provide a
readout in the desired units.

only a single detector is used. The detector, how

ever, must then have a response time of not more
than about 1 second, and must have a fast recovery
from overload.
As shown in the appendix the error in the
difference measurement due to power supply fluc
tuations is reduced by a factor of 10 in a typical
case. Thus although the sequential system relies
on the short term stability of the power supplies,
many commercial power supplies are adequate.
PRINCIPLE OF OPERATION
Figure 2 Is a block diagram of the comparator
system showing the various functional components.
In normal operation, either alternating or direct
current is applied to the "test" and "standard"
thermoelements in a prearranged sequence deter
mined by the power selector switch. The outputs
of the two thermoelements are fed Into the ratio
circuit which has been set Initially to an approxi
mate balance.
To satisfy the condition ETac - ETjc, the
appropriate power supply Is adjusted until the
output of the "test" thermoelement is equal in
magnitude to a reference voltage of opposite
polarity. Equality of the two voltages Is deter
mined by connecting the null detector to the sum
ming point. The magnitude of the reference vol
tage is approximately equal to the output of the
thermoelement at the nominal operating point.

Battery operation has the additional advan

tage that circuits can be isolated to reduce,
spurious Interaction with one another and with
the thermoelements.

The null detector Is then switched to the

ratio circuit to measure the unbalance between the
"test" and "standard" thermoelement outputs. To
restore the ratio null balance a balancing voltage
is added to the output of the "standard" thermo
element. This balancing voltage is generated by
adjusting the analog computer. Four such voltage
adjustments are made, one for each step of the
AC, DC, DC-R, AC measuring sequence. The result
ing four voltages are retained by the computing
circuit and combined to give the final readout
value. The power selector switch is linked
mechanically to switching sections In the analog
computer to coordinate the operation. The cali
bration network provided determines a proportion
ality constant relating the balancing voltage to
the analog voltage. This constant is experimen
tally determined prior to the set of measurements
using a self-calibration circuit not Indicated on
the block diagram.

The preference for electronic detectors over

galvanometers is prompted also by their faster
response. To satisfy the requirement that the
output of the "test" thermoelement remain con
stant throughout the set of measurements the emf
has to be closely monitored. Thus essentially
two measurements are necessary, the magnitude of
the output Ej and the difference Eg - Ej. It
seemed preferable to obtain the two null balances
sequentially rather than simultaneously. Thua
only one adjustment at a time is required and

The calibrating procedure compares a fixed

known fraction of the output emf of the "standard"
thermoelement with the balancing voltage generated
when the analog computer is set to give a "full
scale" reading. The calibrating network is adjus
ted until the balancing voltage equals this frac
tion of the output emf. By proper choice of the
voltage divider, the output reading is then ex
pressed in parts per million of Isdc *n accordance
ulth equations (1) and (6). The voltage divider
which for this purpose has to be accurate only to

By using manual rather than servo adjust

ment of the power supplies, no particular diffi
culties arise with regard to compatibility.
Since the time constants of different thermo
elements vary, satisfactory servo control uould
in any case not be easily attainable.
To make the equipment movable within the
laboratory no galvanometers or other devices
needing firm supports are used. With minor
modifications, in fact, all control circuits can
operate from their own internal batteries, making
the system independent of the power line, except
for the power supplies to furnish the required
test currents.

302-3

the order of a tenth of a percent is formed by two range from -1.35 volts to +1.35 volts. Bach divider
stable resistors and should not require periodic
is associated with one of the four ratio balances
adjustments.
required In the measurement sequence. In turn each
adjustable voltage is connected to Ry in series with
Thus the comparison measurement reduces to a the calibration and range selecting networks. To
calibration adjustment, followed by four succes
avoid confusion, the particular divider that is
sive ratio balances each preceded by a rapid emf
connected at the moment is identified by a pilot
light controlled by the selector switch (not shown) .
check of the "test" thermoelement. After the
After the measuring sequence has been completed, the
last ratio balance, the analog computer provides
a readout of the relative ac-dc difference of the four voltages that are now "stored" in the dividers
are connected in series across the readout vacuumthermoelements being compared.
tube voltmeter (VTVM) , which has been calibrated to
indicate parts per million of ac-dc difference by
the calibration procedure. Switching circuits pro
CIRCUIT DESCRIPTION
vide correct polarities of the "stored" voltages so
that the sum of the d-c measurements Is subtracted
The simplified circuit diagram, Figure 3,
from the sum of a-c measurements.
shows the main components in more detail. The
outputs of the "test" and "standard" thermoele
Three full-scale ranges of 50 ppm, +250 ppm,
ments are connected in series with the ratio re
and +1250 ppm can be selected with a range switch.
sistor R3 which is a 500- ohm Kelvln-Varley divi
For routine lntercomparlsons , the 250 ppm range is
der with three switched decades followed by a
generally appropriate. The "ng" value (equation 3)
ten-turn voltage divider. This high resolution
is preset on a switch (see Figure 3) which modifies
permits the initial ratio adjustment to be made
the sensitivity of the VTVM so that the relative
easily so that the balancing voltages for the
meter indications are Inversely proportional to ns.
final null adjustment can be kept as small as
Using the Input impedance Rw of the VTVM as part of
possible. The range of the balancing adjustment
a voltage divider, the "n" switch adds resistance
is approximately 2 1/2 times the "full scale"
Rn, given by Rn/Rv - (ns/n ) - 1. In the particular
indication in ppn, but precision of readings is
design n 1.80 which represents therefore the
improved if the balancing voltage is kept small.
lowest value that can be selected.
The null detector Is connected to the ratio
Since the calibration network resistors as well
circuit through a three-position low thermal emf
as the VTVM are used in essentially the same con
switch with spring return to the center position.
nection during calibration as during actual measure
The third position of the switch Is used to moni
ments, their long term stability does not Influence
tor the output of the "test" thermoelement with
the result. It is required only that they remain
the null detector by comparing the reference
the same to 0.1 percent during the relatively short
voltage across R4 with the thermocouple emf.
The reference-voltage adjustment is also a Kelvln- period of the measurement. As already mentioned
the calibration essentially depends only on the
Varley circuit with two switched decades and a
ratio of R^ to Ro and on the ratio of the resistors
ten- turn divider used as a current divider.
In the range switch. Since these ratios have to be
Power is supplied from a mercury battery in an
accurate to only 0.1 percent the long term stability
external battery compartment.
requirement is easily satisfied.
When calibrating the system, the "standard"
thermocouple is disconnected from the ratio cir
PERFORMANCE
cuit so that its open circuit emf is applied
across the voltage divider formed by R2 and .
lntercomparlsons of thermoelements were car
The snail fraction of the emf across R} is nulled
ried out on a number of voltage and current ranges.
by the equal and opposite voltage drop produced
Table 1 shows the typical dispersion of readings
across R\ by a current derived from the analog
using two thermal voltage converters (5 mA thermo
and calibration circuits as described before.
elements and appropriate range resistors) at three
frequencies in the audio spectrum. The term "read
Parts of the circuit enclosed by the dotted
ing" here refers to the result of a sequence of
line on Figure 3 are contained in a carefully
four measurements as described. The standard devi
shielded and thermally lagged enclosure to mini
ation for a reading is calculated from the data
mize thermal emfs , leakage, and Induced currents.
shown.
Connection to the thermoelement outputs is made
through gold-plated copper binding pests.
The particular converters used in this test
agreed to 2 ppm or better, as established by pre
The power and control circuits are mounted
on a separate panel. They are linked mechanically vious calibrations. It will be seen that all
readings except some at 50 kHz and 300 volts fall
by the selector switch, but shielded from one
within the limit of 10 ppm. The lower precision
another by a separate enclosure surrounding the
at 50 kHz is probably caused by greater fluctua
power terminals and switch sections.
tions in the power supply.
The analog computing section consists of four
Using two sets of high-current thermoelements
adjustable voltage dividers each connected across
which had not been intercompared before, readings
two mercury batteries providing a voltage output
303-4

were obtained on 1, 3, 5, and 10 ampere ranges as

shown In Table 2. With a few exceptions the
readings fall into a 10 ppm Interval about the
average. The scatter is slightly larger than
with the voltage readings , which again is pro
bably a consequence of the lower stability of
the current source. All readings however are
well within the present accuracy limits of 200
ppm set for the ac-dc difference determinations
of thermal transfer standards at NBS.

Since K and K are very nearly equal we can

write K' - K(l - q) where q 1.
1
Expanding (K/k')" - (1 + q/n + . . . ) neglec
ting higher terms and substituting in (5A) we have
D-l
Designating the "test" thermoelement with the
subscript "T" and the "standard" with "S" , we have
the following conditions for null balance for
direct current:

CONCLUS IONS
Sdc h
A system has been designed for routine intercomparisons of thermal ac-dc transfer standards.
The design aims of reducing the effort involved
in making these measurements have been attained
without greatly increasing the complexity of the
equipment. Results obtained in tests agree with
those of other methods 5 to within 10 ppm on most
current and voltage ranges in the audio frequency
range.

dc
ETdc " Eref

OA}

E_,
Sdc -B.j
Adc
Tdc
where Ea Is the balancing voltage and r ratio of
voltage divider R (figure 1).
And similarly for alternating current:

APPENDIX
The ac-dc difference D is defined as
ac

dc

*Tac " Sac

(1A)

E E r
Tac
ref

(2A)

EcSac " E.Aac

Tac

(8Aj

Mc
when
E ac - E,dc

Combining (7A) and (8A) and dividing by E ^

where I Is the thermoelement heater current and
E is the thermoelement output voltage generated
by the thermocouple.
The following relations are postulated for
thermoelements :
dc

dc
(3A)

E Sac - E Sdc
dc

E Aac - EAdc
"Sdc

From (3A) and (6A)

EAac EAdc
ESdc

E ac - K'l"ac
vie Vs> " V5
where K and K are constants and the exponent n
is the range of 1.8 to 2.1.

(9A)
Vdc

From equations (2A) and (3A) we have

ns
ns
<L - "sV - *dc

Kl"dc - K'l"ac
and thus
I

= (k/k')1,i dc

(4A)

Since Eref - ETac = ETdc we obtain from (1A)

(5A)

I ac = I.dc (1 + D_)
V
Raising the equation to the ngcn power, and neglec
ting higher terms in D^ of the expansion, since

Substituting equation (4A) into (1A)

D - [(K/K')" - 1]

304-5

D_T 1 *, we have

The factor of 2 Is Introduced because two

pairs of Ea values are obtained In the measuring
sequence. Combining (12A) and (13A)

hi < - w - (1 + sdt - w
2EAdc " ^Aac - EAdc>/BSdc
and therefore (9A) can be written as
and therefore,
E
- E
T " S " h (~^c~}

(l0A)

R1/(R1 + Rj) - 2ns'DF

Substituting numerical values

In actual practice there Is some error In adjus

ting ET so that

Dp - 250 ppm, n^ - 1.8

E_Tac - E ref, (1
y)
v + "
then,
It can be shown that equation (10A) then becomes
T

"S

ESdc

ESdc

"T

R1/(R1 + Rj) - 900 ppm

A)

In a typical case the difference between ng and

Op might be 10 percent. E^dc/Esdc
designed to
have a maximum of 700 ppm (250 ppm range) . Even
If one assumes a relatively large error In the
adjustment of the output emf of the "test"
thermoelement, that Is

and Rj^ - 18 ohms, ^ - 20,000 ohms.

ACKNOWLEDGEMENTS
The author thanks Mr. D. R. Flach of the
Electrical Instruments Section of the National
Bureau of Standards for assistance In testing the
system and making the Intercomparison measurements.

y - 100 nV/5 mV - 20 ppm

the error term of (11A) will still be negligible.

REFERENCES

Error term - 1/2 x 20 x 10_6(1 + 700 x 10"6 - 0.9)

1. Hermach, F. L. , Thermal Converters as AC-DC

Transfer Standards for Current and Voltage
Measurements at Audio Frequencies, J. Res. NBS ,
Vol. 48, No. 2, pp. 121-138 (Feb. 1952).

- 1 ppm
In the calibration procedure the balancing
voltage E^ Is adjusted to be equal to a fraction
of the output voltage of the "standard" thermo
element (figure 3).

2. Hermach, F. L., AC-DC Transfer Instruments for

Current and Voltage Measurements, IRE Trans,
on Instrumentation, Vol. 1-7, No. 3 and 4,
pp. 235-240 (December 1958).

K - i-ri; Esdc

3. Williams, E-. S., Practical Aspects of the Use

of AC-DC Transfer Instruments. NBS Technical
Note 257.

<12A>

Ej( Is obtained by setting the voltage in the

analog computer to give a full scale reading on
the VTVM (250 ppm) and adjusting the calibration
network to obtain a null In the detector.
A particular value ng (- 1.8) Is chosen in
the calibration procedure for convenience in the
circuit. An arbitrary choice does not Influence
the validity of the calibration obtained, because
ng is adjusted independently for the actual
measurement. A different choice for ng would re
quire a corresponding change in the ratio of
vv
If D Is the full scale Indication of the
readout, then, rearranging equation (10A) we have

4. Wlddis, F. C, The Theory of Peltier and

Thomson-Effect Errors In Thermal AC-DC Trans
fer Devices, Institution of Electrical Engin
eers (London), Monograph No. 497A (January 1962)
5. Hermach, F. L. , and Williams, E. S. , Thermal
Converters for Audio Frequency Voltage Measurements, IEEE Trans, on Instrumentation and
Measurement (to be published in 1966).
6. Wilklns, F. J., Multifunction Thermal Conver
ters . The Institution of Electrical Engineers,
Proceedings, Vol. 112, No. 4, pp. 794-805,
(April 1965*.
7.

2ns'DF " <EAac - EAdc>/ESdc

<13A>

305-6

RoEhdes tvenskaya , T. B., Electrical Comparators

for Precise Measurements of Current, Voltage,
and Power, p(lll) Translation JPRS:29,373
(TT:65-30626) , U. S. Dept. of Commerce, Clear
inghouse for Federal Scientific and Technical
Information.

Table 1
Relative AC-DC Difference of Fairs of Thermal Voltage Converters
(parts per million)
50 Hz
20V
-1
+3
0
+3
+2

Frequency
Voltage Range
DT" DS

10 kHz
100V
+2. -1.
0, -2,
-2, 0,
-2, -2,
-1, 0,

-2.
+1.
o,
+1.
-4.

50 kHz
300V
-10
-12
-12
- 4
- 6

+1
0
+2
+2
+4

Mean

1.4

-0.15

S tandard
Deviation

1.7

1.8

No. of
Readings

- 8.8
6.0

20

Table 2
Relative AC-DC Difference of Fairs of High Current Thermoelements
(parts per million)
Frequency
Current Range
DT " DS
(ppm)

1 A
0
-1
-4
+8
-4

Mean

-3.2

10 kHz
3 A
5 A
-15
-3
+1
-30
-12
-1
- 6
+5
-1
-17
0.2

-16

50 kHz
10 A
+ 2
- 6
+ 4
- 5
-11

1 A
0
+6
-3
+2
0

3 A
+ 8
+10
+ 8
+14
+20

5 A
- 7
- 6
-11
- 8
- 4

10 A
-2
+4
0
-1
-2

- 3.2

1.0

12

- 7.2

-0.2

Standard
Deviation

No. of
Readings

306-7

AC OR DC

RATIOWWW
ADJUSTMENT

"TEST"
THERMO
ELEMENT
NULL
DETECTOR

"STANDARD"
THERMO
ELEMENT

BALANCING VOLTAGE

FIG. I

BASIC CIRCUIT OF
THERMOELEMENT COMPARATOR

307-8

AC
SUPPLY

DC
SUPPLY

POWER
SELECTOR

"TEST"
THERMOELEMENT

"STANDARD
THERMOELEMENT

BALANCING
VOLTAGE

RATIO
CIRCUIT

CALIBRATION
NETWORK

REFERENCE
VOLTAGE

li
NULL
DETECTOR

ANALOG
COMPUTER

READ OUT

FIG. 2

THERMOELEMENT COMPARATOR

308-9

NOTE DURING CALIBRATION SWITCHES

ARE IN POSITION "C".
FIG. 3

SIMPLIFIED CIRCUIT DIAGRAM

OF
THERMOELEMENT COMPARATOR

n ADJUSTMENT
VTVM
READ OUT

309-10

PPTh,PPTNTT/^\INSTRUMENT SOCIETY of AMERICA Preprint

-L XVJi A XVXi. 1 J. (i^yj) 20th Annual ISA Conference and Exhibit October 4-7, 1965. Los Angeles Number 14.2-3-65
in thti paetr rttll with lh authori

All rigKlM r, ,. ., ,,f . for prrmiuuM fe tutbhtk wnu ISA 530 Wm

CALIBRATION OF PEAK A-C TO D-C COMPARATORS*

By:

Donald Flach, Physicist

Louis A. Marzetta, Electronic Engineer
National Bureau of Standards
Washington, D. C.

ABSTRACT
Recent developments in commercial electronic
power supplies have made practical the use of
peak a-c to d-c comparison for measuring a-c
voltage.
This paper describes two methods
that have been developed at NBS for testing peak
a-c to d-c comparators at the 10 and 20 volt
level over a frequency range of 50 Hz to 4800 Hz.
These two methods yield results which agree well
within 50 parts per million (ppm) .

PEAK A-C TO D-C COMPARATORS

A commercially available peak a-c to d-c
comparator (see Fig. 1) uses a pair of precision
summing resistors, one of which is connected to
the a-c, the other to the d-c voltage, for the
ba6ic a-c to d-c comparison. * *^ The summing re
sistors are followed by diodes which clip off all
but a small portion of the composite wave.
Following the diodes 1b a chopper which provides
a ground reference, a XI, X10 amplifier, an out
put limiter, and an oscilloscope which is
connected to the output as a null detector. This
instrument has a frequency range of 0-20 kHz and
is capable of measuring up to 20 volts peak.
Another type of peak a-c to d-c comparator,
which operates on a time-gate principle^, was
constructed at NBS. This Instrument was designed
to operate over a frequency range from 50 to
2400 Hz, and with a 10 volt peak input. The
basic idea in the development of a new peak com
parator was to search for a circuit that would
offer a simple, direct comparison of the peak of
a sine wave with a precision d-c source, with a
common ground connection for both sources. It
was expected that the desired simplicity would
place some restrictions on operational flexi
bility, but the purpose is a laboratory standard
and not an instrument capable of operating under
adverse conditions. In the design of the instru
ment, the use of precision summing resistors was
avoided in order to obtain an independent method
for calibrating the summing type comparators.
This system uses a pair of synchronized relays
to slice out a section of the sine wave at its
peak (see Fig. 2). Figure 3 shows the basic

operating principle of the time-gate comparator.

For the major portion of the sine wave, the re
lays are connected to the d-c voltage. The
relays are connected to the a-c voltage for a
brief interval as this a-c voltage passes through
its peak value. They are then switched back to
the d-c voltage for the remainder of the cycle.
The comparator is balanced when the peak of the
a-c signal is aligned with the d-c voltage line.
In operation, the input a-c voltage is clipped
and Is used to initiate a trigger circuit whose
function is to generate synchronizing pulses
occurring at the zero crossing of the sine wave.
Since electro-mechanical relays cannot function at
the higher speeds, six binary stages have been
incorporated for the purpose of dividing the input
frequency. A seven position switch allows the
selection of a 1/1 to 1/64 frequency division,
chosen to give a relay operating speed of about 50
operations per second. Thus, for example, a 2400
Hz input would be divided by a factor of 64 aC the
last binary. This lower frequency can be used to
drive the relays. For this purpose a power ampli
fier is located after the dividers. By means of
phase-delay adjusting resistors, the closing time
of the relays can be set to provide a time gate
situated at the peak of the sine wave. For the
example stated above, the gate would be located at
the peak of every 64th cycle of the 2400 Hz input
potential. No information is lost by this proce
dure; however, the oscilloscope trace must have
sufficient brilliance to present a usable expanded
pattern In a low duty cycle.
The 10-volt d-c supply located between the relay
output and amplifier input reduces the 10-volt
average output from the relays to a level which
will not overload the amplifier. The putpose of
the amplifier Is to provide the proper transfer
circuit between the output of signal relays and
the oscilloscope input. A first requirement in
considering a choice of design Is the degree of
amplification Itself. For oscilloscopes having a
vertical deflection sensitivity between 0.01 and
0.1 volts per cm, the instrument amplifier should
have an amplification factor between 10 and 100.
An Input impedance of at least 100,000 ohms was
chosen to reduce loading errors to less than 10
ppm. The amplifier was designed with a bandwidth

Superior numbers refer to similarly-numbered references at the end of this paper.

Contribution of th? National Bureau of Standards, not subject to copyright.
310- 1

In excess of 1 MHz so that the decrease In

response at the highest signal frequency was less
than 10 ppm.
All possible storage elements such as capaci
tors were deleted in the Interest of quick
recovery. Reactive feedback components were
avoided in order to achieve an ideal flat fre
quency response with the desirable upper frequency
roll-off characteristic. Amplification stability
and quiescent d-c output level stability were
compromised in order to achieve a flat frequency
response. A simple control is provided for the
adjustment of output d-c level.

CALIBRATION OF PEAK COMPARATORS WITH DIFFERENTIAL

THERMOCOUPLE VOLTAGE COMPARATOR
The Differential Thermocouple Voltage
Comparator^ (DTVC) may be used for simultaneous
comparison of rms a-c and d-c voltages. If the
wave form is known, it is possible to use the
instrument, in combination with a suitable trans
former, for checking peak a-c to d-c comparators.

the positive and negative peak) corresponding

deflections D3 and D4 are read in the same way.*
Then, as shown in appendix 3 of reference 3,
VQ(1 + sAD + 6)

where Vac is the rms a-c voltage and Vq is the

nominal d-c voltage, s is the sensitivity factor
(in proportional parts per division) and 6 is the
ac-dc difference of the AV converter of the DTVC,
(defined as the proportional difference between
the a-c and average d-c voltages required to
produce the same output of this converter). AD
is equal to l/2(Dac - Dj,,) where Dac Is the sum
of the two galvanometer deflections with a-c
voltage applied to the AV converter, and D^c is
the sum of the two galvanometer deflections with
d-c voltage applied to both converters.
If the ratio of the peak to the rms values of
the waveform is /2 (the same as for a sine wave),
and the voltage ratio of the transformer In Fig. 5
is N, the peak value of the a-c voltage applied
to the peak comparator is
V p - /2NV ac

Figure 4 is a schematic of the DTVC. This

instrument has two nearly identical voltage con
verters, as shown in Fig. 4 and in the simplified
diagram in Fig. 5, each consisting of a chain of
redistors in series with the heater of a
thermoelement. The two thermoelement outputs are
permanently connected in opposition and to a
galvanometer, G. Voltage ranges from 0.5 to 500
volts are available with a frequency response of
20 Hz to 50 kHz. The use of separate resistors
to reduce reactance errors for each range,
shielding and isolation of a-c and d-c converters,
and rf filters on the thermoelements, combine to
reduce ac-dc transfer errors to 10 ppm or less for
this Instrument.
Figure 5 is a schematic diagram of the circuit
used to test peak comparators with the DTVC. The
ac-dc difference of a peak a-c to d-c comparator
is evaluated by adjusting the d-c voltage to
balance the peak comparator, and then measuring
the difference between the rms a-c and d-c
voltages with the DTVC.
A deflection method is used so that it is
necessary to know the scale factor of the galva
nometer. A built-in 0.05 percent sensitivity
check enables the operator to calibrate the
galvanometer scale on all but the lowest range.
A galvanometer shunt is used to adjust the scale
factor to a value convenient for computation.
With the applied voltages adjusted to balance
the peak comparator under test, the AV converter
(shewn at the left in Fig. 5) is switched to the
d-c and a-c voltages respectively, and the
corresponding galvanometer deflections Di and D2
are read. With the d-c voltage reversed, and the
peak comparator again balanced, (to eliminate d-c
reversal errors in the DTVC and to measure both
315-411 O - 68 - 21

(I)

(2)

where Vac is the output rms voltage of the a-c

supply. Let N - (1//2)(1 + n) , where n Is the
small ratio correction factor, and define the
ac-dc difference of the peak comparator as
V - V
5 - -E
S
(3)
P
v
Combining equations (2) and (3) results in:
+ n

CO

Then by changing eq (1) to the form of an ac-dc

difference expression, substituting it In eq (4)
and neglecting second order terms
6 + n + SAD.

(5)

Distortion in the a-c source may introduce an

error into the peak measurement. The harmonics in
the output of the power-suppy- trans former combi
nation were measured with a commercial wave
*A modification of the above procedure, which was
used in most of the tests described in the report,
is to follow the above procedure for the first two
steps. On step (3) the peak comparator is not re
balanced. Step (4) remains the same. A second
set of four readings is taken to measure the peak
of the opposite polarity by reversing the d-c
voltage. Thus eight readings are required to cali
brate the test instrument, but this method is
more convenient than the four-step procedure.

311-2

analyzer preceded by a twln-T network tuned to

reduce the fundamental by at least 60 dB. The
results are glvn In table 2. Supplementary
measurements with a Wlen bridge constructed of
wire-wound resistors and air capacitors of
negligible voltage coefficient verified that the
twin-T network did not Introduce appreciable
distortion. The effect of even harmonic distor
tion on the calibration is virtually eliminated
by measuring both the positive and negative peaks
and taking the average. Odd harmonics introduce
an error which depends (among other things) on
the phase relationship of the distorting harmonic
to the fundamental. At the present time there Is
no known method of measuring this phase relation
ship; therefore, in the worst case there is an un
certainty in the results equal to the arithmetic
sum of the odd harmonics.
A requirement for peak a-c to d-c comparison
is the need for high cycle-to-cycle stability in
the a-c source. Typically the a-c source used in
these tests varied from cycle to cycle by less
than 30 ppm, with occasional extremes of 100 ppm.
With a little practice the operator could visually
reduce the effect of cycle- to- cycle variations
and Ignore the occasional extremes.
By using the a-c supply to trigger the
oscilloscope, a presentation of a single peak is
obtained. With an oscilloscope sensitivity of
lOmV/cm, changes as small as 10 ppm between the
a-c and d-c voltages may be detected when the
peak comparator has a voltage gain of 10, and is
operating at 10 volts peak.

1NTERC0MPARIS0N OF PEAK COMPARATORS

Figure 6 is a block diagram of the lntercomparlson circuit. For these tests the a-c supply
was kept fixed at a value of 10 volts peak. This
a-c amplitude was compared with a d-c voltage
which was adjusted, In turn, to balance each
comparator by means of an oscilloscope. The
voltage settings of the precision d-c supply were
recorded in each determination. This difference
between the d-c voltage settings is equal to
5TG " 6s> where 6jq is the ac-dc difference of
the time-gate comparator, and 6s is the ac-dc
difference of the summing-type comparator. This
may be shown as follows:
UTG

is aa

(6)

V is the nominal d-c voltage.

A rearrangement of the terms in equation (6) and
(7) results in:
i92
6TG

*o

VG " *S-

Assuming that the a-c voltages are equal, and

subtracting the second equation from Che first
gives:
V o(S) - V offG)
'tg " V

(8)

This type of test requires highly stable a-c

and d-c supplies. The d-c supply stability has
been measured and is 1 ppm/10 minutes, the a-c
supply is stable to better than 20 ppm/10 minutes.
If the ac-dc difference of one peak comparator
is measured with an rms standard, the comparator
may then be used to calibrate another peak
comparator without additional waveform error even
If a different a-c source is used. However, a
waveform error may, or course, result if the peak
comparator is used for rms voltage measurements
with a source of different harmonic content.

TEST RESULTS
Table 3 shows the results of the ac-dc
difference tests on the two types of comparators
using the circuits shown in Fig. 5. This) table
also shows the results of the direct intercomparison tests using the circuit shown in Fig. 6.
The relative ac-dc differences, 8g - 8t(J,
compiled from the two methods agree to well
within 20 ppm.
The positive sign in the 5s - &TG differences
compiled from the direct lntercompariaon tests
indicates a higher d-c voltage with the NBS
comparator than with the summing comparator, for
the same a-c source value. Note that all the
differences are In one direction. This Indicates
a small but real offset in one of the comparators.
Some portion of this offset would manifest itself
as a systematic error in the case of other deter
minations Involving the comparator.
The results are shown to the nearest ppm for
the comparison of the different methods. It is
not to be Implied that the values of ac-dc
differences are correct to one ppm.

(7)
where Vac is the a-c voltage when the comparator
is balanced,
V
is the d-c voltage when the time-gate
comparator is balanced,
V0(s)
Cne d-c voltage when the summing
type comparator is balanced,
312-3

The results given in table 3 are the average

of a number of determinations made over a period
of months. The standard deviations of the
individual measurements on which the averages
were based were leas than IS ppm. However,
earlier measurements, (which were made while the
calibration system was being investigated), and

other arrangements of connections showed

occasional unexplained shifts of up to 30 ppm.
The estimated overall uncertainty In the calibra
tion of a peak comparator, including allowances
for systematic errors and for three times the
standard deviation of an average of four determi
nations, is less than 50 ppm with the time-gate
comparator as the standard. It is less than 50
ppm from 400 through 2400 Hz with the DTVC as the
standard, but must be Increased to 100 ppm at 50
and 4800 Hz because of the waveform uncertainty
already discussed. The agreement to 30 ppm or
better between the two methods (lines 2 and 4 in
table 3) indicates that the actual waveform error
Is very probably considerably less than this.
Because of the freedom from waveform errors,
peak comparators are now tested by NBS with the
time-gate comparator at 10 volts, at 50, 400,
1000, and 2400 Hz, with a conservatively stated
accuracy of 100 ppm. Testing of peak comparators
at 4800 Hz (outside the range of the time-gate
comparator) will continue to be done with the
DTVC.

ACKNOWLEDGMENTS
The authors gratefully acknowledge the help of
J. E. Griffin who made the first calibrations of
the commercial peak comparator with the DTVC, and
C. J. Saunders who designed and evaluated the
twin-T networks used with the wave analyzer.

REFERENCES
(1) Rlchman, Peter L. , A New Absolute A-C Voltage
Standard. Session IEEE Record of Convention,
Paper 13, March 1963.
(2) Rlchman, Peter L. , A New Peak-to-DC
Comparator for Audio Frequencies, IRE
Transactions on Instrumentation, Vol. 1-11,
p. 115, December 1962.
(3) Smith, W. E. and Clothier, W. X.,
Determination of the D.C./A.C. Transfer Error
of an Electrostatic Voltmeter, Proceedings of
the Institution of Electrical Engineers, Vol.
101-11, pp. 465-469, October 1954.
(4) Hermach, F. L. , Griffin, J. E., and Williams,
E. S., A System for Accurate D-C and A-C
Voltage Measurements, Session 63 of IEEE
Record of Convention, March 1965.

Table 1
AC-DC Difference of DTVC
AC-DC Difference ppm
Frequency Hz
1 kHz
20 kHz

Range

-30
-10
+ 10
+10
+10
0
0
+10
+10
0
0
+10
+30

Table 2
Waveform Distortion of
Power Supply-Transformer Combination'Distortion ppm
Frequency Hz
Harmonic
400
1000
50
2400
4800
2
B5
3
75
4
5
20
6
7
S
8
9
5
Sum of odd
harmonics 100
Note 1.

35
5

30
10
--

SO
25
5

65
30
10
10
25
10
10

--

10

25

50

1000 fl across primary of transformer and

5000O load on secondary.
10 volts rms applied to transformer.

Note 2. Dashes Indicate measured value less than

5 ppm.
Table 3
Intercomparlson by Two Methods*
400 1000
50
6TC measured with DTVC
5S measured with DTVC**
Difference (6g - 6TC)
6S - 6TC from direct
intercomparlson

+32
+50
+18

- 4
+14
+18

- 8
+24
+32

+ 3
+42
+39

+21

+11

+24

+31

easurements were made at 10 volts peak*

6g measured at 4800 Hz was +41 ppm.

313-4

2400

A-C
INPUT
o-

SUMMATION POINT

10 ML
AAA

I0HJL
AAA

I
0-C
INPUT

PRECISION
LINITER

OSCILLOSCOPE DBPLAY
Figure I. Sunning Type Peak Comparator

&

OSCILLOSCOPE

OC.
"11

II

FREQUENCY DIVIDERS AND

RELAY

DRIVERS

Figure 2. Block Diagram of Time Gate Peak Comparator

314-5

OSCILLOSCOPE

Figure 3,

Time Gate Comparison of Peak A-C and D-C Voltages

315-6

Figure 4. Differential Thermocouple Voltage Comparator

Note: All range resistors not in use are
connec ted to the range shields through
additional switch contacts as shown.
The shields are shown as dotted lines.

316-7

-yjr AUTOTRANSFORMER

OSCILLOSCOPE
,

/
9

0
n
t
D C POWER SUPPLY

PEAK
COMPARATOR

_DTC
A C. POWER SUPPLY
DV

CALV.
NETWORK
AC THERMOCOUPLE

DC THERMOCOUPLE

figure 5. Test Setup for Calibrating Peak Comparators

A-C
SUPPLY

NBS
PEAK
T
COMPARATOR

D-C
SUPPLY
'

Figure 6,

COMMERCIAL
PEAK
COMPARATOR

Direct Intercomparison Test

317-8

OSCILLOSCOPE

A Differential Thermocouple Voltmeter

J. E. GRIFFIN
ASSOCIATE MEMBER AIEE
Summary : An a-c voltmeter has been
developed to indicate directly the percentage
difference between an unknown voltage and
the settings of the instrument dials. It is
accurate to 0.05% from 5 cps (cycles per
second) to 10 kc within the voltage range
from 1 to 700 volts. Two 10-ma (milliampere) thermoelements are used; one in
series with an a-c decade resistor, and the
other energized from a constant-voltage d-c
source, a Zener diode. A built-in galva
nometer, calibrated in % of input voltage,
indicates the difference between the two
thermocouple-output emfs (electromotive
forces), and has a resolution of more than
0.01%. The instrument was designed for
rapidly calibrating other voltmeters, but it
can be used also for a-c- d-c difference
measurementsfrequency response tests
to 0.02%.

F. L. HERMACH
MEMBER AIEE
HIGH ACCURACY in measurements
is generally attained at the cost of
increased time and effort, as with potenti
ometer techniques, or by means of rather
complicated equipment such as electronic
standards using feedback and similar
techniques.
There is always the necessity for com
bining high accuracy with the simplicity
which makes for reliability. This is par
ticularly true in instrument calibrations in
which many measurements of high ac
curacy are often required. Electrodynamic and electrostatic instruments of the
higher accuracy classes suffer from
limited frequency or voltage ranges, and

318-1

thermocouple instruments from poor sta

bility.1
The undesirable characteristics of
thermocouple instruments are best over
come by a-c-d-c transfer techniques
With the thermoelement connected to the
a-c source, the thermocouple-output emi
is observed. The thermoelement is ther.
connected to a d-c source, which is ad
justed to give same thermocouple output
The d-c source is next measured with i
d-c potentiometer of high accuracy.
Ordinarily, this a-c to d-c transfer nrasi
Papr 62-810, recommended by the AIEB Ittdiouing and Integrating Instruments Committee sad
approved by the AIEB Technical Operations De
partment for presentation at the AIEE Mid&t
Eastern District Meeting, Wilmington, Del.
May 7-0, 19G2. Manuscript submitted Jamarr
30, 1962; made available for printing March i
1962.
J. B. Gkippin and P. L. HhkmaCH are with xht
National Bureau of Standards. Washington. D C
The authors gratefully acknowledge the help
T. W. Cannon, who constructed the instrn&eg:
and J. Grant, who assisted with the test.

0.01% stability, which greatly simplifies

the standardizing circuits, and makes
possible a lighter, more convenient, and
much more compact instrument. The
present instrument also differs from
Rump's in a number of other respects,
and is useful to much higher frequencies.
Today, differentially connected thermo
elements are used in a commercially
available 0.02% a-c-d-c transfervoltmeter
but with uncalibrated resistors, so that
a d-c volt box and potentiometer are re
quired for the measurements. Recently,
a calibrated 0.05% a-c voltage source,
using differentially connected thermo
elements in a feedback circuit with a
Zener diode as the reference, was de
veloped, independently of the present
instrument.4

Fig. 1. Simplified circuit dkjna of DTVM

be made for each measurement if better
than 0.1% accuracy is required.'
This paper describes an audio-fre
quency a-c thermocouple voltmeter in
which the transfer need be made only
about twice a day to maintain a stability
of 0.01%. As a result, a-c voltage meas
urements may be made rapidly, conven
iently, and accurately.
Basic Principle
The basic principle of the differential
thermocouple voltmeter (DTVM) may
be seen in Fig. 1. A 10-ma thermoele
ment Tt of negligible a-c-d-c difference
is connected in series with a decade re
sistor Rm. A second thermoelement 7,
having the same output emf at 10 ma, is
connected to a stable d-c voltage source
Zener diode energized by a rectified and
filtered a-c voltage. The outputs of the
two thermoelements are connected to
gether in opposition through a galvanom
eter ft. Thus, when the galvanometer
is balanced by adjusting the decade re
sistor RH, the two heater currents are
equal and the input voltage is a function of
the setting of Rm, which is marked directly
in volts. In addition, the scale of the gal
vanometer is marked to indicate directly
the percentage difference between the
actual input voltage and the voltage
setting of Rm for slight differences between
them. This is particularly convenient
for calibrating other voltmeters.
Several preliminary calibration adjust
ments are necessary to insure that the
resistances of the thermoelements are
correct and that the thermocouple out
puts are equal with equal heater currents.
The resistances of the thermoelement
heaters are checked in the internal bridge
circuit with the galvanometer in position
shown as ft in Fig. 1. First, with the
bridge circuit as shown in Fig. 1, R,
is adjusted to balance the galvanometer.

Then, with Ta and Ra switched in place

of R,, R is adjusted to balance the
galvanometer. To check equality of the
thermocouple-output emfs, the galvanom
eter is switched to position ft-with T
and Ra undisturbed. Also, R which
shunts the 12-ohm thermocouple of the
thermoelement having the highest output
emf, is adjusted to balance the gal
vanometer. These calibration adjust
ments do not require reference to external
standards; are easily and rapidly carried
out; and ordinarily need be made only
twice a day to insure a stability of 0.01%.
An important but infrequent calibra
tion adjustment is setting the reference
current. This is done by connecting the
input of the instrument to a known d-c
1-volt source, with the circuit as in Fig. 1,
and adjusting Rf to balance the galva
nometer. Because of the stability of the
Zener diode which supplies the reference
current, this 1-volt calibration need be
made only semiannually or whenever the
instrument receives a d-c calibration.
To improve stability, both thermo
elements are mounted in a thermally
lagged enclosure. They have closely
matched temperature coefficients of out
put emf and have heaters of low-tempera
ture coefficients of resistance. Highquality a-c resistors having low reactance
are used. Thus, without applying cor
rections, measurements can be made to
0.05% or better from 1 to 700 volts at
frequencies up to 10 kc. High accuracy
is possible because of the a-c to d-c trans
fer, which reduces the effects of drifts and
other changes in the thermoelements.
This DTVM is a modification of an
instrument developed by W. Rump.' In
Rump's instrument, a single-point po
tentiometer with a standard cell was used
to adjust the d-c current through the
second thermoelement. Zener diodes
have now been developed with better than
319-2

Description
Figs. 2 and 3 show a complete circuit
diagram and a view of the DTVM. The
heart of the instrument is, of course, the
differential thermal converter, which
consists of two 10-ma thermoelements,
each having approximately 11 -mv (milli
volt) output, mounted in the thermally
insulated enclosure shown in Fig. 4.
Each thermoelement is contained in an
evacuated glass bulb, and has an electri
cally insulating bead between its heater
and the thermocouple hot junction. Each
has a nickel-chromium heater, which, be
cause of its low thermoelectric effects, has
d-c reversal and a-c-d-c transfer errors less
than 0.01%. The voltage thermoelement
r, connected in series with the decade
resistor, Rm in Fig. 1, is energized by the
voltage to be measured. The reference
thermoelement Tt is supplied by the
internally mounted commercially avail
able temperature-compensated Zener
diodeenergized from a 115-volt 60cycle line-which has a 5.8-volt output
with an 11 -ma load.
The switch connecting the various
calibration and operating circuits was
selected because of its low contact resist
ance and low thermal emfs. Connections
to the switch introduced negligible lead
resistances into the circuits.
A resistor, part of Rf, in the Zenerdiode circuit provides approximately
0.5% adjustment of the reference current.
It is adjusted only in the external 1-volt
check with the dials set in the 1-volt
position. In addition to supplying a
closely regulated current to the reference
thermoelement T, during voltage measure
ments, the Zener diode supplies the bridge
circuit for checking and adjusting the
resistance of the thermoelement heaters.
One 100-ohm resistor R, and two 1,000ohm resistors, Rt and R make up the

VOLTS

Fig. 2. Complete circuit due-iof DTVM

R.&RlManganin, 67.5 ctma, set
text
R.25 ohms
RjftR.Manganin, 1,000 ohm
RiManganin, 320 ohrs vet
text
R,Margamn, 29 ohms
Rk20lcilohms
R,Manganin, 1O0 ohms
T.StlbThermoelement, 10 na, 11
ZZener diode, 5.5 volts 11 w
VRectifier voltmeter, 1 volt, 100
ohms
Switch Position
1. Adjust heater resistor Rfc
2. Adjust heater resistor R,
3. Adjust thermocouple outru
resistor R.
4. Rectifier i
5. Volts

reference resistor and ratio arms of the

bridge. These three resistors are adjusted
to better than 0.01%.
The thermoelement heater circuit,
consisting of Ta and R*, provides the basic
1-volt range of the instrument. The
heater resistance of each thermoelement
is approximately 33 ohms. A 66-ohm a-c
resistor and a 1.5-ohm adjustable bifilar
resistor are connected in series with each
heater to provide a total heater resistance
of 100 ohms with a 1.5% adjustment. An
a-c 5-dial decade resistor, connected in
series with voltage thermoelement T
provides for voltage measurements from
1 volt to 700 volts in 0.01-volt steps. Re
sistors in the 0.01% accuracy class were
used in the X100, X10, and XI dials, while
resistors in the 0.05% accuracy class were
used in the X0.1 and X0.01 dials. For
preliminary setting of voltage, a 1-volt
100-ohm full-wave rectifier instrument,
V in Fig. 2, is substituted for the thermo
element.
The XI dial is marked from 1 to 11
rather than 0 to 10, since the 1-volt heater
of the voltage thermoelement is neces
sarily in series with the decade resistor.

This requires that a voltage of 100 volts,

for example, be set as the sum of 90 volts
on the X10 dial and 10 volts on the XI
dial. Although this arrangement may
seem awkward at first, the correct pro
cedure for setting the dials is quickly
grasped. A much more complicated
circuit would have been necessary to pro
vide for steps of 0 to 10 on the XI dial.
An adjustable resistor in the galvanom
eter circuit. Fig. 5, is used to set the galva
nometer sensitivity. This adjustment
is checked occasionally after the other
checks are made, with TQ and Rc con
nected in place of R, in Fig. 1. Con
necting the 20-kilohm resistor Rk causes
the two heater currents to differ by 0.50%,
resulting in a 25-mm (millimeter) change
in galvanometer deflection if the adjust
ment is correct.
A thermocouple instrument is a squarelaw device. However, to a first-order
approximation, the galvanometer deflec
tion is directly proportional to small
changes in DTVM input voltage. This is
possible since the voltage across the galva
nometer is only a small part of the
thermocouple output voltage. Let E,
320-3

and E, be the outputs of the reference ar-d

voltage thermoelements, respectively, cor
responding to input currents of I, and /
Let A be the change in thermocouple
output caused by change AI in input cur
rent through the voltage thermocouple,
corresponding to an input voltage change
of AV. Let k be the proportiotiaJit.
constant, which, by calibration, is the
same for both thermoelements.
Let
A/ = /,-/,
Then
HE = ,-, = */,'-*/, = 2*/,AJ+ tAT
Since M is never greater than 0.5% of /,
in the normal use of this instrument,
AE ~2E,M/I, 2E,AK/ V
Maximum error from this approximator
is less than 0.002% of the input voltageSensitivity of the galvanometer is
adjusted to cover a change in fafc"
voltage &V/V of 0.5%, or 0.02% pr
division for the full-scale deflectkc 9
25 divisions of the center rero gal1^

Fig. 3. Differential
thermocouple volt-

nometer used in the instrument. Since

the output emf E, is 1 1 mv, the scale fac
tor for the galvanometer, by the prior
equation, must be 4.4 itv (microvolts) per
division. Fig. 5 is a diagram of the
galvanometer circuit, showing provisions
for preliminary balances and the resistor
for setting the sensitivity. The entire
circuit of Fig. 5 is represented as ft,
Gi, and G in Figs. 1 and 2.
Temperature Compensation
The differential-thermocouple principle
affords a means of temperature compensa
tion . To accomplish this, the two thermo
elements must have nearly equal tem
perature coefficients of emf and must be
maintained at the same temperature.
Several thermoelements were tested,
grouped into pairs, and the combined
temperature coefficient of each pair was
measured. The net temperature coeffi
cient of the pair used in the DTVM is 0. 1
(iv/C (microvolts per degrees centigrade),

which causes an error in voltage measure

ment of only 0.0005%/C.
Temperature equality is maintained
by means of a 1/8-in. copper temperaturelag box, approximately 2.5 by 3 by 4 in.,
shown in Fig. 4 The two thermo
elements are mounted in holes in a small
copper block and held in place with sili
cone grease. The block is placed inside
a layer of foamed plastic approximately
1/2 in. thick. The foamed plastic, in
turn, is placed in the copper box. The
two thermoelement leads are shunted
by 1,000-picofarad bypass capacitors.
These capacitors, connected from each
lead to the copper box, reduce any undesired r-f (radio-frequency) signals,
such as television carrier waves, but offer
high impedances to the audio-frequency
signal being measured. Because the cur
rent through the capacitor is small and
in quadrature with the current through
the thermoelement heater, the error
from this source is less than 0.001%, even
at 10 kc.
LOW

150 OHM
WWv-

>

Fig. 4 (left). Thermoelements in

equalization box
Fig. 5(<bove). Galvanometer circuit

Fig. 6. Circuit
for reducing d-c

321-4

Errors and Compensation

The inside of the instrument is lined
with copper, thus providing electrostatic
shielding. The common input terminal
is connected to this shield, and should be
grounded when the instrument is in use
The switching circuit is arranged so that
one side of both the Zener diode and the
voltage thermoelement heater T are
always grounded to this shield. Also,
the thermocouple circuit is connected to
the heater circuit to keep the voltage
across the insulating bead of each thermo
element below 2 volts. For some thermo
elements, errors may be caused by higher
bead voltages.
An error in setting the heater resist
ance Ra of the voltage thermoelement will
introduce an error of equal magnitude in
the voltage measurement at the 1-volt
level. However, this error is inversely
proportional to the voltage. An error in
setting the heater resistance Rt of the
reference thermoelement will introduce
only about 1/6 this error in the voltage
measurements since it is in the 5.8-volt
Zener-diode circuit. In the calibration
checks, the voltage thermoelement is
substituted directly for the 100-ohm
reference resistor R eliminating any
errors in the bridge circuit except for this
resistor.
During preliminary setting of voltage,
using the rectifier instrument, both
thermoelement heaters are connected in
series in the Zener-diode circuit. Keep
ing the thermoelements energized reduces
to a minimum the errors resulting from
the small initial driftup to 0.05%
which affects many thermoelements dur
ing the first few minutes after a large
change in heater current.'
Galvanometer sensitivity is uniform
within 1% of the end-scale deflection.
The maximum error introduced into the
measurement from this source is less
than 0.005% of the input voltage.
Sensitivity control in the galvanom
eter circuit and current adjustment in
the Zener-diode circuit are similar. Each
consists of an adjustable resistor, shunted
by a relatively low-resistance manganin
resistor This type of control is similar to
a Waidner-Wolf element with the adjust
ment continuous rather than in steps.
Such circuits minimize contact resistance
and thermal emfs.
Copper-to-copper connections, thermalfree keys, and other precautions were
taken to minimize residual thermal emfs
in the galvanometer circuit. The Zener
diode and filament transformer for the gal
vanometer light are mounted in a venti
lated enclosure that is thermally insulated

Table I. D-C Calibration of DTVM

Fig. 7. Circuit (or -c-d-c transfer measure

ments
VA-c voltase to be measured
DTVM Differential thermocouple voltmeter
VBVolt box
PD-c potentiometer
from the rest of the instrument. Tem
perature and residual thermal emfs inside
the instrument, although greatly reduced,
require approximately 6 hours to reach
equilibrium after the Zener diode and
lamp are energized. During this warmup period, the total drift in calibration is
approximately 0.05%. However, if the
calibration procedure is carried out at
half-hour intervals, the warm-up error
is reduced to less than 0.005%. After
the instrument reaches temperature equi
librium, it is stable within 0.01%, and a
calibration twice daily should be suffi
cient. The Zener diode and lamp are ex
pected to be continuously energized in
periods of normal use of the instrument.
The self-heating error from the voltage
to be measured is less than 0.01% on all
ranges.
The d-c reversal error of the DTVM
is approximately 0.01%, primarily in
the thermoelement. If the error were
significantly larger, it could be reduced
by the method shown in Fig. 6. The
magnitude of the resistor R, needed to
cancel the effect of reversal error, can be
computed from the equation, R=&E/2I,
where AE is the difference in thermo
couple output for ID=IB = I in Fig. 6.
The correction is so small that it need be
only approximate. The polarity must be
considered when connecting R into the
thermocouple circuit. This procedure for
reducing d-c reversal error also provides
a convenient method of connecting the
thermocouple and heater circuits.
Tests and Calibrations
Stability tests were made on the instru
ment extending over a 3-month period.
During this time, the heater resistance
R and Ri, of the thermoelements did not
vary by more than 0.005%. Voltage
measurements showed that the decade
resistor and the bridge components were
stable to 0.01%. The Zener-diode
voltage did not change by more than
0.005%, with no evidence of drift.
Supplementary tests verified the manu
facturer's specifications that the tempera
ture coefficient of the Zener diode was

XlOO

XlO

0
0
0
0
0
0
0
0
0
0
II
0
0
0

0
5
0
u
0
0
0
10
52
1
0
0
0
0
0
0
0
0
0
0

Dial Sottioft of DTVM

Xl
X0.1
1
1
1
1
1
1
1
1
1
1
1
11
10
51
1
1
1
1
1
1

X0.01

Volts

0
0
0
0
0
0
0
1)
0
0
0
0
0
0
0
0
0
0
10
5
0

601
551
501
401
301
201
101
101
at
21
11
11
10
5
2
1
1 5
1.1
1.1
1 05
1 00

0
0
0
0
0
0
0
0
0
u
0
0
0
0
0
10.r>
1
0
0
0

less than 0.001%/C, and that its out

put was regulated to 0.001% for a 10%
line-voltage variation.
Table I shows results of a d-c calibra
tion of the completed instrument, based
on average values for the two directions
of applied d-c voltage. The corrections
listed are all well within 0.05%, and
agree within 0.01% with those obtained
from resistance measurements of the
decade resistor.
Table II shows the a-c-d-c differences
for several combinations of settings of the
XlOO and XlO dials, where capacitance
effects are most likely to be appreciable,
and at the top, middle, and low positions
of the XlO and XI dials. Additional
difference measurements were made at
several other dial settings including the
X0.1 and X0.01 dials In no instance
was the a-c-d-c difference greater than
0.01% even at 20 kc. However, since
capacitance effects, and consequently a-cd-c differences, are dependent upon the
dial-setting combination, and because

D-C
1) S-i
M 00
0 00
0 00
Out
+0 'I.
^0 01
TO 01
0 00
to Oi
+0 01
T0 0!
+0 01
-0 01
+0 01
to M
rO 02
to 01
0 ft
0 00
0 ft'

tests at all possible dial settings are pr

hibitive, it was felt that the rated hquency should be conservatively set t
10 kc rather than 20 kc.
DTVM Uses
The instrument is particularly sate
for calibrating a-c voltmeters. Wtt
voltage adjusted so that the pointer d
the voltmeter under test is at a sea
marking, and with dials of the DTVM*
at the corresponding value, the percnti?
correction is read directly on the DTK
galvanometer. For highest accuracy, 4i
d-c correction of the DTVM should 5*
taken into account, so that C,=D-:,
where C, and C, are the percentage constions of the test instrument and fc
DTVM, respectively; D is the perce:age reading of the DTVM galvanomea
For a-c measurements of the hips
accuracy-0.02% or bettera direct Kd-c transfer method is usually recumand the circuit of Fig. 7 is used, W

Table II. A-C D-C Differ.net of DTVM

Dial Settlofa of DTVM*

XlO
XI
XlOO

Applied
Volte

20 Cpe

A-C-D-C Difference, %
Frequency
10 Kc
1 Kc
QflCpe

...0.00. . . .0.00 ... . .0.00. . . . . 0 00

01
. . .0.00. . . . 0 00 . . . . .0.00. . .. . +00.00
.0 00. . . .0.00. . . . . 0 00. .
+0 01
-0.01
-0 01
-0.01
. . 0 00. . . .. 0 00
101
. . .0 00. . . . 0 00
101
0 00. . . .. 0 00.
.. 1.. .
5
Q
0 00
1
1
Q
0 00
. . .10. . .
o
0
.
0 00
5 ..
o
o
1
0.00. . . ...0.00... . . 0 00. . .
* The X0.1 and X0.01 dials were set at 0 for these tests.
...10...
5
.. . .10. . .
*
0. .
B. . . .... 0 ...
o
2

...
...
...
...
.

1... .
1...
1...
1...
1. . .

322-5

601
601
501

of the DTVM are set to the nominal value

of the voltage to be measured. With the
DTVM switched to the a-c source, the
galvanometer deflection is read. With
the DTVM switched to the d-c source, the
d-c voltage is set by reference to a volt
box and potentiometer, and the galvanom
eter deflection is again read. The a-c
voltage being measured is then Vm = Vt
[1 + CD,-Ai)/100], where V, is the a-c
voltage; Vt is the d-c voltage, corrected
for potentiometer and volt box errors;
and D, and Dd are the galvanometer read
ings in per cent with the DTVM
switched to a-c and d-c sources respec
tively. An error will be introduced into
this measurement if the lead resistances
between the switch and the points of a-c
and d-c measurement are not equal.
Because of the low d-c reversal difference
of this DTVM, the d-c source and poten
tiometer need not be reversed for this test.
The DTVM can be conveniently used
for frequency influence tests or a-c-d-c
difference tests of other voltmeters.
Again, the circuit shown in Fig. 7 is used,
with the voltmeter under test connected
directly in parallel with the DTVM.
The volt box and potentiometer are not

required. Both instruments are switched

in quick succession to: (1) the a-c source,
(2) the d-c source, (3) the d-c source
reversed, and (4) the a-c source again.
In each case, voltages are adjusted to give
the same readings of the voltmeter under
test, and the deflection of the DTVM
galvanometer is read. The a-c-d-c differ
ence Si of the voltmeter under test in
per cent is simply S,= DDt, where
Dm and D& are the average galvanometer
readings with the instruments connected
to a-c and d-c sources, respectively. A
second a-c source can be used in place of
the d-c source if an a-c reference frequency
is desired. These procedures are highly
useful because the frequency influence of a
good instrument is small and relatively
stable, so that subsequent tests of the
instrument need be made only at the
reference frequencyreversed direct volt
age or alternating voltage.
Conclusions
This instrument is exceptionally con
venient and is a time-saver. The modi
fications that have been incorporated
make it much lighter and more compact

than the instrument on which it was

based.' Detailed descriptions of pre
cautions taken in design should be partic
ularly useful to others who may wish to
construct a similar voltmeter. Equally
important, these precautions illustrate
once again that a careful study of sources
of errors and of means of eliminating
them cannot be neglected if the high
accuracy inherent in a good basic method
of measurement is to be realized in an
actual instrument.
References
1. AC- DC Transpbr Instruments for Current
and Voltage Mba&uxbmbnts, F. L. Hermach.
Transactions, Professional Group on Instrumenta
tion, Institute of Radio Engineer!. New York
N. Y.. vol. 1-8. 1958. p. 23S.
2. A Wiue-Rangb Volt Ampere Converter
FOE Current and Voltage Measurements,
P. L. Hermach, E. S. William!. Al EE Transac
tions, pt. I {Communication and EUctrcmics), vol. 78,
Sept. 1959, pp. 384-88.
3. User DIB gbnaub Ahsolutmbssuno von
WbCHSSLSPANNUNGBN UNO BINBN KoMPBNSATOR
SUV PbOFUNO VON WKCHSBLSTROM PsiNMKSSobratbn, W. Rump. Eitktrolechnik, Berlin. Ger
many, vol. fi. Feb. 19.11. p. 64-67.
4. Audio Voltage Calibrating Standabd,
K. J. Koep, G. B. Ruble. A/EE Transactions, pt. I
{Communication and EUctrcmics), vol. 81, July
19152, pp. 179-86.

A reprint from COMMUNICATION AND ELECTRONICS, published by

American
Institutebyofpermission
Electrical Engineer!
Copyright 1962,
and reprinted
of the copyright owner
The Institute assume! lib responsibility for statements and opinion! made by
contributor!.
Printed in the United States of America
November 1962 issue

323-6

NBS Technical Note 257

March 1965

PRACTICAL ASPECTS OF
THE USE OF AC-DC TRANSFER INSTRUMENTS
E. S. Williams

Electrothermic transfer instruments may he used to

make direct measurements of ac-dc difference or frequency
influence. With a somewhat similar procedure an "unknown"
a-c voltage or current may be accurately measured by de
termining the difference between it and a preset and accu
rately measured d-c equivalent. Test circuits and proce
dures are described and data taking and calculations are
illustrated.

NBS Technical Note 188

April 1963

Calibration of Volt-Ampere Converters

E. S. Williams
These notes have been prepared to describe the National
Bureau of Standards calibration services for volt-ampere con
verters (or transfer volt-ammeters), to suggest procedures
for d-c standardization in the user's laboratory, and to de
scribe a voltage comparator which can be used to make such
calibrations quickly and easily.

324-1

U. S. Department of Commerce

National Bureau of Standard*

RESEARCH PAPER RP1344

Part of Journal of Research of the National Bureau of Standards, Volume 25,
November 1940

STANDARD ELECTRODYNAMIC WATTMETER AND AC-DC

TRANSFER INSTRUMENT
By John H. Park and Arthur B. Lewis
abstract
A description is given of the design and construction of the standard instrument
used at the National Bureau of Standards in testing wattmeters and watthour
meters.
The investigation to determine the accuracy of this instrument for alternatingcurrent testing at frequencies up to 2,000 cycles per second is described and the
results are given.

>urnal of Research oQhe National Bureau of Standards

Vol. 53, No. 2, August 1954

Research Paper 2521

Precise Comparison Method of Testing AlternatingCurrent Watthour Meters

A. W. Spinks and T. L. Zapf
A brief description of the basic method of testing alternating current watthour meters
at the National Bureau of Standards is given, followed by a description of equipment for a
faster and less laborious method.
Equipment with several novel features has been assembled for making precise tests of
alternating-current watthour meters by a comparison method employing a group of carefully
selected alternating-current watthour meters, which serve as a secondary standard group.
One of this group, designated the "Standard Watthour Meter", is used with multirange
instrument transformers as a reference standard to test other watthour meters with good
precision. The testing procedure is explained, and the formulas used in computing the
results of the tests are derived.
An analysis of the possible errors of measurement and data from numerous tests indicate
that the measurement of energy applied to a watthour meter under test can be relied upon
to better than 0.06 percent.

325-2

NBS Technical Note 266

December 1965

VOLTAGE RATIO DETECTOR FOR

MILLIVOLT SIGNALS
by J. R. Houghton

ABSTRACT
A voltage ratio detector circuit for measuring
ratios of a-c and d-c signals 5 millivolts or larger
is described. The ratio is determined with a pre
cision voltage divider which is accurate to within
0.001 percent of the indicated ratio when the ratio
is near one. The detector has sufficient sensi
tivity and stability to indicate differences between
two signals of 0.01 percent. Experimental results
are presented to show the relative improvement in
sensitivity of this voltage ratio detector over the
previously used transfer admittance method for the
calibration of vibration pickups.
Key words: Voltage ratio, vibration, calibration,
thermal converter, electronic circuit, test method.

Instruments
Volume 23, Number 12, December 1950

Notes on the Care and Use of Electrical Instruments

F. D. Weaver
This article presents notes on precautions that should be
observed by users of electrical instruments to enable them to
achieve greater accuracy in measurements and to extend the use
ful life of their equipment.

326-3

Transformers and Inductive

Voltage

Dividers

Papers
An international comparison of current-ratio standards at audio fre
quencies, B. L. Dunfee and W. J. M. Moore
Comparators for voltage transformer calibrations at NBS, W. C. Sze..
An international comparison of voltage-transformer calibrations to 350
kV, F. K. Harris, W. C. Sze, N. L. Kusters, O. Petersons, and W. J. M.
Moore... ...
The precision measurement of transformer ratios, R. D. Cutkosky and
J. Q. Shields.
Comparison calibration of inductive voltage dividers, R. V. Lisle and
T. L. Zapf
Comparator for calibration of inductive voltage dividers from 1 to 10
kHz, W. C. Sze
_
.
..
.....
An international comparison of inductive voltage divider calibrations at
400 and 1000 Hz, W. C. Sze, A. F. Dunn, and T. L. Zapf

329
335

342
349
357
362
371

Abstracts
Equipment for testing current transformers, F. B. Silsbee, R. L. Smith,
N. L. Forman, and J. H. Park
The design and performance of multirange current transformer standards
for audio frequencies, B. L. Dunfee
Inductive voltage dividers with calculable relative corrections, T. L.
Zapf, C. H. Chinburg, and H. K. Wolf
The calibration of inductive voltage dividers and analysis of their opera
tional characteristics, T. L. Zapf
315-411 O - 68 - 22

327

379
379
380
380

An International Comparison of Current-Ratio

Standards at Audio Frequencies
BERNADINE L. DUNFEE, senior member, ieee, and W. J. M. MOORE, senior member, ieee

AbstractThe results and analysis of an intercomparison beween the National Research Council (NRC), Canada, and the
rational Bureau of Standards (NBS), Washington, D. C, of two conrasting types of current ratio standards (current comparator and
urrent transformer) are presented. The agreement achieved beeen the two laboratories when their respective designs and
lethods of measurement were quite different is emphasized. To this
nd, the basic theories underlying the design and operation of the
urrent comparator and current transformer are contrasted; the
rigin and significance of their respective errors are summarized ; the
V-f I) method used at NBS to measure the errors of the multirange
ansformer standards up to 10 kHz is contrasted with the (rV+1)
ethod (one of several) used at NRC in calibrating the comparator
andards. The comparison circuit used at NRC to compare the
spective standards of the two laboratories is described. The errors
the transformers as measured at NBS and NRC are presented
id contrasted in both tabular and graphical forms.
Introduction
N MANY INSTANCES the measurement of an ac
electrical quantity is made possible or, at least,
facilitated if accurately known ratio and phase rela>ns can be established between two or more currents,
lis is particularly true when the measurements cover
broad frequency range or when the parameter being
;asured is relatively large (or small) in magnitude.
Anticipating future needs, the National Research
mncil (NRC), Canada, and the National Bureau of
indards (NBS), Washington, D. C, have been concting independent programs in current-ratio meaements which include: 1) the design and construcn of current-ratio standards for use at audio fre;ncies; 2) the development of methods for measuring
errors of the standards; and 3) the study of their
and application in the measurement of electrical
intities.
"ontrasting approaches in both the design of the rertive standards and in their calibration have been
sued. NRC has explored the current comparator
[4\ and extended its use up to 16 kHz; NBS, on
other hand, has focused on the current transformer,
:ring approximately the same frequency range. The
iective work of each laboratory is described in two
pan ion papers [5], [6].
he two different approaches have provided a more

lanuscript received June 28, 1%5.

crnadine L. Dunfee is with the Electricity Division, Electrical
jments Section, National Bureau of Standards, Washington,
'. J. M. Moore is with the Division of Radio and Electrical
leering;. Electrical Engineering Section, National Research
cil of Canada, Ottawa, Canada.
329

complete understanding of the behavior and limitations

of current-ratio standards and their associated measur
ing systems, and an international comparison has pro
vided an independent method for verifying results and
estimated accuracies.
This paper incorporates information from the com
panion papers pertinent to the present discussion and
describes in detail the results obtained when the NBS
and NRC standards were compared at NRC.
Comparator and Transformer in Contrast
In its simplest version, the current comparator com
prises a toroidal core of high permeability which carries
a single-layer detection winding of many turns for sens
ing the flux condition of the core, and two windings,
identified as primary and secondary, that carry the
currents to be compared. It embodies the principle ex
pressed by Ampere's Circuital Law tfH dl = 2I, with
the added provision that when the line integral equals
zero, the net electromotive force (EMF) induced in the
detection winding, as measured at the terminals, is
also zero. The significance of this equality is that the
core operates under a condition of zero flux at the time
of measurement so that errors that would arise from
magnetizing and loss currents are avoided. Although
this kind of magnetic error is eliminated and a greater
inherent accuracy is assured, other errors of magnetic
origin are present unless modifications are made in the
basic design. As described by analysis and experiment
[2], leakage flux assigned to either winding or to the ex
ternal environment induces an EMF in the detection
winding which is not a simple function of the currents
being measured. Furthermore, the nonuniformity of
permeability around the magnetic path produces an
EMF even though the line integral may be zero. How
ever, the inclusion of toroidal magnetic shields which
surround and separate the critical windings can reduce
the magnetic error to negligible proportions.
In contrast, the single-stage current transformer car
ries only two windings (primary and secondary) dis
tributed around a toroidal core of high permeability.
Since an EMF is required to establish and maintain
a current through the total impedance of the secondary
circuit, the operation of a current transformer requires
a flux to be established in the magnetic core. Thus a
small component of primary current must supply the
working flux and the losses associated with the magnetic
circuit so that, unlike the comparator, an error arises
from this cause. Because it is of magnetic origin, this
172

error can be identified as the magnetic error of the cur

rent transformer. Furthermore, since the flux is a non
linear function of the magnetizing current, the error be
comes a function of burden, frequency, nominal ratio,
and secondary current. Some advantage is gained when
the current transformer is to serve as a standard, since
it need operate at only one burden which can be chosen
with discretion. Thus at low frequencies, where the error
is chiefly magnetic in origin, the current comparator is
appreciably superior to the current transformer.
The only type of error considered thus far is the
magnetic error, which, at frequencies not greatly ex
ceeding 60 Hz, is the only one of concern. However, as
the frequency increases another error of different origin
is introduced. This error is associated with the windings
and arises from the distributed capacitances that are
present and acted upon by voltages that are either im
pressed or induced. These capacitances may exist within
a winding, between windings, dr between windings and
shields. Thus, they are present in both types of ratio
standards and their contribution to both the in-phase
and quadrature components of the error increases with
frequency. In contrast, the magnetic or low-frequency
error of the current transformer decreases with increased
frequency so that gradually its contribution to the total
error becomes negligible and the capacitive or highfrequency component becomes dominant. Also, at higher
frequencies the change of error with secondary current
and the effects from residual magnetization become less
significant. Thus, in the upper audio frequency range,
the source of error is essentially the same in the two
types of ratio standards and, except for differences in
design that may alter the respective capacitance dis
tributions, the marked advantage of the current com
parator over the current transformer, as a ratio standard,
diminishes.
Complete analysis of capacitance errors in compara
tor and transformer designs would be a formidable task
and outside the scope of any one paper. However, gen
eral discussions and analyses pertinent to certain con
figurations appear in [7]-[ll], and the two companion
papers [S], [6]. It is sufficient here to state that, in gen
eral, the capacitive error is a complex function of sev
eral parameters with the ratio and phase angle com
ponents varying with the second and first power of the
frequency, respectively. There does exist, however, an
important concept that is necessary to the present dis
cussion. A primary or a secondary current, as such, can
no longer be assigned to the respective windings. In the
presence of capacitance currents, the current entering
one terminal of a winding will not be the same as that
leaving the other terminal. Furthermore, the distribu
tion of capacitance currents changes with any alteration
in the voltage distribution or when a change is made in
the "tie" connection between primary and secondary
terminals.
Thus, several ratios for a given nominal ratio are
available and a measurement can have meaning only
by 1) specifying the potential of each winding, and

2) identifying the currents at the terminals where the

measurement is made. The ratio discussed in this text,
unless otherwise stated, refers to the currents at the
terminals of like polarity when those terminals are at
ground potential.
Calibration Prior to the International
Comparison
The current comparators used in the intercomparison
were part of the group described in one of the com
panion papers [5]. Each of the group was multirange
and together covered a large number of selected ratios
which permitted their calibration by three independent
methods in a buildup or "boot-strap" process. In addi
tion, ratios of the same nominal value were measured
in a comparison circuit similar to that used in calibrat
ing current transformers. Measurements were made at
selected frequencies to a maximum of 16 kHz. The
estimated uncertainty in the error e assigned to the
nominal ratios was 1 part per million (ppm), where the
error t is defined by the equation /, = /,/n(l +). [/,
and /, refer to the primary and secondary currents at
terminals of like polarity when those terminals are at
ground potential and n is the nominal ratio. ]
The two essentially identical NBS current transform
ers are discussed in the other companion paper [6]. Thev
were multirange and carried the consecutive ratios of
1 to 1 to 6 to 1, inclusive. Measurements were made at
NBS on all ratios at selected frequencies from 400 Hz to
10 kHz. A single "boot-strap" method was employed
with verification of the results by an independent meth
od provided by the intercomparison. The accuracy of
the measured values was estimated at 10 ppm or better.
The "boot-strap" techniques employed at both
laboratories required an initial calibration of the 1-to-l
ratios, which then served as the basis for the successive
calibration of the higher ratios. The self-calibration
circuits for measuring the 1-to-l ratios and the balance
equation (applicable to both the current transformer
and current comparator) are indicated in Fig. I.1 In
each case, the conductance G and capacitance C are
adjusted for a null on detector D.
The higher ratios of the current transformers were
measured in succession using the (iV+l) circuit of F;e
2(a) [for the sake of clarity only the essential elements
are included]. After each transformer was calibrated or.
its n ratio, it served as the standard 5 in calibrating the
( + l) ratio of the other, identified as X. Thus, begin
ning with 5 connected for its 1-to-l ratio with currentsadding at A, calibration is made for the 2-to-l ratio of
X by measuring, in effect, the difference between their
nearly equal secondary currents; the process is con
tinued by comparing the 3-to-l against the 2-to-l etc ,
throughout the total range. The inserted voltage indi
cated by V adjustable in magnitude and phase, corr.1 The current comparator shown in this design is of the compen
sated type [3]. It has an additional winding (known as the compe.sation winding) with the same number of turns as the secondarv
winding located inside the magnetic shields.

330-173

(b)
Fig. 1. Self-calibration circuit (1-te-l ratio), (a) For current
transformers, (b) For current comparators < = r(G +jaC).

(W
Fig. 2. (AT+1) calibration circuit for ratios>l. (a) For current
transformers, (b) For current comparators.

pensates for the added burden impressed on X by its of 5 differs from that used when it is calibrated as the
unknown X ; furthermore, measurements are made on X
own primary winding.
The (N+ 1) circuit of Fig. 2(b) represents one of three at the ends opposite to where the terminals are joined.
"boot-strap" methods used at NRC in the calibration The balance equation for this circuit contains two un
of the comparators. Bearing the same identification as known parameters which must also be evaluated from
the one used at NBS, it serves to emphasize the similar a set of auxiliary measurements.
ities and differences between the two methods. Com
The Intercomparison
parators S and X serve in the same capacity as that out
The intercomparison was accomplished by using the
lined for the transformer calibration and the succes
sive buildup through the higher ratios is identical in the NRC current comparators to measure the errors of the
two cases. The secondary burden imposed on either 5 NBS current transformers (identified as M-l and M-2).
sr X remains constant and, therefore, unlike the current The circuit is described in one of the companion papers
[5] and is reproduced in Fig. 3, greatly simplified for the
transformer circuit, requires no compensation.
Insidious effects from capacitance currents occur in sake of clarity.
An over-simplified approach is perhaps useful here in
30th circuits which could introduce errors in the meaiurements if not properly considered. The measuring understanding the circuit. Disregarding any capacitive
jranch in the network of Fig. 2(b) is confined solely to or magnetic errors in the current comparator, assume
:he secondary circuit so that the currents being com- first that the current transformer is free of error and that
>ared are consistent with the definition; however, be- the G-C network is disconnected from point M,. The
ausc of capacitance currents, the current that leaves primary and secondary windings of the comparator to
me terminal of the secondary winding of 5 differs from gether with the magnetic shield constitute, in effect, a
Jiat which enters the other terminal. Since this current current transformer and hence the ampere-turns of the
nters the primary circuit at Mp a correction /3B, previ- primary and secondary windings will nearly balance one
msly measured, appears in the balance equation. A another. However, complete ampere-turn balance must
imilar but more complicated condition exists for the exist in the comparator as a whole. The residual ampere^BS circuit [Fig. 2(a)]. The measuring branch is not turns must, therefore, be provided by the current in the
onfined to the secondary circuit but bridges across compensation winding. This winding has the same num
irimary and secondary. The "tie" between the windings ber of turns as the secondary winding, hence the division
331-174

of current between the two windings is immaterial. that enter and leave terminals of like polarity througha
When the current transformer under test is not error- pair of shielded leads when the ends of the leads are at
free, the procedure is to make it appear so to the current ground potential; a noninductive burden of 0.10 ohm
comparator by injecting current into the circuit through was imposedlead resistance was small but was essen
the G-C network. The values of G and C at balance are tially the same in both calibrations.
Measured values and differences are shown in Table!
then a measure of the current transformer's error. Any
error in the current comparator, of course, enters the for transformer M-1 where the errors are given in ppm
(for phase angle this is the equivalent of microradians)
balance equation.
It was imperative that the quantity measured and the the errors plotted against frequency for ratios 1 to 1
conditions imposed on the current transformer be iden and 6 to 1, appear in Figs. 4 and 5. The measured errors
tical in the two laboratories to the extent that the de for current transformer M-2 were identical to those
sired accuracy is not impaired. In this respect, the bur shown in the table for M-1 to within 2 ppm;* the differ
den imposed was the same in the two cases and the ences between the laboratories are essentially the same
ratios measured refer to the same currents, conforming with a maximum again of 8 ppm. Although the values in
to the definition specified earlier. The latter can be Table I and the graphs refer to a secondary current oi
clarified further by considering other aspects of the mea 5 amperes, measurements were made, also, at other
suring process. The current transformer, equipped with secondary currents to a minimum 0.5 ampere. The
shielded leads, is connected to the comparator at junc largest change in going from 100 to 10 percent rated
tions Mp and M,. The potential condition is satisfied by current occured at 1 kHz and amounted to 0.5 ppmic
bringing points M, and M, to ground potential. This is both the in-phase and quadrature components.
The errors as recorded in Table I are the total errors
achieved by 1) making a Wagner arm balance (not
shown) and 2) injecting a small voltage at g to compen that include both the magnetic and capacitive com
sate for the impedance drop in the compensation wind ponents. The capacitive portion is the sum of all con
tributions from the winding impedances, lead imped
ing.
The use of leads to connect the transformer to junc ances and any capacitance that shunts the primary- The
tions Mr and M, requires comment. Measurements at first two can be determined experimentally [5 ] so that
NBS were made at the terminals, i.e., the shielded leads with some reasonable estimate of the shunt capacitance,
(shields not shown) used in the NRC circuit and con an artifice is established for computing the magnetk
sidered as part of the transformer were not present in error. The magnetic error was estimated in this manner,
the NBS circuit. Since the potentials of the windings, its value for the six ratios at 10 kHz is given in Table II.
and therefore capacitance currents, are altered by the using NRC values and assuming a shunt capacitance a
insertion of leads, the transformer errors as measured 150 pF.' Since the contributions from capacitance to the
in the two laboratories would differ at the higher fre real and quadrature components of error vary with the
quencies. It was necessary, therefore, to measure the second and first power of the frequency, respectively
"lead effect." This was done at both NRC and NBS estimates of the magnetic error can be made for the re
with an agreement of 1 ppm. NBS values for the mea maining frequencies.
sured error t were adjusted accordingly, the maximum
adjustment of approximately 10 ppm in the real com
This applied to both NRC and NBS values with one except-"
ponent occurring on the six-to-one ratio at 10 kHz.
at 1 kHz on the 3-to-l ratio the NBS value for phase angle differs:
from the M-1 value by 6 ppm. It was believed that polarizatk:. e
be discussed later, accounted for this discrepancy.
Results (1 to 10 kHz)
' The estimate is based principally on the expected capacitan*
the magnetic core (which is not grounded) and the winda?
To re-emphasize, the errors measured at NBS (ad between
turn-to-turn capacitance or that resulting from terminal geocK^
justed for lead effect) and NRC refer to those currents would be much smaller for this design.
332-175

TABLE I
Errors of Transformer M-l Measured at NRC and NBS at a Secondary Current of 5 Amperes
Parts Per Million
NoraP inal
Ratio
1/1
2/1
3/1
4/1
5/1
6/1

NBS
-5.1+720.0
-5.0+718.9
-5.1+/20.1
-5.6+7'21.6
-5.8+/20.3
-6.3+/21.9

1 kHz
NRC

-5.4+j20.2
-5.3+719.0
-5.3+;19.4
-5.3+/20.0
-5.4+^20.1
5.5 +j"2 1 . 7
A = the difference (NBS-NRC).

NBS
-3.8+il.S
-2.4+72.7
-2.0+/3.2
-1.9+/4.2
-1.5+j4.6
-0.9+/4.2

+0.3 -jO. 2
+0.3-j0.1
+0.2+/0.7
-O.3+7I.6
-0.4+70.2
-O.8+7O.2

5 kHz
NRC
-4.0+/2.1
-2.8+7'1.3
-2.5+i2.3
-2.3+J2.9
-2.0+73.3
-1.8+73.9

+0.2 -70. 6
+0.4+/1.4
+0.5+70.9
+0.4+71.3
+0.5+71.3
+0.9+70.3

10 kHz
NRC
-2.I-72.6
+2.I-74.3
+3.I-72.3
+3.9-7I.4
+5.2-70.7
+6.3-7O.I

+0.3+70.2
+2.0+72.2
+3.3+7I.6
+4.5+72.6
+6.5+72.5
+8.4+72.5

20

15

10
j
/
?
i

NBS
- I.8-72.4
+ 4.1-72.1
+ 6.4-7O.7
+ 8.4+7I.2
+ II.7+7I.8
+ 14.7+j2.4

/*

in 15
z
<
o
4
(E
O
or
o 10

//\ Rc
8
cc

6/l Rotio
\

LlJ
<a1
0-

'/l Rotio

4
6
8
10
FREQUENCY - kHz
Fig. 4. Ratio error of current transformer M-l; NBS measure
ments
, NRC measurements
.

otio

\;

4
6
8
10
FREQUENCY - kHz
Fig. 5. Phase error of current transformer M-l; NBS measure
ments
-, NRC measurements
.

TABLE II
Estimate of Capacitive Errors for Transformers M-l and M-2 at 10 kHz
Parts per Million
Nominal
Ratio
1/1
2/1
3/1
4/1
5/1
6/1

NRC Calibration
M-l
-2.I-72.6
+2.1-74.3
+3.I-72.3
+3.9-7I.4
+5.2-7O.7
+6.3-7O.I

M-2
-1.9-72.7
+2.3-74.4
+3.2-72.5
+4.I-7I.4
+5.4-7O.9
+6.6-7O.3

Capacitive Error Components Due to

150 pf Primary
Winding
Lead
Shunt
Impedance
Impedance
Capacitance*
O.O-7O.6
+2.2-73.3
0
+3.I-73.9
+2.9-7I.O
+O.5-7O.8
+2.4-72.4
+0.2-70.4
+4.6-7I.2
+ I.7-7I.2
+6.5-7I.6
+0.2-70.2
+ I.I-7O.3
+8.I-7I.7
+O.I-7O.I
+0.7+70.4
+9.5-71.9
+O.I-7O.I

Residual Error
M-l
-4.3+7I.3
-4.4+7I.4
-4.1+71.7
-4.5+7I.6
-4.I+7I.4
-4.O+7I.5

M-2
-4.I+7I.2
-4.2+7I.3
-4.O+7I.5
-4.3+7I.6
-3.9+7I.2
-3.7+7I.3

' Assumed
Results (400 Hz)
The results at 400 Hz are examined separately in
rder to emphasize the effects of polarization observed
t low frequencies. The values and differences for both
1-1 and M-2 are given in Table III. The relatively poor
greement for phase angle is in startling contrast to that
btained at higher frequencies. This fact appears conary to that expected, since the role of capacitance

currents has essentially disappeared. However, polar

ization and its effect in current transformers become
more critical at the lower frequencies and extreme care
must be taken in demagnetizing the core and in avoid
ing sources of direct voltage in the measuring network.
From evidence obtained at both N RC and N BS, it was
concluded that the large differences evident at 400 Hz
were caused by polarization, established periodically

333-176

TABLE III
Errors of Transformers M-l and M-2 at a Frequency of 400 Hz and a Secondary Current of S Amperes
Parts Per Million
M-l
Nomina!
Ratio
1/1
2/1
3/1
4/1
S/1
6/1

NBS*

NRC

-6.9+j49.3
-7.0+746.9
-6.8+>49.9
-7.9+7'61 .4
-8.1+759.6
-8.9+7'65.4

-6.9+7'49.8
-6.9+746.1
-7.1+746.4
-7.1+747.7
-7.3+7'47.5
-7.4+752.4

M-2
(NBS-NRC)
O.O-7O.5
-O.I+7O.8
+O.3+73.5
-O.8+7I3.7
-O.8+7I2.I
-1.5+713.0

NBS*

\RC

-6.6+7'54.0
-6.8+7'59.2
-7.4+765.1
-7.4+>66.4
-8.2+772.0
-8.3+768.2

-6.2 +748.6
-6.6+747.8
-7.0+750.5
-7.0+751.9
-6.5 +746.9
-6.5+747.7

A
(NBS-NRC >
-0.4+jj 4
-O.2+7IH
-O.4+7H.6
-0.4+714.5
-1.7 +725.1
-1.8+720.S

NBS values are an average of maximum and minimum values when repeat measurements were made following a demagnetization proas
during the "boot-strap" measuring process at NBS. different personnel responsible for the measurements
This evidence is discussed quite fully in Dunfee [6]; The good agreement implies that the "boot-straphowever, certain of its aspects should be noted here.
method of NBS together with the auxiliary measure
Small magnetizations can produce relatively large ment give results accurate to 10 ppm and verifies fur
effects at 400 Hz. A maximum change of 20 ppm in the ther the validity of the methods used at NRC in
quadrature component of the error immediately follow evaluating the comparators. Experimental evidenoe in
ing demagnetization was observed at NBS; essentially dicates that careful demagnetization of the transfonrthe same value was obtained at NRC when the trans ers must be observed at frequencies near and belos
formers were calibrated on their 1-to-l ratios "as re 400 Hz.
ceived" and followed immediately by a calibration after
References
demagnetization. In both laboratories, the maximum
[1]
N.
L.
Kusters
and
W.
J. M. Moore, "The current compara:'
change was reduced to about 8 ppm at 1 kHz, and con
and its application to the absolute calibration of current tni-*tinued to decrease approximately with the first power
formers," Trans. AIEE (Power Apparatus and System], P'III, vol. 80, pp. 94-104, April 1961.
of the frequency. The NBS values in Table III result [2] P.
N. Miljanic, N. L. Kusters. and W. J M. Moore. Tkt*
from averages taken of the maximum and minimum
velopment of the current comparator, a high-accuracy a-c r?~<
measuring
device," Trans. AIEE (Communications and /-
measured values. Selection of the minimum values
Ironies), pt. I, vol. 81. pp. 359-iCS. November 1962.
would improve the agreement between the laboratories:
[3J N. L. Kusters and W. J. M. Moore, "The compensated rarent comparator; A new reference standard for current trans
however, there is no justification for this choice because
former calibration in industry," 1964 IEEE Inlernatl Com. fo
of the complex manner in which uncertainties are propa
ot. 8, pp. 204-212.
[4] N. L. Kusters, "The precise measurement of current rati*'
gated through the "boot-strap" process. Strong evidence
presented at the 1964 International Conference on Prec:*:"
supports the fact that when the measuring circuit is
Electromagnetic Measurements, Boulder, Colo.
|5)
N.
L. Kusters and W. J. M. Moore, "The development and per
devoid of direct voltage sources, careful demagnetiza
formance of current comparators for audio frequencies.' tki
tion just prior to calibration assures a repeatability of
issue, page 178.
[6] B. L. Dunfee, "The design and performance of multirrj!
results to 1 ppm.
current transformer standards for audio frequencies," this skx
page 190.
Conclusions
|7] P. N. Miljanic, "Capacitive error in current comparators,* prfr
sented at the 1964 International Conference on Precision FiffThe use of two or more distinct methods to measure
tromagnetic measurements, Boulder, Colo.
|8|
A. H. M. Arnold, "Dielectric admittances in current trans
the same quantity offers the best approach in establish
formers," Proc. IEE (London), pt. 2, vol. 99, pp. 727-734. 15V
ing the accuracy limit of a measured value and verifying [9]
" The effect of capacitance on the design of toroidal er
rant transformers," Proc. IEE (London), pt. 2, vol. 97, pp. "J"the methods employed. This arises principally because
808, 1950.
it is quite unlikely that the sources, magnitudes, and dis [10] N. L. Kusters and W. J. M. Moore, "The effect of wincrf,
potentials on current transformer errors," Trans. AIEE i6
tribution of the errors of measurement are identical in
munications and Electronics), pt. I, vol. 81, pp. 186-191, J-:the several methods. In this intercomparison, not only
1962.
[11|
J.
J. Hill, and A. V. Miller, "Design and performance of rsi
were the methods and circuits markedly different in the
precision audio-frequency current transformers," Proc. iti
two laboratories but possible bias was reduced in having
(London), pt. B, vol. 108, pp. 327-332, May 1961.

1965

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT

334-177

DECEMBE!

JOURNAL OF RESEARCH of the National Bureau of StandardsC. Engineering and Instrumentation

Vol. 69C, No. 4, October-December 1965

Comparators for Voltage Transformer Calibrations at NBS

Wilbur C. Sze
(July 28. 1965)
An inductive comparator is described for measuring ralio-correction-factor and phase-angle error
of voltage transformers by comparison with reference transformers. The simple circuit utilizes a
resistance-capacitance network and an inductive voltage divider. The relative magnitude of the ratiocorrection-factors is obtained from the dial readings of the inductive voltage divider, and the phase
angle between the voltage phasors is equal to a>CR. The accuracy of these measurements at 60
and 400 hertz is within 2 parts-per-million for ratio and within 10 microradians for phase angle at the
limits of the phase-angle range of 38 milliradians. Only a few ranges of reference transformers are
required since fractional ratios from 1/1 to 1/4 may be measured as readily as ratios near unity, without
diminution in accuracy. A general discussion of a resistive comparator, which has been in con
tinuous operation at NBS since 1939, is included.
Key Words: Voltage divider, inductive voltage divider, resistive voltage divider, voltage trans
former calibration, ratio-correction-factor, phase-angle errors, voltage transformer
comparator, voltage-transformer-test-set, comparator, test-set, calibration.

1. Introduction
The method most widely used by public utilities
nd standardizing laboratories to determine the ratio
id phase angle of voltage transformers is by comarison with calibrated reference transformers. How
ler, it is a relative rather than an absolute method,
i that the accuracy of the results obtained depends
pon the knowledge of accuracy of the reference
ansformers [1].'
At the NBS laboratories, a group of reference
ansformers is available by means of which test transrmers having any ratio from 1 : 1 to 2000:1 can be
ilibrated. The accurate initial establishment of
le voltage ratio and phase angle for these reference
ansformers has been discussed elsewhere and will
>t be repeated here [2, 3].
The purposes of this paper are (1) to describe the
sistive comparator which has been in continuous
>eration since 1939; (2) to describe a new inductive
imparator; and (3) to analyze various factors affecting
e accuracy with which measurements can be made.
2. Resistive Comparator
This comparator was designed and constructed at
BS by F. K. Harris and has been in continuous
.eration since 1939, except for a very brief period
tring which new switches were installed to replace
e worn out original ones [4, 5].
The schematic circuit of this comparator is shown
figure l.2 The in-phase voltage balance condition
achieved by adjusting the tap of the resistive diier, and the quadrature voltage, by varying the mual inductor, M.
figures in brackets indicate the literature references at the end of this paper.
Simplified by not showing the shielding arrangements around each segment of A\.

Figure 1.
2.1. Description of Components
The R, and Rm circuits are supplied from the sec
ondary of transformer S. Each can be varied from
2000 to 4000 n in increments of 100 ft. The exact
values are determined by the ratios of the test and
reference transformers. The 100-ft and 1000-ft unit
resistors of R, are of woven wire and mounted on
ceramic forms. Thus the residual, L,, of R, is re
duced to a minimum [6]. The Rm circuit, nominally
equal to R,, include the primary circuit of M which
has a d-c resistance of 200 ft. The capacitor, C,
shunting the lower portion of the resistance, serves
to compensate for the self-inductance introduced by
M [5]. Multiple taps are brought out from Rm to sup
ply the shield potential of corresponding sections of
Rs and of the detector circuit [7]. The shielding is
shown in figure 4.
As shown in figure 2, the portion of the R, circuit
between 1800 and 2000 ft carries a continuously ad
justable tap, Rx, which is connected through the de
tector circuit and M to the secondary of the test
transformer.

335-257

isoon

tooa

9 x10a

x io

FlCURE 2.

The mutual inductor used in this circuit consists

of a 10 X 20 mH per section unit ( 200 mH maximum)
and a continuously adjustable inductometer ( 25
mH maximum) [8, 9].
Both the RJRX ratio of the resistive divider and
the mutual inductance, M, are periodically redeter
mined to insure that maximum accuracy is maintained
for transformer testing. The residual reactances,
L, and Lx, were also accurately determined at the
time the comparator was built.
The design is such that the resistors must dissipate
considerable power. Therefore, the entire resistive
divider, Rs and Rm, is immersed in oil to minimize
the effect of temperature coefficient on the RJRX
ratio.
The detector circuit consists of a doubly shielded
impedance-matching transformer (1000-4000 turns)
and a wave-analyzer used as a tuned null detector.
2.2. Comparator Desk
Figure 3 shows the general layout of the desk. The
resistive divider is located to the right with the R,
adjustment dials in the back and the Rx dials in front.
The step mutual inductor and the inductometer are
located in the left portion of the desk. The detector
is in the rear right. Since this comparator is adapt
able for calibrating voltage-transformer-test-sets, the
panel shown at the lower left side of the desk is for
arranging the connections for calibrating either trans
formers or test-sets. Figure 4 shows the internal
wiring of the resistive divider and the individual
shields around units of R,.

Figure 3.

terminal voltage so that

l^secl

NJ

where N' is the actual voltage ratio, N the m

ratio, and /the ratio-correction-factor. If 0 lepra*
the angle between the primary and reversed second
voltage phasors, then
W I
v'=ffr (cos 0,+j sin $,)
and 3
V***^ (cos 0X + j sin 8X)
By inspection of figure 1, the following mesh a"
equations are obtained

Vx = hZx + UZX + ZM.) + /jj'wM

2.3. Theoretical Relations
V, = hjwM + hZm
The equations for obtaining the ratio-correctionfactor and phase-angle error are developed as follows.
By definition, the voltage ratio of a transformer is the
ratio of the primary terminal voltage to the secondary

'Subscripts j and x denote (hat the symbols refer to (be reference or W ^

respectively.

336-258

and 4
(Lx

L.M\

0X = 0 + w
As previously indicated. Rs and Rm circuits con
stitute a burden on the reference transformer. There
fore,/, and 8, are determined with these burdens.
The design of the divider is such that the balance
conditions require

n;

Rs

'

this requirement is met by using reference transform

ers whose voltage ratios have been adjusted to values
2 percent less than the integral values usually en
countered as nominal ratios in instrument
transformers.
The minimum resolution of this setup is such that
the precision for ratio measurements is 2 ppm and for
phase angle, 15 fjurad at 25 and 60 Hz. However, sus
ceptibility of the mutual inductor to the gradients of
stray magnetic fields, large residuals and uncertainty
in ratio calibration of the resistive divider limit the
accuracy of these measurements to 10 ppm and 60
/xrad, respectively. These errors increase rapidly
with frequency. Therefore, this comparator is not
suitable for accurate calibrations beyond 60 Hz.
3. Inductive Comparator
Figure 4.

2 = 0 when a null is indicated on the detector, then

nd

ince Zm = Rm by the addition of C, and Rm = Rs

ominally, the balance equation becomes
>s 0x+j sin gj)_(cos 6g+j sin 6)

This comparator avoids a number of the objection

able features of the resistive comparator. The shield
ing is simple and is easily accomplished; the method
is capable of extension to cover the audio-frequency
range; the apparatus is less bulky and requires very
little power for operation; the components are in
herently more accurate and stable and readily avail
able commercially; the ratio difference is indicated
directly on the decade dials; and the set (except the
detector) is self-contained in one portable case.
The schematic diagram of the circuit is shown in
figure 5, which indicates the positions of resistor, R,
and detector when the voltage phasor of the test trans
former, X, leads that of the reference transformer, S.
For lagging angles, the positions of R and detector are
* The
1) and 9, for the NBS reference transformers are very small. Therefore,
ihe approximations made are justifiable.

Dttector
id on separation

H.V. (} v>
Supply V-/
/ cos ex

cos 0g sin 6gO)

\Rs

R.

Rj'
5Figure 5.
337-259

simply interchanged. The in-phase voltage balance

is obtained by adjusting the tap on the inductive volt
age divider, and the quadrature voltage balance is
obtained by adjusting either the resistor or the ca
pacitor.

power factor), the effect of loading of the fz anc

of the test transformer can be expressed as
(Equivalent Total Impedance Referred \
to Transformer Secondary
/
(External Impedance)
= ( Correction to fx and1-.

3.1. Theoretical Relations

A transformer with an equivalent resistance of 1 .
(which is easily measurable) will result in a correct"
of approximately 1 ppm [5].
The effect of loading by the quadrature balar
network on either the inductive voltage divider *
the reference transformer is usually insignificr
At the limits of phase-angle range of : 37.8 mrad
min), the correction to ratio and phase angle ui t*
inductive voltage divider is ~ 1 ppm and < 4 /ui,
respectively; or, to the reference transformer. < <
ppm and < < 1 find.

As previously defined, the voltage ratio is

y,=^(cos e,+j sine,),

and
\v I
V
Vx = -j^-f- (cos 0X +j sin ex) = -**

3.3. Description of Components

m being the actual ratio of the inductive voltage di
vider. Corrections to the ratios and the phase angles
of the inductive voltage divider used are insignificant
and can be neglected. Then from figure 5,

The balance equation of the circuit becomes

Nxfx(cos 8,+j sin fl.) _
"/V,/,(cos 6X +j sin dx)

2rtB,

. rt}

and on separation and by neglecting second order

terms

All the components in this circuit are readily su

able commercially or accessible in most testing iii
oratories. In figure 5, R is a 010 kft, Kkir
continuously adjustable resistor with small readreactance. C is a 10 X 1 nF decade capacitor wri
dissipation factor of < 0.0003. Both R and C *rt
calibrated before and after assembly. The devi*>from nominal values were found to be within 5ft ^
5 pF, respectively.
The inductive voltage divider is a six-decade k
designed for optimum performance at 60 Hz.5 F
tests at higher frequencies (up to 1 kHz), it can em
be replaced by a unit of appropriate design, or r
application of appropriate loading corrections to "i
ratio-correction-factor and phase angle. Ho*tif
due to the simplicity of the circuitry and infreqBB
requests for calibrations at higher frequencies, a tr

Bx = ^kh (ft. + <> CR + d.^CR2) ~6, + <oCR

mN,f,
* Sine* maximum voltagr on this divider is limited to 130 V al 60 Hz. it t
across the test transformer rather than thr reference transformer whose
may exceed this value in sol

radians, and
mNJ, (c
/cos dx-rfOW + uCR sin 6r\
, _ mN,f,
cos 6,
)

For relative lagging phase angle, with the detector

and resistor, R, interchanged,
dx = 6, wCR radians, and
/x^+(/.-D+|(aC/?)
0-10 nF
3.2. Loading Effects
Since the input impedance of the inductive voltage
divider at 60 Hz is in excess of 0.8 Mfl (nearly 1.00
338-260

Figure 6.

Inductivt
Voltogc
Divider
G

Figure 7.

aorary setup is generally used at NBS. This consists

jf a high frequency divider, commercially available
lir capacitors, and an a-c decade resistor.
The input impedance of the inductive voltage di
vider was measured in a capacitive voltage divider
letup; see figure 7 for circuit diagram [10]. The voltige-ratio corrections and phase-angle errors were
.lit amed by a tracking method [11]. However, for
i well designed inductive voltage divider, these corections are usually < 0.2 ppm of input in ratio lin:arity deviation and < 8 /xrad of input in phase
leviation.
The detector circuit consists of a wave-analyzer used
is a tuned null indicator and a specially constructed
hielded transformer, shown in figure 6. This transormer is used to match impedances and to isolate
he measuring circuit from the detector circuit electro
statically. It consists of a high permeability toroidal
:ore, a 700-tum inner primary winding, two electrostatic
hields, a 3500-turn outer secondary winding, and an
verall Mu-metal case. The use of a toroid minimizes
he magnetic pickup from stray fields in the laboratory.
The polarity checking voltmeter is a 0-150 V, 3-in.
panel-type instrument of 3 percent accuracy. The
1 kfl resistor matches its internal resistance. The
oltmeter indicates the secondary voltage of the test
ransformer when the selector switch is in the
CALIBRATION" position.
Figures 8 and 9 show the exterior and interior of
le comparator. The cylindrical shield shown in
le interior view encloses the 10-kfl resistor.

3.4. Accuracy Determination

After the completion of the comparator, the caliration of each component was rechecked.
The accuracy of the completed comparator was
erified by the following procedures: (1) connecting
single voltage source to both "Vs" and "Fj" terlinals to obtain the zero corrections for C and R and
le unity correction for dial settings of the inductive
oltage divider; (2) calibrating several pairs of trans>rmers of the same ratio to reference transformers of
ppropriate ratios, thus determining the relative ratioi>rrection-factors and phase-angle errors with several
alues of secondary burden for each pair of transform339-261

ers; (3) comparing each pair of these calibrated trans

formers to each other with the comparator alternately
as reference and test units and with various combina
tions of previously used secondary burdens; and (4)
connecting a single voltage source to "Vx" terminals
and to the input of a calibrated inductive voltage
divider, whose output is connected to "V," terminals
in order to check the fractional ratio measurements.

load corrections. Such errors are estimated to k

0.5 ppm and 3 firad for each transformer.
Finally, one should add 3 times the standard dni
tions (0.3 ppm of ratio and 1 /irad) of the results <itained in repeated balances of the comparator to ik
combination of the systematic-error estimates.
Uncertainties to
ratio-correctionfactor

3.5. Transformer Testing Procedure

Unctrtimt*lo phis?

Source of errors

A reference transformer is selected so that N, 3 Nx.

The selector switch is set to "CHECK" position; and
the ratio, m, is set to equal Nx/N,. The secondaries
of the test and reference transformers are connected
to "Fx" and "V," terminals respectively. With the
transformers energized and the voltmeter switch de
pressed, the voltmeter indicates (V, VX)I2 if the po
larities of V, and Vx are correctly connected to the
comparator. [The polarity of Vx must be reversed if
voltmeter indicates {V,+ Vx)\2.\ The test proceeds
with the selector switch turned to "CALIBRATION."
Vx is now indicated on the voltmeter.
If Nx > N, by more than 10 percent, the test can still
be made, but the reference transformer must be con
nected to "Fj" terminal and the test transformer to
"Vg\ At the balanced condition, the ratio, m, will
indicate the reciprocal of the relative ratio-correctionfactors, and the relative phase-angle switch will show
the negative of the markings as shown in figures 6
and 8.

n
PPm
From assumptions made in
equations
0.5
Calibration of the inductive
voltage divider
.2
Obtaining secondary load cor
rections (S)
.5
Obtaining secondary load cor
.5
rections iX)
.5
Leads and contacts

n*

find

0.25

.04

.25

.25
.25

3
2

1.04

Sn

'

3 x standard deviations

1.02
0.90

7.1
3.0

Maximum error

1.9 ppm

10.1 JITld

i1
|J

4. Conclusion
3.6. Consideration of Errors
An examination of errors requires a careful appraisal
of all components of the measuring circuit as well as a
justification of estimations, even though in this case
one would expect most of the associated errors to be
negligible.
The first to be considered are those possible errors
concealed through approximation in developing the
expressions for obtaining the ratio-correction-factor
and phase-angle relations, where the effects of the
residual of the 10-kfl resistor and the dissipation factor
of the decade capacitor are ignored. Approximate
magnitudes of these errors can be obtained by adding
the estimated equivalent residual reactance across R
and estimated equivalent series resistance to C.
These effects were found to be less than 0.5 ppm for
ratio and 5 /ttrad for phase angle at the limits of the
phase angle range ( 38 mrad) at 60 Hz.
A second source of uncertainty is in the calibration
of the inductive voltage dividers. However, past
experience indicates that errors should not exceed
0.2 ppm of input in ratio linearity deviation and 2 fixad
of input in phase deviation [12].
Errors caused by the leads, switch contacts, and
imperfect shields are estimated to be < 0.5 ppm and
< 2 /irad.
The last important uncertainty that should be con
sidered is in obtaining the transformer secondary

A high accuracy inductive comparator has be*

developed for measuring the relative voltage ^
and phase angle of voltage transformers and let*
sets at power and audio frequencies. An ma
accuracy of < 2 ppm in ratio measurements 3n&-/irad in phase angle at 60 and 400 Hz is achif*
with a slight reduction in the accuracy attain
at higher audio frequencies.
The simple circuit does not require any sp
equipment. Only a few ranges of reference tri
formers are required, since fractional ratios fiwr
to 1/4 may be measured as readily as ratios near;"
This inductive comparator could replace ifsets now in general use. It is compact and poru
has a broader useful frequency range, and excre
the accuracy of many commercial sets presently J"
able by one or two orders of magnitude.
The resistive comparator will continue to be a
at NBS for calibration of voltage transformers
Hz at which frequency the operating voltage is uabove the maximum operating voltage of the indc
comparator.
The author acknowledges the valuable help
advice received from Mr. F. L. Hermach. 1
Electrical Instrument Section, NBS. Mr. R. J
contributed skillfully in constructing the "
comparator.

340-262

5. References
i] F. B. Silsbee and F. M. Defandorf, A transformer method for
measuring high alternating voltages and its comparison with
an absolute electrometer, J. Res. NBS 20, 317-336 (1938)
RP1079.
2] W. K. Clothier and L. Medina, The absolute calibration of
voltage transformers, Proc. Institution of Electrical Engi
neers, 104A, 204-214 (1957).
3] F. K. Harris. W. C. Sze, N. L. Kusters. O. Petersons, W. J. M.
Moore, An international comparison of voltage transformer
calibrations to 350 kV, IEEE Trans, on Communication &
Electronics No. 70. pp. 13-19 (1964).
4] Private communication from F. K. Harris, Chief, Absolute
Electrical Measurements Section, NBS.
F. K. Harris, Electrical Measurements (John Wiley and Sons,
Inc.. New York. N.Y.) pp. 410-411. 565-569 and 611-612
(1951).

[6] W. Duddell and T. Mather, Improvements in non-inductive

resistance. British Patent No. 5171 (1901).
[7] F. B. Silsbee, A shielded resistor for voltage transformer testing.
Bull. BS 20, 489-514 (1926).
[8] Private communication from F. K. Harris and B. L. Dunfee,
NBS.
[9] H. B. Brooks and A. B. Lewis, Improved continuously variable
self and mutual inductor, J. Res. NBS 19, 493 (1937) RP1040.
[10] M. C. McGregor, J. F. Hersh, R. D. Cutkosky, F. K. Harris
and F. R. Kotter, New apparatus at NBS for absolute capaci
tance measurement, IRE Trans., Instr., 1-7, 253-261 (1958).
[11] NBS Technical News Bulletin, 49, 1 (1965).
[12] W. C. Sze, A. F. Dunn, and T. L. Zapf, An international com
parison of inductive voltage divider calibrations at 400 and
1000 hertz, IEEE Trans, on Instr. and Meas. IM-14, No. 3,
124-131 (1965).

341-263

(Paper 69C4-206)

An International Comparison of
Voltage-Transformer Calibrations to
350 Kv
F. K. HARRIS
FELLOW IEEE

W. C. SZE
SENIOR MEMBER IEEE

N. L. KUSTERS
MEMBER IEEE

O. PETERSONS
MEMBER IEEE

W. J. M. MOORE
SENIOR MEMBER IEEE
IN JUNE 1960, the Hydro-Electric
I Power Commission of Ontario inquired
whether the United States National Bu
reau of Standards (NBS) in Washington
could measure for them the ratio and
phase-angle errors of a 350-kv voltage
transformer then being built to their
order. This request was also made at
about the same time of the National Re
search Council of Canada (NRC) in
Ottawa. Neither laboratory had, at that
time, made calibration measurements on
instrument transformers at this voltage
level, but both had given the problem
considerable thought; and it appeared
that the request by Ontario Hydro
afforded a unique opportunity not only to
examine some of the problems involved in
voltage-ratio measurements at 350 kv,
but also to compare measurement results
between the two national laboratories.
The transformer, as first built, de
veloped troubles during shipment that re
quired structural modifications; but
finally, early in 1962, the completed trans
former arrived in the High-Voltage Lab
oratory of NRC for calibration tests.
From Ottawa it was carried by truck
to Washington for further measure
ments in the High-Voltage Labora
tory of NBS, and then returned to
NRC for repeat measurements. In this
way, it was believed that any changes re
sulting from its shipment could be de
tected, and perhaps separated out from
the interlaboratory comparison.
The transformer is of unusual construc
tion and, since this had considerable bear
ing on the test results, it will be described
in some detail. Following this, the NRC

and NBS tests will be separately dis

cussed, and finally the results from the
two laboratories will be compared.
Description of the Transformer
The transformer is a cascade type made
by the Canadian General Electric Com
pany.1 It consists of three sections, each
with a separate core and primary winding
encased in a porcelain cylinder. The
primary voltage is distributed equally
among the series-connected primary wind
ings of the three sections. To insure
equal voltage distribution among the
sections and to provide a low-impedance
path for the load current, the cores are
coupled by low-voltage windings. The
secondary winding of the transformer is on
the section nearest ground. Voltage
ratios of 4,000: 1, 3,000: 1, and 2,000: 1 are

Paper 63-992, recommended by the IEEE Funda

mental Electrical Standards Committee and
approved by the IEEE Technical Operations
Committee (or presentation at the IEEE Summer
General Meeting and Nuclear Radiation Effects
Conference. Toronto, Ont., Canada. June 16-21,
1963. Manuscript submitted March 18. 1963,
made available for printing May 7. 1963.
F. K. Harris and W. C. Szb are with the National
Bureau of Standards, Washington, D. C; and
N. L. Kustbrs, O. Petersons, and W. J. M.
Moore are with the National Research Council of
Canada, Ottawa, Ont., Canada.
342-1

available at the rated primary voltage

of 350 kv, by means of taps on the
secondary winding. Additional ratios i
1,200: 1 and 1,000: 1 are obtained by ranecting the primary voltage across the
two lower sections of the transformer
The 1 ,200 : 1 ratio involves a nomntepa!
number of secondary turns, that is, one
turn links only part of the core.
Shielding
Since the case of this transformer is
made up of the porcelain cylinders that
insulate its three sections, the windingsar;
not electrically shielded ; and the capadtive currents in the windings will depsxi
on the location of nearby objects and tit
potential difference between them and tie
transformer section. The internal m
pedance of the transformer being hizh
appreciable voltage drops may result fron
these capacitive currents, so that tie
voltage ratio and phase angle of the trareformer are dependent on its emvonmertt.
and will change depending on the po
tential and distance of neighboring ob
jects.
To reduce this proximity effect, shield;
were constructed and were connected v.
the ground and high-voltage termhi
as shown in Fig. 1. Both shields wee
used for the three highest ratios, whit
only the lower shield was used for tie
1,200:1 and 1,000:1 ratios. It will be
seen from the result that these shield

Fig. !. Transformer errors at measured by

high-voltage capacitance bridge

Fig. 3. Transformer errors as measured by

capacilive voltage divider

make the transformer less sensitive to

the location of external objects, but they
do increase capacitance currents in the
transformer windings, so that the ratio
and phase angle of the transformer are
substantially different in the shielded and
unshielded condition.
NRC Tests
Calibration Methods
At NRC the transformer was calibrated
with a new high-voltage capacitanceratio bridge, based on the current-com
parator principle, and two gas-dielectric
capacitors." The high-voltage capacitor
was a commercial 50-pf (picofarad) com
pressed-gas unit rated at 500 kv, while
the low-voltage capacitor was of parallelplate construction with air dielectric.
The capacitance ratio of the two capaci
tors is first determined with the same volt
age applied to both, the current ratio,
measured by a current comparator, being
equal to the ratio of capacitances. The
primary voltage is then applied to one
capacitor and the secondary voltage to
the other, and the ratio of currents is again
measured. This ratio is the product of the
I voltage ratio and the capacitance ratio.
The bridge is direct reading in trans
former errors as defined by the vector
diagram of Fig. 2 and the following equa
tion:
= K.(l-<,Kl-J/3)
(1)
n
where Vp is primary voltage, Vs is secondP ary voltage, n is nominal ratio, a is in! phase error, and (3 is quadrature error.
The voltage applied to the capacitors
during their ratio determination may be
' very different from that of the second
J measurement, and the stability of the ca315-411 O - 68 - 23

pacitors with applied voltage must be

known. An investigation covering this
point has been reported previously.4
A capacitive voltage divider such as
that used by Clothier and Medina was
also available.' In this divider the ratio
of capacitances is equal to the nominal
ratio of the transformer. To permit the
use of capacitors with gas dielectric, the
capacitance of the low-voltage arm must
be limited to a few thousand pf. For the
high-voltage arm a 1-pf capacitor rated
at 80 kv was available. Thus up to 80
kv, it was possible to calibrate the trans
former by two methods. The balancing
circuitry associated with the capacitive
divider is direct reading in transformer
errors ; and the balance condition may be
represented by the diagram of Fig. 3 and
the following equation:

(2)
n - vAi-je)/a+a)
The stability of the capacitors with volt
age is equally important in both methods.
Proximity Effect
While the influence of nearby objects
on the transformer errors was reduced by
the addition of the shields, these shields
shifted the in-phase error in the negative
direction by 1 ,500 ppm (parts per million)
and the quadrature error in the positive
direction by 730 ppm at the top three
ratios. The calibration results tabulated
later in this paper are applicable only
when the shields are used, and with ample
clearance around the transformer.

To illustrate the severity of the prox

imity effect, the changes in transformer
errors corresponding to the introduction
of certain objects in the vicinity of the
transformer were recorded. One of these
objects was a piece of tubing suspended
from the high-voltage bus bar, approxi
mately 4 feet from the transformer. The
other was a high-voltage capacitor
approximately 8 feet from the trans
former. The outer part of the capacitor
is a high-voltage electrode, and could
either be grounded or excited from the
high-voltage bus. These objects are
shown in Fig. 4. The variations in
transformer errors (Aa, A0) at a 2,000:1
ratio are recorded in Table I. The values
given represent transformer error with
the object present minus the error with
the object removed. It will be noted
that the largest variation occurs from
proximity of an object at high voltage.
The shields reduce this effect by a factor
of three. The shift in transformer errors
is much less when the object is grounded,
and the shields are not as effective in re
ducing this shift.
Comparison of Test Methods and
Transformer Stability
A limited number of calibration points
were remeasured at NRC following the
return of the transformer from NBS.
These are summarized in Table II, to
gether with data that are typical of the
results obtained using the two methods
of calibration described above.

Tabic II

T.bU I

a (Ppm)

Transformer Transformer
Without Shield! With Shields
Off Ha
Aa
(Ppm) (Ppm) (Ppm) (Ppm)
Tubing suspended
-187. + 95. -53.. .+29
Capacitor energized
-273. + 123 -88. + 38
Capacitor (rounded . .+ 24. - 10. +14. . - 9

It will be noted that the two test

methods yield results that are in complete
agreement, that is, their differences are
random and well within the estimated
accuracy (10 ppm) of the two methods.
It will also be noted that there was a
very marked shift in the 1,200:1 ratio
values and a smaller but definite shift in
the other ratios, after the transformer had
travelled to Washington and back to
Ottawa. Additional calibration points,
not included in the table, indicated that
differences between the measured values
obtained before and after the NBS tests
were constant over the whole voltage
range of the transformer.

Secondary
Volte
1,000:1.
1.200:1.
2.000:1.
3.000:1.
4.000:1.

80 .... ...+1,202
88.7.... ... -1.114
8*3
40 ....
42
28.7....
992
20

+ 1.195
-1.289
- 912
- 982
-1.010

^-(l-a-i
(3)
nV,
and the balance equation of the network
shown in Fig. 5,

(Before)

+ 7
+175
+ 19
+ 20
+ 18
t (Ppm)
- 1
- 23
- 2
- 4
- 5

+1.206..
-1.111..
- 894..
- 944..
- 992..

Method 112"
-4
-J
+1
+2
0

-4
+ 341
+ 344
.+ 340.
SO .
1.000:1.
.4
+ 254
+ 235.
. + 231 .
.88.7.
1.200:1.
.

-1,175
-1.177.
. -1.177.
.40 .
2.000:1.
-1,141
-1,141.
-1.145.
.28.7.
3,000:1.
-1
-1.068
-1.059.
-1,061.
.20 .
4.000:1.
In using Method 2. the 1 pf capacitor (shown in F.g. 4) ... connertecl to
therefore contributed to the "proximity effect." Hence, to make a vahd comparison
S
nd Method 2 it was necessary that volta,e be Impressed on this capacitor for the Method 1 i
^^^'*^?ZZ* the data of Table II are not strictly equivalent to thos. of 1
for which the 1-pf capacitor was grounded.

'i+s = i
we have
nV,(.\-a-jff)-ju>C,

and on separation, this becomes

NBS Tests
At the National Bureau of Standards it
was decided to use the method of Clothier
and Medina, and a 1-pf free-air capacitor,
rated at 350 kv, was constructed for use in
the capacitive voltage divider. Its con
struction and characteristics are described
in a companion paper.* The low side of
the divider was made up of one to four
1,000-pf air capacitors of special construc
tion, depending on the nominal ratio of
the transformer. The circuit is shown
schematically in Fig. 5, but simplified by
not indicating the shielding in detail.
Balance was achieved by adjusting the
tap points on inductive voltage dividers
connected across the secondary winding
of the transformer. Commercial 6-decade
inductive voltage dividers designed for
60-cycle operation were used. Positive or
negative phase-angle errors could be com
pensated by moving the tap point to one
side or the other of the indicated ground
point of the divider, and so reversing the
voltage of the branch with respect to the
secondary voltage of the transformer.
Using the following expression for the
relation between primary and secondary
voltage of the transformer

Before -Alter

and
0--N,/(umRQ)
where A\ and A5 are the fractions of V, im
pressed on Cj and R, respectively, by the
voltage dividers. Since this test network
was not expected to become a permanent
laboratory setup, it was not considered
worthwhile to make it direct reading. It
will be noted that the in-phase balance
expression contains a term representing
the equivalent shunt capacitance across
the resistor R used for quadrature bal
ance. In any practical case this would
never be more than a few ppm and, with

Flg. 5. NBS test circuit

344-3

-s

the resistance element and shielding used,

it was found to be negligible. Thus the
capacitance ratio becomes CV'G and,
since Ci is a free-air capacitor whose vaht
is subject to ambient temperature, pres
sure, and humidity, this capacitance ratio
was determined immediately before ate
after any series of transformer ratio
measurements, and the mean of the ratio
values was used in computing the trans
former errors. If the two capacitance
ratio values were markedly different, in
dicating a considerable drift in the vahe
of Ci, the run was discarded. The ratio
Cj/Ci was established in three steps with
a 10:1 transformer-ratio-arm bridge
using equipment very similar to thit
described earlier by McGregor et al.7
As has already been pointed out in dis
cussing the NRC test methods, the capaci
tance ratio is determined at a low voltage
whereas a very different voltage is re
pressed on one of the capacitors in voltafftdivider operation. It is therefore im
portant that any change of capacitance
with voltage be known, so that an apprc-

Fig. 6. Voltejc
to NBS c.picitor

AO

100

ISO

200

250

priate correction to the ratio may be ap

plied. This was investigated for the 1-pf
350-kv capacitor, and is separately re
ported.' The corrections arising from
impressed voltage are shown in Fig. 6,
capacitance increasing with increasing
voltage. Measurements on the 1,000-pf
low-voltage capacitors indicated that they
change by not more than 1 part in 107
over the voltage range of interest.'
Another correction that must be taken
into account results from the burden im
posed on the transformer by the test-cir
cuit elements connected across its second
ary winding. This correction can be com
puted in terms of the impedance of the
circuit elements, and the burden calibra
tion of the transformer.
As to proximity effect, the shield sys
tem was used in the same way as at NRC
and a cleared area with a radius of 1 1 feet
was in all cases maintained around the
transformer at ground level. In the
space above the plane of ground shield,
the clearance was nearly double this value,
except for the high-voltage conductor
between the transformer and the highvoltage capacitor. The test arrangement
is shown in Fig. 7.
Results of the tests in the two national
laboratories will be summarized below,
using values which are generally the
average of several determinations. First
however, it will be instructive to look at
- the maximum spread in the NBS values
used for these averages, to show the com
bined uncertainties and instabilities of the
measuring method and equipment, and of
the transformer. This is given in Table
III. While most of this spread is believed
to represent uncertainties of variations in
balancing techniques, instabilities in
measuring-circuit components, or errors
in setting test burdens or voltages,
-there is some reason to suspect small varia
tions in the transformer itself. Precau
tions were regularly taken to prevent core
magnetization that could significantly
affect ratio and phase-angle values at

_i300

390I

the lower test voltages; but on one oc

casion the transformer was inadvertently
magnetized, and its subsequent stabiliza
tion was quite tedious. It is believed
that day-to-day variations of a few ppm
in transformer ratio were seen.
Comparison of NRC and NBS Test
Results
In Table IV are listed the average ratio
and phase-angle errors in ppm, observed
for the transformer in the two national
laboratories for various test conditions.
All values were taken at a 60-cycle fre
quency. The NRC data tabulated are
those taken before the NBS test.
As has been noted already, the trans
former errors changed during the series
of tests, and it is supposed that these
shifts occurred as a result of road shocks
during shipment of the transformer be
tween the two laboratories. In Table V,
the differences of Table II between the
transformer errors before and after its
trip to Washington are compared with
the average differences from Table IV

345-4

Tabl. Ill

Nominal
R.lio

Number of
Determination!

1.000:1
1.200:1
2.000:1
3.000:1

5
r
12
IS

Mu. num Spread

(Pjm)
J.
. 6.
. .13.
. .17.
. .23.

4
8
8
8

between the initial NRC values and the

NBS values.
Only for the 1,000: 1 ratio are the differ
ences between the in-phase errors found
by the two laboratories greater than the
observed differences in the NRC values
before and after shipment to NBS. Since
these measurements required that the
upper shield ring of Fig. 1 be removed,
with a consequent increase in the proxim
ity effect, it may well be that this differ
ence results in part from a minor differ
ence in the arrangement of the highvoltage bus-bar and transformer connec
tion in the two laboratories. The same
statement concerning shield and lead
arrangement could be made about the
1,200: 1 ratio, but here any possible differ
ences from this cause are masked by the
very large "transportation" shift in the
transformer itself. While it is rather
suprising that such a large change in
ratio would occur, it seems reasonable that
the behavior of this ratio could be very
different from the others since one turn of
this secondary winding links only a por
tion of the core.
It may be noted from Table IV that the
interlaboratory differences in quadrature
errors appear to show a trend with chang
ing voltage for the 1,000:1 and 1,200:1

Conclusions

Table IV
HB8

KRC
Nominal Secondary
VolM
Ratio
(SO
0
68
80
100
120
1,000:1... 130
SO
60
69
80
100
120
130

...
...
...
...
...
...
...
...
...
...
...
...
...
...

SO
60
69
80
100
120
1,200:1. . . 130
SO
60
69
80
100
120
130

...
...
...
...
...
...
...
...
...
...
...
...
...
...

[ 50
75
100
125
150
50
75
100
125
150
33.3
50
88.7
83 3
100
116.7
33.3
50
88.7
83.3
100
116.7
25 .
37.5.
50 .
62.5.
75 .
87.5.
25 .
37.5.
50 .
62.5.
75 .
87 5.

386. .
371. .
360. .
348. .
334. .
326. .
323. .
67 . .
51. .
43. .
29. .
18. .
8. .
6..
Atc.
-1.108. . . 241.. . . -1.236. . . 270.
263.
.
-1,118. . . 235. . . . -1 .248. . .
-1,124 . . . 231. . -1,256. . . 260 .
229. . . . -1,265. . . 255
-1 .130. .
-1.138. . . 231 .. . . -1 .274. . . 252. .
-1.140. . . 241. r. . -1 .277. . . 251. .
-1,140. . . 249. . . . -1.277. . . 250. .
-1.212. . . - IS... . -1.338. . . + 10. .
-1.224...- 20. . . -1.354... + 2..
0..
-1,232. . . - 24. . . -1,365. . .
3. .
-1,241...- 25. . . . -1,374 . . .
7. .
-1.249...- 23. . . -1,385. . .
-1,252...- 14. . . .-1.389... - 10. .
-1.252. . . - 9... -1,388... - 12. .
Ays.
848 -1.178.
845 -1.174.
780 -1.175
7H.-, -1.170.
781. -1.171.
749. -1.164.
727 -1.169.
726 -1.160.
718. -1.165.
715 -1.158.
968 -1.505.
959. -1.495.
907 -1.505.
898. -1.492.
871. -1.501.
862 -1,488.
848 -1,498.
-1,484.
836 . -1.495.
-1 482.
Ave.
1.142. .
906
897. . -1.142.
819
1 .142. .
840. . -1,138.
815
1 .135. .
805. . -1.135.
1.132.
794
783. . -1.130.
783
1.129. .
773 . -1.128.
1 .127. .
782
773. . -1.128.
96(1
1,276
.
948. . -1 ,276.
901
1.276. .
892. -1.272
866
1,272.
857 . . -1.268.
846
1.268. .
833.. -1,264.
833
265. .
822 . . -1,262.
832 . -1.261 . .
822. . -1.262.
Ave.
954 . -1.060. .
955. . -1,062.
811. -1.080 .
910. . -1,061.
882. -1,059 .
882. . -1.058.
S62 -1.057. .
86a. . -1,054
852 -1.055. .
850. . -1.056.
s:,i . -1 .051. .
850. -1,056.
993 -1.154. .
994. . -1 .157.
951. -1 ,154.
946 . . -1.155.
922 -1,154
920 . -1.152.
903 -1 .151. ,
898 . -1.150.
892 -1.148.
-1.150.
890 -1.147..
-1.149.
Ave.
1.208. . .
1.201. . .
1.196. . .
1,192. . .
1,186. . .
1 ,184. . .
1.185. . .
1.110. . .
1.108...
1,100...
1,098. . .
1,091. . .
1,089. . .
1.088...

373. . .
358. . .
349 .. .
340. . .
330. . . .
325. . .
326. . .
55. . . .
42...
33. . . .
23 . . .
13. . . .
8... .
9... .

1,227...
1 , 222 . . .
1.218...
1.214...
1,208...
1,206...
1,207...
1.131...
1.123...
1.121...
1.115...
1.109...
1.107...
1.109...

NRC- NBS
A*
ia

Burden

.- 19. -13
.-21. . -13
. - 22. . -11
. - 22. .- 8 . . open circuit
. - 22. .- 4
. - 22. . - 1
.- 22. .+ 3
.-21. . -12'
-17. -
.-21. -10 . . 882 ohms, unity
power factor
.- 17. . - 8
. - 18. . - 5
0
. - 18.
. - 21. +
- 20. . - 6
. +128. .-29'
+ 130. . -28
+ 132. . -29
+ 135. . -26 . .open circuit
+ 136. . -21
. +137. .-10
. +137. - '<
. +126. -25*
.+130. . -22
+ 133. . -24 , . . 882 ohms, unity
power (actor
+ 133. . -22
+ 136. .-16
.+137. .- 4
.+136. .+ 3
+ 134. . . -18
.+ 41
+5|
.+ 7 .open circuit
.+ 9
+ 7,
.+101
. +13 \ . 882 ohms, unity
power factor
. +13 |
.+ 10. .+14 I
.+ 10. .+13J
+ 6... +10

Within the past 3 years, this is the

second international comparison of trans
former calibrations in which these two
national laboratories have partkipateu.
In the previous comparison, the agree
ment between NRC and NBS results a:
ratios of 1 ,000 : 1 and higher was between
100 and 300 ppm, based in both labora
tories on resistive voltage dividers whose
voltage coefficients of ratio could be
evaluated only with difficulty and to a
rough approximation.* In the present
comparison the results agree between the
laboratories within 20 ppm at ratios to
4,000: 1 and voltages to 350 kv, except in
the 1,200:1 ratio where the transformer
instability made the measurements less
meaningful. It may therefore be said
that an improvement of an order of mag
nitude has been achieved, and that ratio
and phase-angle measurements on voltage
transformers are now reliable in either
laboratory within 20 ppm and 10 microradians (2 arc seconds), at voltages to
350 kv.
Considering the ratio instability of the
transformer and its sensitivity to neigh
boring objects, it appears that closer
interlaboratory agreement would not be
possible with a transformer of the type
used.
References

. 882 ohms, unity

power factor

1. Cascade-Type Potential
G. CamUU. General Electric Rtvirm. Scbesecta^r
N. Y.. vol. 39, Feb. 1936, pp. 95-09.
2. A TRANSPORMBR-RaTIO-ARM BKIPOE Ft*
Hioh-Voltage Capacitance MsASumainxTS SL. Kusters, O. Petersons. IEEE Transactions w
Communication and Electronics (Paper 6J-n!
3. The Development op th
pabatoe, A Hioh Accuracy A-C Ratio Meastt
ino Device, P. H. MUjanic, If. L. Kusters, W. J.
M. Moore. A1EE Transactions, pt. I (Cummb
lion and Electronics), vol. 81, Nov. 1963, pe

!!
i
.'.
3
1
2 . . 882 ohms, unity
power factor
I
2
2
+ 2. . . - 1

4. Trb Voltage Coefficients op

Capacitors, If. L. Ktuters, O. Petersc
Transactions on Communication and
{Paper 63-169).
5. The Absolute Calibration op Voltasi
Transformers, W. K. Clothier, L.
Proceedings, Institution of Electrical Enfiseen
London, England, vol. 104A, June 1957. pp. S04-:i
6. An Experimental 350 Kv-1-Ptcofaxh
Air Capacitor, A. B. Peterson. IEEE Trsx-

+ 10. . . + 2

Table V
ratios. No such trend is observable in
the in-phase error, nor is such a trend
apparent in any of the three higher ratios.
No explanation can be suggested for the
quadrature peculiarity noted in the two
lower ratios; but it may be concluded with
considerable assurance that any voltage
dependence of the capacitors was prop
erly allowed for by both laboratories.

A slightly better internal consistency

is to be noted in the NRC results of
Table IV, which can probably be ex
plained on the basis of better stability in
the compressed-gas high-voltage capacitor
(used by NRC) than in the free-air capacftor (used by NBS). The spread of NBS
measurements given in Table III is further
evidence toward this conclusion.
346-5

Nominal
Ratio

NRC (Before) KRC (After)

A0
ac

1 .000:1
1.200:1
2,000:1.
3,000:1. .
4,000:1. .

.+ 7..
+ 175. .
..+ 19..
..+ 20..
..+ 18..

IfRC (B.efareHIIS
Act
i

. - 1 . . ... - 20 - 1
. -23 . . . +134 ..-a
. - 2 ...+ 6..
.- 4
..+ i
..- i

actions on Communication and Electronics (Paper

63.926).
7. New Apparatus at the National Bureau
op Standards for Absolute Capacitance
Measurement, M. C. McGregor, J. F. Hersh,
R. D. Cutkonky, F. K. Harris, F. K. Kotter. IRE
Transactions on Instrumtntalion, vol, 1-7. Dec.
1958, pp. 253-81.
8. A Report on Canadian Participation in an
International Comparison op Instrument
Transformer Calibrations-1959, If. L. Kusters,
W. J. M. Moore, O. Petersons, C. A. E. Uhlig.
Report ERB-S71, National Research Council of
Canada. Radio and Electrical Engineering Division,
Ottawa, Ont., Canada, 1960.

Discussion
K. M. Mitchell (The Hydro-Electric
Power Commission of Ontario, Toronto,
Ont., Canada): In this paper the authors
have pointed out the limitations of this
transformer with respect to its ratio
stability and the influence of neighboring
objects on its errors. With measuring
techniques sensitive to a few ppm, these
effects are clearly evident. This suggests
the possibility that other influences on
the errors of the transformer, such as
temperature, frequency, and waveform,
may also be significant in terms of these
references. However, the paper does not
liscuss these influences. It would be
jppreciated if the authors could comment
yn their magnitude, and outline what pre
cautions, if any, were necessary to minimize
their effect on the intereomparison.
The proximity effect which the authors
joint out as a characteristic of any un
shielded transformer has been shown in
Table I to be an effect of practical signifi
cance, at least with respect to the cascadeLransformer design used in this intercomparison. That such could be the case,
[ do not think has been generally appre
ciated. Considering the possible signifi
cance of the proximity effect and the
rrowing numbers of potential transformers
>f the cascade type in service on electric
itility systems, it would seem desirable
:hat any revisions to national standards
or instrument transformers should recoglize this characteristic of unshielded or
mrtially shielded transformers and possibly
letail acceptable magnitudes for it. This
nfers the availability of a test method to
establish compliance of a transformer
vith the requirement. Have the authors
iny suggestions for the formulation of a
iuitable test method?
With reference to the terms or and 0
vhich have been used to express the in>hase error and quadrature error, re
flectively, three different equations expressng the relationship between these terms
ind the primary and secondary voltage
ire given (see equations 1-3). Two of
hese are illustrated by the vector diagrams
>f Figs. 2 and 3. In the first of these
xpressions (Fig. 2), the in-phase error,
r, is expressed as a fraction of the true
econdary voltage and /3 = tan y, where y
s the phase-angle error. In the second
xpression (Fig. 3), although 0 may still
>e defined as tan y, a is expressed as a
raction of the nominal secondary voltage,
hat is, VT/n. In the third expression
equation 3),

nV,
a is the in-phase error expressed as a fraction
of the true secondary voltage, but 0 must
be denned as (I or) tan y. It is appre
ciated that these differences in definition
of a and /3 arise from the balance equations
of the several test methods employed by
the authors and that for small values of
a and 0 they may not be significant.
However, the errors of this transformer
on the higher ratios are such that approxi
mations of the order of 1 to l'/i ppm are
involved and differences of this magnitude
are tabulated. I should like to ask whether
in the preparation of Tables IV and V a
common definition of a and f) has been used
or whether the differences in or and 0
tabulated here include differences in the
definition of these quantities?
J. M. Vanderleck (The Hydro-Electric
Power Commission of Ontario, Toronto,
Ont., Canada): It is gratifying to know
that the results agree between the two
laboratories within 20 ppm. Phase-angle
accuracy of 20 ppm is desired for coronaloss measurements of extra-high-voltage
test lines in order to limit the phase-angle
error owing to the voltage transformers
to an acceptable value of 60 watts per
3-phase mile (60 watts is based on a charg
ing kilovolt-ampere of 3,000 and phase
angle of 20 ppm).
The temperature
influence on phase angle could be important
since corona-loss measurements may be
made from 0 to 95 degrees Fahrenheit
ambient.
Other instances wherein instrument
voltage-transformer accuracy is par
ticularly significant are as follows:
1. Revenue metering of high-voltage
circuits.
2. Acceptance tests on voltage trans
formers.
In the first case, it is desirable to know
the error owing to the group of installed
instrument transformers within 0.1% (1,000
ppm). To achieve this result, the errors
of each of the five or six instrument trans
formers installed in service should be
measured within 300 ppm. Thus, it is
desirable that the errors of the referencestandard voltage transformer be known
and certified by a national laboratory
within 100 ppm.
A similar accuracy (100 ppm) is useful
to minimize the doubt whether a voltage
transformer conforms with purchase speci
fications during acceptance tests. An over
all testing accuracy of 300 ppm can be
achieved in practice if the accuracy of the
reference-standard transformer is 100 ppm.
An example will illustrate the extent of
doubt using these data.
Suppose accuracy class 0.3 is specified
for the voltage transformer and the round
ing off method recommended by American
Society for Testing and Materials' is
stipulated, then true radio correction
factor (RCF) of 1.0035 indicates non
conformance with the accuracy specifica
tion, and any lesser value indicates con
formance. (In this example, a noncritical
phase angle is assumed, for example, 7
minutes.) With a testing accuracy of
347-6

300 ppm, it is uncertain whether test data

will indicate conformance or nonconform
ance when the true RCF ranges from
1.00320 to 1.00379.
The foregoing indicates that good use
can be made of the accuracy provided by
the national laboratories.
Reference
1. Designating Significant Placrs in Specpied Limitino Valubs. ASTM Designation E2960T, American Society for Testing and Materials,
Philadelphia, Pa.
F. K. Harris, W. C. Sze, N. L. Kusters,
O. Petersons, and W. J. M. Moore: As
Mr. Mitchell has pointed out, the present
report shows in a very striking fashion the
degree to which capacitance currents to
nearby objects can affect the ratio and
phase-angle errors of a voltage transformer.
We agree with him that such effects may
be significantly larger in a cascade design
than in a one-stage configuration. However,
since cascade designs lead to economies
of space and insulation at high voltages,
it is probable that they will continue to
find favor with instrument-transformer
engineers. Thus it would certainly seem
appropriate that proximity errors be
considered in future revisions of the trans
former standard. The data presented in
Table I were intended simply to illustrate
the performance changes that result from
nearby objects at ground or high potential.
A serious study of proximity effects on
transformers of various designs should
precede any attempt to describe a test
method for inclusion in a transformer
standard.
Mr. Mitchell is quite correct in assuming
that temperature, frequency, and waveform
could also influence our results, and we
regret that we do not have very much
information for him on the magnitude of
these influences. In both laboratories
tuned detectors with negligible harmonic
response were used, and no anomalous
behavior was observed in the presence of
known waveform distortion in one of the
laboratories at the highest voltage. Meas
urements were made on two ratios at fre
quencies that differed somewhat from
60 cps. In each instance the change in
ratio error amounted to about 200 ppm
and in quadrature error about 25 ppm for
a frequency difference of 2 cps. Data on
temperature effects are also meager. In
each laboratory the temperature was
controlled and the working area was shared
with other groups, so that the determination
of a temperature coefficient was not pos
sible. The temperature difference between
the two laboratories was approximately
2 C (degrees centigrade), the NRC labora
tory being maintained at 21 C and the
NBS laboratory at 23 C. On one occasion
when laboratory temperature controls failed,
measurements at the three highest ratios
were made at a temperature that differed
by approximately 2 degrees from the usual
value. It was observed that the phaseangle error was changed in a negative
direction by approximately 0.8% of its
value (that is, about 8 ppm in the quad
rature error), and no significant change
was observed in the ratio error. One
would of course anticipate that the largest
temperature effect would be on the quad
rature error, contributed by the product

of the quadrature component of the excita

tion current (capacitive for these ratios)
and the change with temperature in the
primary winding resistance. Neither
laboratory is really prepared to determine
the temperature coefficients of transformer
errors, since the units we test are generally
used as reference standards under normal
laboratory conditions.
The in-phase (a) and quadrature (0)
error terms tabulated are denned differently
in the different methods; and approxima
tions amounting to a part in a million are
used, as Mr. Mitchell has stated. In view
of the fact that uncertainties in our measure

ments and the observed transformer in

stability and proximity effects were all
substantially larger than this figure, it
did not seem important to reduce the
approximations of our formulas to a com
mon base, although we did report our data
to the nearest ppm to give a complete
picture of their statistical scatter.
The degradation of accuracy at successive
echelons of measurement has been the
subject of much discussion and some argu
ment in recent years. We agree with
Mr. Vanderleck that at present it is reason
able to require an accuracy of 100 ppm
in the assignment of ratio and phase-angle

values for reference transformers calibrated

at a national laboratory, although then
are already some instances in which better
accuracy has been requested. It is indeed
comforting to know our two laboratories
have a substantial margin of accuracy u
voltage-transformer calibration beyond the
100-ppm requirement. If, however, as
Mr. Vanderleck suggests, the transforms
is to be used over a 50 C temperature
range for corona-loss measurements where
quadrature errors of 20 ppm begin to be
significant, there should certainly be i
determination of temperature influence os
its calibration.

A reprint The
fromInstitute
COMMUNICATION
ELECTRONICS,
published by
of Electrical andAND
Electronics
Engineers, Inc.
Copyright
1964,
and
reprinted
bv
permission
of
the
copyright
The Institute assumes no responsibility for statements and opinionsowner
made by
contributors.
Printed in the United States of America
January 1964 issue

348-7

The Precision Measurement of Transformer Ratios*

R. D. CUTKOSKYf and J. Q. SHIELDSf
I. Introduction
A CONVENIENT and accurate method for cali
brating a given capacitor in terms of another ca
pacitor having a value ten times or one tenth
the value of the first is by means of a precise, stable
ransformer with a ten-to-one ratio and low equivalent
ieries impedances. 1,1 Such a measurement requires a
* Received by the PGI, June 24, 1960. Presented at the 1960 Consrence on Standards and Electronic Measurements as paper 5-4.
t National Bureau of Standards, Washington, D. C.
1 A. M. Thompson, "The precise measurement of small capacimces," IRE Trans, on Instrumentation, vol. PGI-7, pp 245-253;
>ecember, 1958.
M. C. McGregor, J. F. Hersh, R. D. Cutkosky, F. K. Harris, and
. R. Kotter, "New apparatus at the National Bureau of Standards
>r absolute capacitance measurement," IRE Trans, on Instruentation, vol. PGI-7, pp. 253-261 ; December, 1958.

knowledge of the open-circuit voltage ratio and equiva

lent series impedances of the transformer.
One method for measuring a nominally ten-to-one
transformer ratio has been described by Thompson.*
This method consists of determining the relative values
of eleven nominally equal three-terminal capacitors by
simple substitution and then setting up a bridge with
the circuit of Fig. 1 with ten of the capacitors in paral
lel on the low voltage winding of the transformer and
the eleventh capacitor on the high voltage winding. The
bridge is balanced by means of a small variable admit
tance in parallel with one arm of the bridge.
This series of measurements, when suitable correc* A. M. Thompson, private communication; 1957.

349-243

Fig. 1 Basic transformer bridge circuit.

lions are applied for the effect of finite lead impedances

and load admittances, yields a value for the open-circuit
transformer ratio.
If we are to obtain reliable ratio measurements, one
requirement is that all of the capacitances involved
drift at nearly the same rate. This can be achieved by
constructing them to be identical and mounting them
in good thermal contact with each other.
It can be shown that if eleven nominally equal threeterminal capacitors are mounted in this fashion with
their detector leads connected together, and if eleven
balances are made with the line lead of each capacitor
in turn on the high-voltage side of a 10:1 transformer
and the other ten line leads on the low-voltage side,
one may obtain directly the voltage ratio at the points
of interconnection independent of mutual coupling and
ground impedances within the capacitor housing. The
following section will show in detail how this is accom
plished in the more general case of a transformer with a
nominally m:n ratio. The technique is similar to that
used to establish known dc resistance ratios.

E,
Fig. 2Equivalent circuit of measuring system
employing permutation method with i = l.

switch position. Balancing the bridge with a small ad

mittance y assumed for simplicity to be connected to
the transformer winding of voltage E\, we have
t'i = - E,y
and
m+n
m+n
Ei y. + {mEi + nEt) Yk = 0.
i-i
t-i
Making use of the relationship
Ei

Ei
we have
H
*
i

II. Theory of Permutation Method

Fig. 2 shows a transformer with open-circuit ratio
Ei
= -M=
Ei

m
(1 + a + J/3),
n

m
(1 + +#),
ii

a+j& =

y>
=

ll'

EKt
kI
where bars denote averages over the m+n indices.

where m and n are integers and a and /3 are to be deter

mined. Connecting m-\-n admittances Y\, Yi, ,
Yk, , Ym+* as shown, we have with the detector
voltage zero,
n+m
n
, = , Yk + Ei L Yk.
k-n+l
t-1
If the admittances are permuted cyclically, each ad
mittance Yk is replaced by Yt+i, and Y+ is leplaced
by Ki. If this permutation is repeated m + n times, each
admittance appears with voltage E2 n times, and with
voltage Ei m times. We then have, after adding the
m+n equations,
m+n
m+n
m+n
m+n
i. = mEi ^ Yk + tiE2 Yk = (mEi + nEi) X) Ft,
-1
tI
t-1
t-1
where i, is the short-circuit detector current for the 5th

III. Lead Impedance Corrections

In the derivation of (1) in the previous section, it j=
assumed that the voltage ratio at the points of inter
connection is constant and independent of the permring switch position. Although this requirement can bt
satisfied, it is not necessary for the accuracy desireo
Formulas will now be developed which relate the ope
circuit transformer parameters to the average balam
ing admittance and the average loading associated i'
the assembly of permutable capacitors called the "rats
device" in the sequel.
Let the voltages at the interconnection pohvi
after permuting 5 times be given by ,(1 +e,'i
2(1 +/') where 1 and 2 are the open-circuit volt
ages and e,' and e," are functions of the lead impedance;
and load admittances. Then the short-circuit detect:
current corresponding to this configuration is given tV

350-244

the special case m = 10 and n = 1 by

E,(l*e's)

i. - ,(1 + O E ^* + t(l + e.")F.

- ,(1 + e.') ( E 7* - 7.)
- 10,(1 + a + #)(1 + e.")Y
where k runs from one to eleven. Balancing with an ad
mittance y, assumed to have a voltage i, we have,
ifter summing over all s from one to eleven,
E + E (i + o ( E f - k.) - ioE(i + .")f.
E,(Ues)
- 10(a+i/3) E (1 + OF. = 0.
Fig. 3Equivalent circuit of measuring system showing
lead impedances and load admittances.

.Vriting 7.= 7.+A7,, we have

Ey. + 10 (e/ - .") lFt- E A7.(10." + O

'here

- - Z,[K.' + (Jf + 1) 7."']

1
E ^ (io<." + ,.')
HOT 7.

+ zA y: + (m + i) r/"]

2,2,(7/7."+ 7.' Y."'+ 7." 7."')

E F.e."

_ -

EF.(l + OLl07.
1

'

A7.

e.=-Z,r7.+(l+iW'l
"I

riting -y, = 7a>c, + 2, and Y,=joiC. + G., and making

e tacit assumption that G,uC we have
jgt
- r
a +7(S = = - = + e.' - e." + ., + e,,
10C. 10u>C.

Af/

_ ^'^H*. J1, -f r . r .
1 +*.

t r , z. ) ^

where
^ = zx{Y,' + 7."') + Z2(7." + 7."')

ere

+ Z,Z2( 7/ 7." + 7.' 7."' + 7." 7.'")_

' = CT.'

Thus,
.+J,__i_*! + *rip + LIwn

isr. 3 represents the equivalent

circuit of a threeM
ding transformer and the ratio device at the sth
tch position. i and Ei= -MEi= - 10(1 +a+j0)E1
the open-circuit voltages; Zi and Zj are the impedes of the transformer and leads; 7.', 7.", and 7."'
the load admittances of the ratio device. A more acite equivalent circuit would show a transmission
connecting the transformer to the ratio device, but
our purposes the transmission line is short enough
be considered as lumped at the interconnection
its. Solving for e,' and e,", we have

.
7 fv'
j. 11
, J. * J. <
t w. J t h t i t i,
where
j

/ <t> \
El
' j
' *
( T
(
1\
r
' \Z*[Y' + \ + ^f)' ]~Z

ta =

351-245

_ (a _|_ y^) y,'" ( 10Zi + \

\
M/

'\\
+ {M+l)Y, \j

Zi = juLi + Ri
Zt = juL, + R,,
we have, after separating real and imaginary parts,
a = -^= + U*L1(C7+ MCT) - JLr(C7' + X.XCT)
10C
- RiiG? + UGT*) + R^GT7 + l.lGT^ + ReCf.), (2)
/? = - -r== - *i(c? + ncr) + wRitc? + i.ic7")
IOojC,
- Li(g7 + hgT77) + wL1(g77+ i.igT77)
(3)

+ Im (o),
where
0 = <1 + l + ff
Making the assumptions that
| a+jff\
| Zi| , | Zt\
| F.' I , | F." | , | F.'" |
9 X 10- mhos < | F.|
AF.I
F.

<
<
<
<

1 X 10-'
0.1 ohm
1 X 10-" mhos
11 X 10-' mhos

< 1 x io-

AAAAFig. 4Equivalent circuit of transformer showing

impedances and interwinding capacitance.
secondaries having the turns of one side in close proiimity to the turns of the other in order to reduce leak
age inductance.1 This construction produces a trans
former with a large distributed capacitance between
windings.
Fig. 4 represents the two secondary windings of a
transformer of voltages . and . C represents the in
terwinding capacitance and Er is the open-circuit volt
age between the two leads G, and G. For use in a bridge
circuit, the terminals G. and G4 of the transformer art
connected together and to the ground. We then havt
with
YZ.\ and | FZ| 1,
El.
EJ

/ Er

ST
< 1 x io-,
uC,
we can show that
| to | < 4.3 X 10-' | F'| max + 8.1 X lO-'
| Y" | max + 7.3 X 10-
| Y"'\ max+ 120 |*| ,

Et+ErZjY
E.+ ErZ.Y
Er
)]

(4)

where | Y'\ max is the largest of the eleven values for

| Y.'\ , and | Y"\ max and | Y"'\ max are defined simi
larly. It should be observed that o is the error caused
by neglecting certain terms in mathematical expan
sions, and does not represent an inherent limitation of
this method.
In general, it is desirable for the correction terms in
(2) and (3) to be small. The impedances may be signifi
cantly reduced by careful attention to the methods of
connection. For this reason, it is necessary to digress on
the subject of transformer design.
IV. Transformer Design
Precision transformers with low series impedances
are commonly constructed with heavy strap-copper

Ey is typically quite large and for a 10: 1 transform?

has a value between and
C is unavoidably lan;
because of the close spacing between windings, and vu<
be as great as 0.02 iif. From the above comments it ca
be seen that the stability of the transformer ratio
pends upon the stability of Z and Z.
For this reason, it is necessary for Z and Z* to :
small and well defined. This can be accomplished :
connecting the two secondary windings together w
manently within the transformer housing with sh*
low-resistance leads.
With the above connection made and a single ^
brought out from this connection, the complete tr".
former with its primary can be represented by "
equivalent circuit based on an ideal transformer
shown in Fig. 5(a). A, B, and G represent the tennis
of a connection block to which the secondary lead?
attached. The circuit of Fig. 5(a) is equivalent to :~
of Fig. 5(b). The accompanying formulas show that'
common impedance f< enters into the series impecr
terms, but not in a way basically different from v; <

352-246

It
AAA/V
///// '// '"////////,
'V//

AAAA
Si
AAA/^

'//

7S
;

/ '///////////V

(a)
ft
AAAAA

z1:

WW
AAA/^
(b)
Fig. 6Use of high-permeability core to
reduce equivalent impedances.

(b)
v-ft+r.(. + 9

f.'

(p.)' -P.'(l-Ply)
<ZV-*(l-P.)
lf>ipi pi'pipi ~ 1.
-Ideal transformer representation for three-winding trans
formers, where the p s are ideal transformer ratios.
Vith reference to Fig. 5(b) and the associated
li las, the parameters of interest are
10, ft " ti + 1.11-4, and f,' f, + llf,.
Pi
s usually best to make {V and fj' as small as possitcause of the factor 1 1 multiplying f4 in the expres"or f it is particularly desirable to make f as
sls possible. This may be achieved in effect by
rig separate ground leads from the common con>n point p within the transformer housing to the
crtion block terminal G, as in Fig. 6(a), and link-

ing one of these leads with one of the high-voltage leads

by means of a high-permeability core. This is electri
cally equivalent to breaking the connection at p. Fig.
6(b) shows the resulting equivalent circuit of the trans
former whose terminals are A, B and G of the connec
tion block. The terms Zi + Zj and Zj+Zi are the equiva
lent series impedances of the transformer, and Eb/Ea is
the open-circuit voltage ratio we desire to measure.
Since the value of Eb/Ea depends upon the loading ad
mittances present in the transformer, and its leads to
the connection block, these admittances must be well
defined. This is generally accomplished by shielding all
parts of the transformer and maintaining a fixed ge
ometry by means of rigid construction.
It can be seen from the pictorial nature of Fig. 6(a)
that the equivalent series inductances of the transformer
can be reduced by reducing the shaded areas in the fig
ure. In practice, each lead pair consists of a coaxial
lead which, in addition to reducing the inductances as
measured at the connection block, also minimizes the
external fields produced by currents in these leads and
hence interactions with other components of the meas
uring system.
The coaxial lead plus high-permeability core tech
nique can also be profitably used to reduce effective

353-247

impedances from the connection block to the ratio de

vice. The use of high-permeability cores in a related ap
plication has been discussed by Thompson.1
V. Lead Impedance and Load
Admittance Measurement
Having discussed methods of obtaining small, welldefined impedances, we next turn our attention to the
measurement of these impedances. One method of
measurement may be developed by reference to (2) of
Section III. If C,'", for example, is increased by the
addition of components within the ratio device (beyond
the permuting switch), one obtains
c. + Ac. + u,sL,(C.' + 11C.'" + 11 AC.'")
iocT
- u>'U(c77+ 1.1CV + 1.1 AC")
- r,(g7 + war) + R2(G/' + l.iGT)
+ Re (to + Ae0),

(5)

where AC,'", Ac,, and Ato are the increases in C,'", c

and to, respectively. Subtracting (2) from (5) and rear
ranging terms, we obtain
U - 10Z,, =

Ac.
Re (A0)
+
1WC,(AC."')
l.la>2(AC."')

(6)

Similarly, from (3),

R2 - WR,

Ag.
n^C,{AC."')

Im (Ato*
l.lo)(AC.'")

(7)

One might suppose that if similar measurements

were made with components added beyond the permut
ing switch to increase G", for example, the value of Z.2
by itself might be obtained with comparable precision.
However, since the permutation process places each
added ground capacitance on the high-voltage side
once and on the low-voltage side ten times, a change of
AC," in C," produces a corresponding change of 10
AC,' in C,', so that one arrives again at values for
Lt 10Li and R2 lOJtt. In fact, were it not for the
loading admittances associated with the wiring be
tween the permuting switch and the connection block,
we would have C7' = 10C7' and CV=10CY', in which
case knowledge of L\, Li, R\, or R% alone would not be
needed. This fact can best be seen if we let =Lt 10Z.1
and (R = i?i-10i?i, and rewrite (2) and (3) in the fol
lowing manner:
a = -^= 10C.

10C.

\10

+ o>L

Im (to).

Since the loading admittances associated with the

leads can be made very small, the quantities C," C.'/IO
and CV'-CV/10 in (8) and (9) will also be very small
and, consequently, the accuracy to which L2 and R;
need be measured is greatly reduced.
The principal significance of this' procedure for th
measurement of and <R is that any component added
past the interconnection points produces essentially no
error by virtue of its mutual inductance with other por
tions of the circuit. Mutual inductance between the
added component and other elements of the ratio device
changes the effective values of the transfer admittances
Yk defined in Section II, but the derivation is still valid
if these new values for the Yk are used. In practice,
since Yt is needed only to an accuracy of a few per cent,
the effect is insignificant.
Errors resulting from mutual inductances between
the added component and the leads preceding the in
terconnection points also cancel to first order, since any
change in the voltage at the interconnection point',
caused by such mutual inductance, reverses sign when
the component is switched from one side of the trans
former to the other. It can be seen that the net effect is
to leave the average voltage ratio at the interconnec
tion points independent of such mutual inductance.
It can now be seen that the only mutual inductances
which may produce errors in the measurement of a and
/3 are those between the nonpermuted portions of tie
circuit. Since the nonpermuted circuit consists almost
entirely of coaxial leads, we may expect all errors caused
by mutual inductances to be negligible.
Since it is difficult to reduce the load admittances of
the leads and switch to zero, values of Li and J?j art
still required in formulas (8) and (9), but relatively lo*
accuracy is sufficient. One method of measurement mav
be developed by referring to (8). If components are
added outside the ratio device so as to increase C." b>
an amount AC," with corresponding increases Ac, '
Ato, one obtains
Ac.
/C7 + 1.1C.'")
\
a = c. +=--u,JjE(
10C.
\ 10
/

( + I.TtA
\ 10
/
- o.'z.'^CT7 + ac7' - -^0

- 0,'Z, (C/'-f)+<R(^+11G'"')
/
+ R: lG."

CA
+ Re (to),

(91

(8)
-I- Re (to + Ae0).
354-248

btracting (8) from (10) and rearranging terms, we

tain
10u2C.(AC.")

a>s(AC.")

nilarly, from (9), we have

R2

Ae,

10u2C.(AC.")

Im (Ato)
-
u(AC.")

(12)

The capacitance coefficients C,' and C," in (2) and

represent principally the ground capacitances of the
io device and are determined by the design of the
nponent capacitors. Capacitors could be designed
:h ground capacitances significantly smaller than
>se incorporated in the ratio device, but this would
>bably be at the expense of their stability. C,'" could
made zero by careful shielding, but it is normally
y small and causes no particular trouble in measurent. We are faced then with the problem of measurC, C", and C.'".
n use, the ratio device is operated with the bridge
anced, meaning that the detector voltage is zero,
us, a correct set of values for C.', C,", and C,'" may
obtained if the detector terminal is shorted to
und. The ratio device then has three remaining
minals to which are attached the leads from A, B,
1 G of the connection block. Representing this by
. 7, we see that C,'" is the easiest coefficient to measand is just the direct capacitance between points A
I B. C,'" is the average of the eleven measurements of
' corresponding to the eleven switch positions. C,'
be determined in practice by shorting B to G and
isuring C,'' + &"'. This can be done with simple twoninal capacitance measuring techniques. Since the
isformer ratio is defined at the connection block,
desired value for C.' + C,'" is the increase in ground
acitance of the connection block when the ratio de: is connected to it.
veraging the eleven measurements, we obtain C,'+
', from which C,' can be determined, and similarly,
G,', G,", and G,'" may be obtained simultaneously
l the same type of measurement.

II
c".
Cs

Cs

G
'Ratio device admittances with detector terminal shorted.

VI. Numerical Results

The procedures developed in the preceding sections
of this paper have been applied to the measurement of
a particular transformer, NBS 89785. This transformer
was built according to the principles described at the
1958 Conference on Electrical Measurements2 except
that the thickness of the copper-strap secondary wind
ing has been reduced to 0.030 inch to allow more room
for insulation.
The ratio device and the transformer were placed in
an oil bath with + 0.01C regulation. This decreased
the drift rate of the capacitors in the ratio device, and
hence improved the accuracy attainable with the meas
uring system. Since the transformer ratio is slightly
temperature-sensitive, temperature control of the trans
former was also required (see Fig. 8).
The work described below was all done at an angular
frequency of 104 radians per second (about 1592 cps).
The voltage level was monitored with an expandedscale vacuum-tube voltmeter accurate to + J per cent.
The small variable admittances, c, and g consisted
of the low-range decade capacitor and the conductance
balance control described by McGregor, et al.,2 but al
most any capacitor variable in steps of 10~* pf and ac
curate to 10~7 pf would have been sufficient. The same
applies to the conductance balance control, which must
have steps of 10-> /imhos accurate to 10-9 /xmhos. Since
only a small total range of the variable admittances is
required, this is not difficult to achieve.
In Section II, it was assumed that the small variable
admittances were connected to the low-voltage side of
the transformer. It is necessary to reverse the signs of
these admittances in order to balance the bridge at
certain switch positions. We can accomplish this by
connecting one or both of the variable admittances to
the high-voltage side of the transformer, but this
changes the apparent scale of the variable by a factor
of ten as well as changing the sign. The transformer
under investigation had a tap on the high-voltage side
of the transformer at which the voltage was approxi
mately equal to that of the low-voltage side, but 180
out of phase. Use of this tap for the balancing admit
tances, when required, greatly simplified the reduction
of data.
Since the variable admittances have, in addition to
their direct admittances, relatively high admittances
to ground, it is necessary to measure these ground ad
mittances as well as the effect on the ten-to-one trans
former ratio of a ground admittance at each of the trans
former terminals to which the variable admittances
may be connected, and apply appropriate corrections.
If it is borne in mind that the average voltage ratio is
the quantity of interest (see Section III), the correction
procedure is straightforward. It involves, in addition
to the quantities mentioned above, a knowledge of how
many times each of the variable admittances is on each
low-voltage transformer tap.
The following parameters were determined for the

355-249

described system:
= - 3.49 X 10-' henry

<R = - 66.3 X 10-' ohms

Li = 1.0 X 10"{ henry

R2 = SO X 10-' ohms

C7 = 1101 X 10"" farads

C?7 = 127.1 X 10"" farads

G7 = 44 X 10-' mhos
G.n = 5 X 10-* mhos

CT' = 18.3 X 10-" farads

OT = 1 X 10-* mhos.

Correction to a for loading of variable admittances

= -4X10-*.
Correction to /3 for loading of variable admittances
= -19X10-'.
Total correction to a =+39X10-'.
Total correction to 0= -97X10-'.
These corrections were determined several times at
various temperatures and voltages. They were ob
served to be constant to 1X10-' over the range of
conditions reported below.
c, and g, were determined from twenty-one separate
bridge balances, with the ratio device switch running
from one to eleven and back down to one. This pro
cedure tends to eliminate the effect of a steady drift
rate of any of the ratio device capacitors. The twentyone balances could be made in a period of time ranging
from twenty to thirty minutes, depending upon the ex
perience of the operator.
Tables I and II represent the transformer ratio pa
rameters as functions of temperature and voltage. Each
number represents the mean of at least five ratio meas
urements. The probable error of each mean is 2 X 10"' or
less.
TABLE I
VARIATION OF a AND 0 WITH VOLTAGE24.83C
Voltage
90
100
110

or
-16X10-'
+26
+62

fi
-12X10-'
-16
-14

TABLE II
Variation of a AND 0 WITH TEMPERATURE 100 VOLTS
Temperature
24.62
24.83
25.42

a
+ 12X10-'
+26
+34

- 4X10'
-16
- 9

Fig. 8Photograph of measuring system. From left to right in ti

bath: ratio device, connection block, and transformer.
Attempts to repeat the numbers presented in Tab*I and II were not completely satisfactory, the discrep
ancies occasionally being as large as 10X10-'. Thor
ough investigation of the measurement equipment daclosed no sources of error large enough to account far
discrepancies larger than 2X10-'. It is believed th.:
variations in the transformer ratio were being observed
Since it is necessary to wait at least a day after chang
ing the oil bath temperature to obtain good tempera
ture regulation, the apparent temperature depender.f
of the transformer ratioin particular its phase defec
angleindicated by Table II, is probably at least par
tially due to random variations within the transformer
Since the theoretical development of Section II as
sumes a linear volt-ampere relationship for the corponents of the ratio device, a method is needed to de
termine whether or not the capacitors have voltaif
coefficients large enough to produce an error. Sensitivity
and drift problems make such a determination difficult
but comparisons with capacitors of radically differenconstruction indicate that any voltage coefficwct
which may exist could not produce an error larger thar.
3X10-' in a or 0 (50 per cent confidence interval). We
conclude that with the equipment described above, tecto-one transformer ratios may be determined v
4X10"* with a 50 per cent confidence intervaL A
more extensive investigation of voltage coefficient
might reduce this figure.
ACKNOWLEDGM ENT
The authors wish to thank Lai H. Lee for his assistance
in obtaining the numerical results of this paper.

Reprinted from IRE TRANSACTIONS

ON 1NSTRV MENTATION
Volume 1-0, Number 2, September, 1960

356-250

TRANSACTIONS: 3 (238-242) 1964

Comparison

of

Calibration

Inductive Voltage

Dividers*

RAYMOND V. LISLE |
RCA Service Company
Missile Test Project
Patrick Air Force Base, Florida
and
THOMAS L. ZAPFt
Radio Standards Laboratory
National Bureau of Standards
Boulder, Colorado

y Calibrated inductive voltage dividers serve as excellent standards of voltage ratio

for use in the comparison calibration of other AC voltage dividers. Differences in phase
angles in the dividers can be accommodated by the production of a voltage in the
detector circuit in quadrature with the reference voltage. This paper discusses the
comparison circuit, the handling of stray impedances, guarding, measurement pro
cedures, quadrature voltage production, and errors. The comparison method described
can be used in standardizing laboratories for the calibration of inductive voltage
dividers and other voltage ratio equipment.
INTRODUCTION
iductive voltage dividers may be described as
tapped inductors wound on high-permeability cores.
; relatively recent metallurgical development of
omagnetic core material having extremely high
gnetic permeability has made possible the construci of inductive voltage dividers having excellent
>ility and loading characteristics. The effects of
rnal loading are inherently small in such inductive
age dividers, and errors in voltage ratios are
[uently less than 1 x 10"* in the low audioiiiency range. The absolute measurement of voltage
os with uncertainties less than this is difficult,
vever, known voltage ratios embodied in an inducvoltage divider may be preserved as a standard and
stablished again and again. The accurate initial
blishment of voltage ratio for the calibration of
ictive voltage dividers*1-4' is not discussed here, but
;thod of calibrating an inductive voltage divider by
paring it with a standard is presented. The comparimethod, essentially as described herein, has been
at the National Bureau of Standards for several
ented at the 1 8th Annual ISA Conference and Exhibit,
ember 9-12, 1963, Chicago, Illinois,
idards Engineer.
iicist. Radio Standards Engineering Division; MemberISA
ver Section.

years and appears worthy of general use in standardizing

laboratories. This method provides efficient, accurate
calibration of inductive voltage dividers using common
laboratory equipment with a standard inductive voltage
divider. Comparison circuits have been mentioned in
several publications'4 7) without detailed elaboration.
This paper discusses in considerable detail a particularly
successful type of comparison circuit.
THE COMPARISON CIRCUIT
A simple bridge circuit for comparing inductive
voltage dividers is shown in Figure 1. The standard
inductive voltage divider is the divider for which the
corrections are known.'The unknown inductive voltage
divider is the divider for which the corrections are to be
obtained.
Other elements of the circuit are the guard divider
and shielding circuit, flexible coaxial cables to the
detector transformer, the detector, a resistor and
capacitor for the introduction of a quadrature voltage
in the detector circuit, and several fuses. The dashed
lines indicate the connections of a capacitor that may
be used to reduce loading.
The input terminals of the unknown and standard
dividers are joined together by short, heavy leads having
low impedance to reduce voltage drops between the

357-238

GUARD

STANDARD

UNKNOWN

Figure I. Bridge circuit for On

calibration of an inductive
voltage divider by comparison
with a standard.

dividers. The electrical connection to the case of the

standard divider should duplicate the connection used
when the standard divider was calibrated. Similarly, the
electrical connection to the case of the unknown divider
should duplicate the connection to be used in subsequent
work. In Figure 1 the cases are shown connected to the
common terminals. These connections fix, in a re
producible manner, the stray capacitances from the
windings to the case of each divider. Connecting the
case to the output terminal is not recommended, as
this would not fix the stray impedances from the case
(and output terminal) to the surroundings, which are
often quite uncertain and variable. The output of the
guard divider is connected to the shield around the
detector circuit. When comparing an unknown induc
tive divider with the standard, the guard divider is set
at the same setting as the other dividers and drives the
shield at nearly the same instantaneous potential as
that of the conductors within. The guard divider may
be either inductive or resistive bzcausc its only function
is to maintain the proper potential on the shield. The
guard circuit minimizes current leakage through the
stray impedances of the detector circuit.
Usually there is a relatively large stray capacitance
to ground from the detector transformer shield and the
capacitor case. The current in this stray capacitance to
ground must not be allowed in the shield of the flexible
coaxial cables going to the output terminals of the
dividers because a current in the shield can induce a
voltage in the inner conductor. To minimize this error,
the connection between the detector transformer shield
and the capacitor case should be made as short as
possible, and the output of the guard divider should be
connected to the shield at some point in this vicinity.
The major components of stray capacitive current then
are conducted directly to the guard divider. The
deleterious effect that is avoided by this procedure may
be observed by increasing the length of the leads in the
detector circuit and connecting the lead from the guard
divider to a different point on the shield. If all leads are
short the error may be insignificant.

It should be noted that, with the circuit shown ::.

Figure 1, the guard circuit and the capacitor case are
at a potential somewhat above ground, and care should
be exercised in touching the shields to avoid hazardeo
shocks. Alternatively, the circuit may be grounded al
the output of the guard divider, but the cases of the
dividers will then be at a high potential. With this
connection it may be necessary to use an isolation
transformer between the power supply and the bnd?e
circuit to avoid double grounding. Regardless of the
grounding system selected, it is good practice to
provide an outer covering of insulation for the shielding
wherever needed to prevent accidental short circuits.
Many inductive dividers are provided with to
"common" terminals connected together and to one
extremity of the winding by a wire of low impedance so
that one terminal may be used with the input circuit
and the other may be used to make connections to the
load. In the circuit shown in Figure 1 the inductee
dividers are treated as three-terminal dividers, i.e.. the
common output terminal is ignored. A small voltase
drop in the internal connection between the twocommer.
terminals would make the three-terminal measurement*
differ from four-terminal measurements in which hot1'
common terminals are used. However, the differences
usually small and often insignificant. Fuses are inserted
in the circuit as shown in Figure 1 to prevent dama
to components in the event of inadvertent short circuits
The fuse in the detector circuit prevents damage to the
resistor, /?, if a large unbalance of the bridge circuit
encountered.
VOLTAGE RATIO MEASUREMENT
Let D + d represent the dial reading of an inductive
voltage divider, where D is the reading of the decades to
be calibrated (usually the highest three decades) and
d is the reading of the lower decades. In most calibra
tions D will be a step setting of one of the highest three
decades with the other two set to zero. The true ratio, I
can then be represented by D + d + r, where c is the

358-239

TABLE I
Voltage ratio
D.
0.900
0.500
0.100
0.090
0.050
0.010
0.009
0.005
0.001

d.
x 10-
0.0
0.0
0.0
+ 0.1
0.0
0.0
0.0
+ 0.1
0.0

Cg
d,
x 10-' x 10-*
+ 0.4
+ 0.1
0.0
+ 0.4
-0.1
+ 0.1
-0.5
0.0
-0.3
0.0
0.0
0.0
0.0
0.0
-0.1
0.0
0.0
0.0

Cx
x lO"1
+ 0.3
-0.4
-0.2
-0.4
-0.3
0.0
0.0
0.0
0.0

rrection to the setting D. For all practical purposes c

independent of d. At balance
Dz + dz + ci = D, + d, + c,

(1)

ere the subscripts x and j denote the unknown and

! standard, respectively. Solving for the correction
the unknown divider and realizing that Dz = D,,
c obtains
Cx = c, + d, - dx
(2)
dinarily, dz would be set to zero. Under this condition
cx = c, + d,
(3)
)wever, if cz < c and if the standard divider cannot
set to a negative d,, the lower decades of the unknown
'ider, dz, can be used to obtain the balance. If ds is
to zero the ratio balance equation is
cx = c, - dx

Ratio,

Phase angle
y prad Ay, jirad fx, urad

0.900 000 3
0.499 999 6
0.099 999 8
0.089 999 6
0.049 999 7
0.010 000 0
0.009 000 0
0.005 000 0
0.001 000 0

PHASE-ANGLE MEASUREMENT
\ true voltage null at the detector can be obtained if
juadrature voltage is introduced into the circuit to
npensate for the phase-angle difference between

0
+1
+3
+4
-4
+ 20
+ 10
+ 30
+ 130

-2
+5
+ 52
+ 40
+ 46
+ 90
+ 70
+ 110
+ 460

the unknown and the standard inductive voltage

dividers. The RC combination in Figure 1 is used to
develop and introduce a voltage in quadrature with the
output of one of the dividers. If the capacitive reactance
of C is made considerably larger than the resistance, /?,
the current in R will lead the voltage across the im
pedance R jXc by nearly rr/2 radians and will cause
the quadrature voltage to appear across R. At balance
the current in the detector is zero, and the voltage at
the junction, of R and Cis , as indicated in Figure 2a.
An expression for the current in R and C can be written
i = E.jXc

(5)

The quadrature voltage, e, is the product of i times R,

e = iR
(6)
and
e = E,R Xc = Esa>RC

(4)

pical data and results for a calibration at 100 V and

30 cps are shown in Table 1.

-2
+4
+ 49
+ 36
+ 42
+ 70
+ 60
+ 80
+ 330

(7)

From Figure 2a, y, and yz are the phase angles

associated with the standard and the unknown inductive
voltage dividers, respectively, and Ay is the phase-angle
difference between the unknown and the standard
dividers. An expression for Ay can be written from
inspection of Figure 2a as
tan Ay = e E

(8)

ERROR

Figure 2. Phasor diagram:

(a) voltage phajors; (b) voltage
ratio phasors; (c) phasor dia
gram showing small error
from quadrature circuit.

(0)

315-411 O - 68 - 24

(b)

359-240

Combining the last two equations yields

tan Ay = <oRC
For small angles tan Ay x Ay, and
Ay fu wRC

(9)
(10)

This equation provides a design formula for the RC

combination in the detector circuit. For any test
frequency w a variable capacitor can be chosen in
combination with a resistor so that Ay in microradians
is a multiple-of-ten times the reading of the variable
capacitor in picofarads. Various multiplying factors
can be obtained using resistors differing by multiples
of ten. As an example, at a frequency of 1000 cps,
resistors having values of 1592, 159.2, and 15.92 Q, will
produce multipliers of 10, 1, and 0.1, respectively.
Thus a 0-100 pF variable air capacitor can be used to
cover a wide range of phase-angle differences.
In the above discussion with reference to Figure 2a,
it has been assumed that the phase angle of the unknown
inductive voltage divider is larger than that of the stan
dard inductive voltage divider. Under these conditions,
Ay is positive, and the phase angle of the unknown
inductive divider can be obtained from the equation
yx = y + Ay
(11)
If the difference yx y> is negative, the comparison
circuit can be balanced by putting the RC combination
on the other side of the detector with R connected to
the standard divider. Under these conditions, the phaseangle balance equation would be
yx = y, - Ay
(12)
Typical data and results from phase-angle measurements
are shown in Table I.
DISCUSSION OF ERRORS
If errors from incorrect connections, incorrect
frequency, or incorrect applied voltage (i.e., errors
arising from conditions different from those under which
the standard was measured) are precluded, the ponder
able systematic errors, such as those arising from the
assumptions made in the derivation of the equations,
are of very small magnitude. These will now be
investigated.
The error introduced by the approximation that
tan Ay x Ay can be evaluated from the trigonometric
expansion of the tangent, namely,
Ay3 Ay5
tanAy = Ay +
+
+ ...
(13)
3
15
The second term, Ay3/!, evaluated for Ay = 300 /xrad
would be of the order of 1 x 10-11 rad, which is negli
gible, and higher order terms are smaller yet.
It is convenient to consider a phasor diagram of
voltage-ratio phasors, as shown in Figure 2b, rather
than one formed of voltage phasors, as in Figure 2a.
The diagram in Figure 2b is formed by normalizing the
voltage phasors by dividing the circuit voltages by the
input voltage magnitude. The voltage ratio A of an

inductive voltage divider is then Eo/\Em\, and the

quadrature component, q = e/|in|.
A small error in ratio from the voltage drop across R
is illustrated in Figure 2c in which the angles have been
grossly exaggerated to illustrate the error. In this
diagram A, is the apparent ratio at balance, smaller
than the true ratio A', because the phasor q is added at
right angles to the phasor A, by means of the RC
circuit. The error, A A,, can be evaluated from the
trigonometric series expansion for cos Ay, namely,
Ay2 Ay*
cos Ay = 1
-I...
(14)
2!
4!
From Figure 2c,
A, = Ax cos Ay s Ax(l -

(15)

The error in ratio is represented by the term Ari&rll'.).

In comparing two good-quality inductive voltage
dividers at a frequency of 1000 cps a Ay of 100>rad
would be experienced only at the lower values of ratio
below about A = 0.01. The error in ratio would be
approximately 5 x 10_u. An error of this magnitude
would be negligible in most cases.
Another source of error in both magnitude of voltage
ratio and phase angle is the loading on the unknown
divider by the capacitor C. To determine the magnitude
of these errors in practical work, the difference in
voltage ratio and phase angle can be observed when
another capacitor having the same value as C is con
nected from the output terminal of the unknown divider
to ground in the circuit in Figure 1 (i.e., across R and
C). The observed differences may be subtracted from
the original readings to yield the values of ratio and
phase angle with no load. Alternatively, the loading
may be eliminated by adding a capacitor as shown by
the dashed lines in Figure 1 : C and C form a capacitive
divider to reduce the loading of C on the output circuit
of the inductive divider. The magnitude of C" must be
adjusted depending on the ratio being measured and.
in particular, must have a value equal to CA,I(\ - ^i)The power source should have a reasonably lo
harmonic content, and the null detector should be
tunable to suppress harmonics in the quadrature
voltage, which would otherwise prevent the attainment
of a precise null. To understand the need for a tunable
null detector, consider that harmonic components
exist in both E, and Ex having a frequency nu>. equal
amplitude hE,, and the same phase. The total voltage
across R is
e = ea + enu (phasor sum)
(161
where eu is the fundamental component and efl is the
harmonic component as shown in Figure 3. The phasor
enu rotates n times faster than e, and its magnitude can
be expressed in terms of the phase-angle difference
of the fundamental output voltages, Ay. The expression
is similar to that for equation (7), except the harmonic
current is hE,/Xc and the angular frequency is n; thus
e = hEtntoRC = nhEAy

360-241

The corrections to both inductive voltage dividers for

the setting of A = 0.5 can be obtained from the comple
mentary ratio measurements without prior knowledge
of either of these corrections. If dt is the reading of the
standard for the normal circuit as shown in Figure 1,
and d', is the reading of the standard with the unknown
divider connections reversed, then
c, = -{d, + d',)l2 and cx = {d, - d',)\2 (20)
nire 3. Phasor diagram showing effect of a harmonic in the
adrature voltage.
this component is not suppressed by the detector, it
>uld not by itself create a systematic error, but it
mid result in a limitation on the attainment of a null
uivalent to a voltage-ratio uncertainty
Cum
fnu
fnw
UA = TF1 = 777 = AT = AnhAy
(18)
1 a phase-angle uncertainty
turn
uy =
= nAAy [radians]

(19)

i typical measurement at 1000 cps, where A = 0.1,

3, h = 0.01, and Ay = 20 ^rad, it is found that
= 0.06 x 10~6 and uy = 0.6 /*rad. It is evident that
ined indicating amplifier that suppresses the third
monic by a factor of 10 would be sufficient to permit
discrimination of ratio to within 0.01 x 10~8 in
case.
should be noted that if a detector is used that has a
ir response to small signals, it is possible to use it
iterpolate between steps on the lowest decade of the
ier.
MEASUREMENT VERIFICATION
simple partial check on the bridge circuit and the
ration results can be performed by reversing the
; to the unknown divider and measuring the
jlement of the ratio. <4) Caution should be exercised
jse the case of the unknown divider will now be at a
potential if it has been connected to the common
nal. It can be seen that 1 Az is now to be
iared with A,. With this connection cz =
+ d, + dx). The correction, c*, obtained in this
nay be compared with that obtained in the normal

By a somewhat analogous method a check on the

phase-angle measurements can be made. The formula
for phase angle can be derived simply from the phasor
diagrams of voltage ratio for the reversed connection.
The phase angles of both inductive voltage dividers set
at A = 0.5 can be obtained by the method of comple
mentary measurements without prior knowledge of
either of these phase angles.

CONCLUSION
If care is exercised in setting up the circuit, using good
shielding practices, and a detector having adequate
amplification and harmonic suppression is employed,
the voltage ratios and phase angles of two good-quality
inductive voltage dividers may be compared with
negligible loss in accuracy. Several systematic errors in
the circuit, for which corrections could be made, have
been discussed and found to be negligible under the
typical conditions described. The precision of the
comparison is limited primarily by the resolution and
stability of the standard and unknown dividers. The
existence of a method for accurate comparison makes
possible the use of calibrated inductive voltage dividers
as standards for the calibration of other dividers and
these, ultimately, for the very accurate measurement of
voltage ratio and impedance ratio.

REFERENCES
1. Sze, W. C, Trans. A.I.E.E., Part 1 76, 444, 1957.
2. Harris, F. K..and W. C. Sze(NBS Washington), unpublished
communications to the authors.
3. Zapf, T. L.,J. Res. NBS 66C, 25. 1962.
4. Engineering Bulletin No. I (Ratio Tran), Gertsch Products
Inc., Los Angeles, California.
5. Engineering Bulletin No. 4 (Ratio Tran), Gertsch Products
Inc., Los Angeles, California.
6. Morrison, N. E., Engineering Bulletin No. 20, Electro
Scientific Industries, Portland, Oregon.
7. Morgan, M. L., Engineering Bulletin No. 29, Electro
Scientific Industries, Portland, Oregon.

361-242

PREPRINT

$\ INSTRUMENT SOCIETY of AMERICA

21it Annual ISA Conference and Eihibit October 2<-27, 1966, New York Number 12.3-2-66

COMPARATOR FOR CALIBRATION OF INDUCTIVE VOLTAGE DIVIDERS FROM 1 TO 10 kHz*

Wilbur C. Sze
Electrical Engineer
National Bureau of Standards
Washington, D.C. 20234
ABSTRACT
A high accuracy comparator Is described for
measuring the relative deviations In voltage
ratios and phase angles of inductive voltage
dividers. The new technique overcomes several
inherent limitations of the existing comparison
methods. The balance Is accomplished by uti
lizing special shielded transformers and a
resistance-capacitance network for ln-phase and
quadrature voltage Injections. The measurements
art accurate to within 1x10"' of Input in the
frequency range from 1 to 10 kHz. Resolution
is better than 2x10" 10
INTRODUCTION
The Inductive voltage divider, since its
first appearance about 15 years ago, has proven
to be an increasingly valuable tool In the
field of accurate electrical measurements as its
striking characteristics become more widely
known .
The development of material of extremely
high magnetic permeability has made possible
the construction of Inductive voltage dividers
with excellent stability and with deviations in
ratio and phase angle of less than 5x10"' of
input. 1** The accurate initial establishment of
the voltage ratio and phase angle has been
presented elsewhere and will not be repeated
here. These known voltage ratios and phase
angles are preserved in the calibrated
inductive voltage divider so that they may be
used as a reference and re-established again by
a satisfactory comparison circuit without
significant degradation in accuracy.
At the National Bureau of Standards, a new
comparison method, based on suggestions by
Robert D. Cutkosky, has been developed for
relative ratio and phase angle measurements
from 1 to 10 kHz to an accuracy of better than

This work was supported in part by the Metrol

ogy Engineering Center, Bureau of Naval Weapons,
Pomona, California.
However, in demonstrating the accuracy of the
comparator, commercially available dividers
were used.

one part in 10'. This technique, which appears

worthy of general use in standardizing labora
tories, overcones several inherent limitations of
the existing comparison methods: (1) the
dividers are treated as four-terminal networks as
they were designed to be operated; (2) the test
circuit imposes negligible burden on either the
reference or the test divider; and (3) it is not
necessary to use lower dials of either divider
for in-phase voltage balancing.
DESCRIPTION OF CIRCUIT
This technique uses a double toroidal trans
former, one transformer for null detection and
the other for injection of differential voltages.
Figure 1-a shows the complete schematic circuit
diagram simplified by omission of the shielding
arrangements. Dj and T>2 represent the reference
and test dividers whose voltage ratios and phase
angles are being compared. Their input and oatput terminals are connected together with short
coaxial leads. The lead Impedance between the
Input terminal of each divider and its junction
point with the supply voltage should be very
small and nearly equal. The detector circuit
consists of a tuned null-indicator and a
specially constructed shielded transformer, T3,
to match the Impedances and to isolate the
detector from the measuring circuit. The parallel
pair of inductive voltage dividers, D3 and D4,
inject, through the R-C network shown and trans
former, T2, in-phase and quadrature balance
voltages into the lead connecting the output
terminals of the dividers, Di and D2. Trans
former, Ti, is also specially constructed with an
accurate 1/1 voltage ratio. The mid-point of the
secondary winding is grounded so that both
polarities of the injection voltages can be
obtained from the outputs of D3 and D4 without
the necessity of altering the circuit.
In order to realize an accuracy of better
than one part in 10', coaxial chokes are required
in the circuit to reduce the small but trouble
some loop and ground currents to a ulnimum. * > "
Each coaxial choke consists of an appropriate
number of turns of a coaxial cable threaded
through a high permeability core as indicated in
Figure 1-a.
Photographs 1, 2, and 3 show the comparator
and the circuit arrangement.
Superior numbers refer to similarly numbered
references at the end of this paper.
362-1

rHEORETICM. RELATIONS
By definition, the voltage ratio of a
ivlder U the ratio of the secondary or output
erminal voltage, Vgtj-p, to the primary or input
erminal voltage, Vju, which can be expressed as
vOUT
VIN - d+t*d +
here d is the nominal ratio or setting of the
Lais. Wd la the ln-phase deviation, and $d
3 the quadrature deviation. Both are expressed
i parts-per-mlllloa (ppm) of Input.

^2^1

'
rdltRi+XC<1 + RT) 1

If Xc 2 Ri and Rj>>R2, then the equations

for the relative deviations can be approximated
as
dR
r(_D
R2
d.R,
dl(?2-^l>rXc-

The equations for the relative ln-phase

viatlon and quadrature deviation measurements
in be derived from Figure 1-b as follows:
DESCRIPTION OF COMPONENTS
- Vi + V,

Tj, T2, and T3 are specially constructed

shielded transformers using high magnetic
permeability toroidal cores to insure close
magnetic coupling and to conserve space in the
circuit. The shields minimize the magnetic pick
up from stray fields in the laboratory.

VC - "ICJXC + V3
R2Z
re Z is the impedance of transformer T2.
>ce in the circuit described here, ZR2, the
jve approximation is justified. Vq and VR are
i voltages between the output terminals of D3
I D4 and ground.
By solving for IR and Iq from the first two
at ions and substituting in the third,

2 2
Ri 2
Ri + xc<1 + nf>

If r represents the turns ratio of T2, then

voltage Injected between output terminals of
md D2 is equal to V-j/r. However, V= and VR
also be expressed as decimal fractions of the
ily voltage or by dial settings of O3 and
Iq and dp, respectively, and V3 as d3.
A3aln froji Figure 1-a, when the detector
catea a null,
d2(l+c<2+ jl52)-d1(l + 6(1 J^j) - -*
e d. and d2 are the equal dial settings on
dividers Dj and D2, at the point being
ired .
cy and i are expressed in parts-perlon of dial setting, d. By substituting
xpression for d-j in the previous equation
separating real and j terms,
dCRl2+dRXC<1 + Rf_
rd1(R^ + X^(l + Rl)2]

363-2

Figure 3 shows the schematic circuit of T^.

It consists of a toroidal core, an Inner primary
winding, an electrostatic shield, an outer
secondary winding, and an overall Mu-metal case.
Each winding is wound on the core in a single,
equally distributed layer from 0 to 360, and the
wire is brought from 360 back to 0" or the point
of origin. Thus no loop is formed, and suscep
tibility to stray magnetic pickup is greatly
reduced. (This same technique is used for
winding T2 and T-j.) The mid-point of the
secondary winding Is grounded. The size of the
core and number of turns for 1 to 5 kHz and 5 to
10 kHz operations at 100 volts input are indicated"
in Figure 3.
Figure 2 shows the details of construction of
Tj and T3 transformers as a single unit. T2 and
T3 are Identical and interchangeable. Each
consists of a toroidal core and a 100-turn
primary winding. The shielding arrangements are
such that the cross capacitances at the gaps are
reduced to a ulnlmum.' The secondary winding
consists of a single conductor connected between
the outputs of the test dividers. The capacitance
between this conductor and the shield Is less than
4 pF. Therefore, it imposes no significant
burden on either divider.
The inductive voltage dividers, Dj and D4,
are co.nraerc ial 1 y available units. Each consists
of three decades and a continuously adjustable
resistive divider. Their small voltage ratio
and phase angle deviations were determined and
found to be negligible, for the Injection voltage,
V}, rarely exceeds 0.002 percent of the input
voltage .

C is a three -terminal capacitor having a very

small dissipation factor. Rj and R2 are threeterminal multi-decade resistors of very small
residual reactance. For 1 kHz calibrations, C is
generally set to equal 15.92 nF, Rj - 10 kilohms,
and R2 10 ohms. (For 10 kHz calibrations, air
capacitors of 1392 pF are used.) Thus one step
on the top decade of Dj and
represents a dif
ference of 1 ppm in ln-phase or quadrature
deviation.
The detector circuit consists of a preampli
fier and a cathode-ray oscillograph. A narrow
band-pass filter is used to discriminate against
any noise before the first stage of amplification.
A double elliptical pattern, suggested by Clothier,
is shown on the cathode-ray oscillograph to
indicate balance condition. The unbalance volt
age from the comparison circuit is impressed on
the vertical plates. On the horizontal plates is
a signal of the same frequency whose phase can be
adjusted so that angular departure of the major
axis of the elliptical pattern represents inphase unbalance and opening of the minor axis
quadrature unbalance. At balance the figure
appears to be a single line. 2, 4, 8 The resolution
is such that an unbalance voltage of 0.1 nanovolt
is easily observed.
TABLE OF RESULTS
The following table lists the in-phase devi
ations, djQ< , and d20<2, and quadrature deviations,
d^Bi and &2P2' ^ two inductive voltage dividers
at 10 kHz and with an input of 50 volts RMS.
These values of da< and d $ for each dividar
were measured with this circuit using a third
divider as a reference. Relative deviations
were calculated from these results and listed in
the table as "computed" d^(o<2-c*,) and d^(@2~
for comparison with those determined by a
direct comparison of the two test dividers. The
latter are listed as "measured" d^(o<2"0<i) an<*
dl(<?2-Pl>- '
The computed and measured results listed in
the table agreed to 0.02 ppm for d|(o<2"Q'i) an!*
0.1 ppm for d1(^2-Pl)The voltage ratio is defined by the equation
^OUT
VIN
and phase angle by

CONCLUSIONS
A high accuracy comparator has been
developed for measuring the relative deviation
in voltage ratio and phase angle of Inductive

voltage dividers from I to 10 kHz. An overall

accuracy of better than one part in 10^ is
achieved.
The new technique overcomes several inherent
limitations of the existing comparison methods
and has a broader frequency range . It requires
only a few pieces of special equipment, and these
are easy to construct. This circuitry appears
worthy of general use in standardizing labora
tories.
ACKNOWLEDGMENT
The author acknowledges the valuable help
and advice received from Mr. Robert D. Cutko3ky.
Mr. R. J. Berry contributed skillfully in
constructing the components.
REFERENCES
(1) Hill, J. J., and Miller, A. P., "A SevenDecade Adjustable -Ratio Indue tively -Coupled
Voltage Divider with 0.1 Part Per Million
A-curacy," JIEE, Vol. 109, Part B, pp. 157162, March 1962.
(2) Sze, W. C, Dunn, A. F., and Zapf, T. L.,
"An International Comparison of Inductive
Voltage Divider Calibrations at 400 and 1000
Hertz," IEEE Trans, on Instr. and Meas.. Vol.
IM-14, No. 3, pp. 124-131, September 1965.
(3) Cutkosky, Robert D., Is with the Absolute
Electrical Measurements Section, NBS.
(4)

"Electrical Calibration Ssrvice ExtendedInductive Voltage Dividers Calibrated up to

10 kHz," NBS Technical News Bulletin, Vol.
49, No. 1, January 1965.

(5) Thompson, A. M., "A.C. Bridge Methods for

the Measurement of Three -Terminal Admit
tances," IEEE Trans, on Instr. and Meas.
Vol. TM-13, No. "4, pp. 189-197, December
1964.
(6) Cutkoaky, R. D., "Evaluation of the NBS Unit
of Resistance Based on a Computable
Capacitor," J. Res. NBS. Vol. 65A, No. 3,
p. 147-158, May-June 1961.
(7) Gulldner, L. A., and Edsinger, R. E., "The
NBS Gas Thermometer II. Measurement of
Capacitance to a Grounded Surface with a
Transformer Ratio-Arm Bridge," J. Res. NBS,
Engrg. and Instr., Vol. 69C, No. 1, pp. 1318, January-March 1965.
(8) Clothier, W. K., National Standards
Laboratory, Chippendale, N.U.W., Australia,
private communications.

364-3

TABLE

"l

df*l

PPM of Input
Computed Measured
d2>2
diVV

.X
.9
.8
.7
.6
.5
.4
.3
.2
.1

+0.15
+1.27
+1.22
+2.21
+1.90
+1.63
+1.17
+0.41
+0.34
+0.11

+0.16
-0.23
-0.31
-0.33
-0.42
-0.27
+0.03
0.00
0.00
+0.0'+

+0.01
-1.50
-1.53
-2.54
-2.32
-1.90
-1.14
-0.41
-0.34
-0.07

+0.01
-1.51
-1.53
-2.53
-2.32
-1.90
-1.14
-0.42
-0.36
-0.08

tA
-0.4
-0.4
-0.3
+0.1
-0.1
0.0
-0.1
-0.4
-0.3
-0.3

-0.3
-0.3
0.0
0.0
-0.2
-0.4
-0.4
-0.5
-0.6
-0.5

+0.1
+0.1
+0.3
-0.1
-0.1
-0.4
-0.3
-0.1
-0.3
-0.2

0.0
+0.1
+0.3
-0.2
-0.2
-0.4
-0.3
-0.2
-0.3
-0.2

.OX
.09
.08
.07
.05
.05
.04
.03
.02
.01

-1.03
-0.96
-0.99
-0.83
-0.96
-0.70
-0.43
-0.09
+0.02
-0.01

+0.21
+0.18
+0.15
+0.12
-0.16
-0.01
-0.21
-0.01
+0.01
0.00

+1.24
+1.14
+1.14
+1.00
+0.80
+0.69
+0.22
+0.03
-0.01
+0.01

+1.25
+1.16
+1.15
+1.01
+0.81
+0.71
+0.22
+0.03
-0.01
+0.01

-0.2
-0.2
-0.3
-0.2
-0.4
-0.3
-0.2
-0.1
0.0
-0.1

-0.4
-0.4
-0.4
-0.3
-0.4
-0.3
-0.4
-0.2
-0.1
-0.1

-0.2
-0.2
-0.1
-0.1
0.0
0.0
-0.2
-0.1
-0.1
0.0

-0.2
-0.2
-0.1
-0.1
-0.1
0.0
-0.2
-0.1
-0.1
0.0

.OOX
.009
.008
.007
.006
.005
.004
.003
.002
.001

-0.01
-0.01
+0.01
+0.03
+0.04
+0.05
+0.06
+0.07
+0.09
+0.10

0.00
-0.01
0.00
0.00
+0.01
+0.01
+0.01
+0.01
+0.02
+0.02

+0.01
0.00
-0.01
-0.03
-0.03
-0.04
-0.05
-0.06
-0.07
-0.08

+0.01
0.00
-0.01
-0.02
-0.03
-0.04
-0.05
-0.06
-0.07
-0.08

-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1

-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1
-0.1

0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0

3.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0

.ooox

+0.11

+0.02

-0.09

-0.09

-0.1

-0.1

0.0

0.0

365-4

PPM of Input
Computed Measured

Photograph 1

366-5

367-6

368-7

He

?T2 ^(|

Fig. l-o

Fig. l-b

Figure 1-* Schematic Circuit Diagram

Figure l-b R-C Network

NATIONAL
WASHINGTON. O.C. !
COMPARISON CIRCUIT
VOLTAGE DIVIDER CALIBRATION
H. P K.
SZE
wtALiiuiw
wni

iM
*y

1211.06
4- 16 66
369-8

66-04-16-548

|
NOMENCIATUM
J ar*
NATIONAL MMEAU OF STANDARDS
WASHINGTON. D C 20SS4
INJECTOR-DETECTOR TRANSFORMER
'voltage DIVIDER CALIBRATION
NOOIL
|TW
,L>afcH tmm -III
h"p k*"
"""""
sze'
DCCIMALS
2 *
M
FRACTION*
ANOLK
*' IUIBB
M WOT KAU DM Ml
THIS
IWV. MC

0. D. 1.5"
1.0. 10"
H -0.25"
Fig. 2.
Figure 2

Injector-Detector Transformer*

|4-I6 66 |

66-04-16-549

- PRIMARY { 100 ,UrM S-'PH)

200 turns (l-S *Hl)
^^^^.,r>,ov/50-50 turnt (5-10 kHi)
SEC0N0ARY(g)O.|Ooturn(l-5
kHi)

-Mu-METAL CASE

NOMENCLATURE:
NATIONAL MWEAU Of STANDARDS
WASHINGTON D C 20234
SHIELDED TRANSFORMER
VOLTAGE DIVIDER CALIBRATION
WCALM
TTRC
MODEL
IKCHM MtriMM
DIXIM MMI IIIIN *W>tfMl
iIMm
M.R K.
SZE
DECIMAL*
t.CM
FRACTION*
*
ANOLM
t* tuaran A*
do hot CAU nm wmm

*tV. SK. PRINTTHIS

ICMfUCO
211.06
1 4 16 66 | , 4-04-l>-4T
.

ELECTROSTATIC SHIELD
-l-mil SUPERMALL0Y TOROIDAL CORE
0.0. = 2.5"
1.0. 15"
M f 0.5" (5-10 kHz )
" I 1.0" ( 1-5 kHz)
Fig. 3
Figure 3

Shielded Transformer
370-9

Reprinted from IEEE TRANSACTIONS

ON INSTRUMENTATION AND MEASUREMENT
Volume IM-14, Number 3, September, 1965

An International Comparison of Inductive Voltage

Divider Calibrations at 400 and 1000 Hz
W. C. SZE, SENIOR MEMBER, IEEE, A. F. DUNN, MEMBER, IEEE, AND T. L. ZAPF, SENIOR MEMBER, IEEE
AbstractThe establishment and maintenance of the highest
jssible order of accuracy for the measurements of voltage ratio and
lase angle are responsibilities of the national laboratories. This
iper reports the results of an international comparison of a seven;cade inductive voltage divider by the United States National
ureau of Standards and the National Research Council of Canada,
le calibration techniques used in each laboratory are discussed in
.'tail, and the results of calibration at 400 and 1000 Hz are compared,
le accuracies of measurements are within 0. 1 ppm of input for volte ratio and 1 (setting of dials)"1 microradians for phase angle.
Introduction
N THE FALL OF 1954, inquiries were received
about the calibration of voltage ratios and phaseangle errors of six-dial inductive voltage dividers at
dio frequencies. An early test method, developed by
e, in which the accuracy of voltage ratio determinams was dependent on the dc calibration of standard
sistors and the ac-dc differences of the resistive dijer, achieved a maximum accuracy of 0.005 per cent
voltage ratio and 60 microradians for phase-angle
or measurements [l].
Since then, considerable progress has been made in
asurement techniques by the United States National
reau of Standards laboratories at Washington, D. C.
BS-W), and Boulder, Colorado (NBS-B), and the
tional Research Council of Canada at Ottawa, Onio (NRC). At present, each laboratory is capable of
ibrating a voltage divider with uncertainty within
X10~' of input for voltage ratio measurements and
hin l/(setting of dials)"2 microradians for phase;le errors at 400 and 1000 Hz.
'he inductive voltage divider that was used for this
rnational comparison had been used on several prejs occasions for 1000 Hz interlaboratory comparison
alibrations between NBS-W and NBS-B. For the
i Its reported in this paper, the divider was initially
brated at NBS-W before it was shipped to NBS-B
calibration tests. From Boulder, it was returned to
ihington for repeated measurements, sent to NRC
calibration, and then returned to Washington for
measurements. It was believed that in this way any
anuscript received January 12, 1965. This paper is National
rch Council Rept. No. 8284.
. C. Sze is with the High Voltage Sect., Electricity Div., NaBureau of Standards, Washington, D. C.
F. Dunn is with the National Research Council of Canada,
'a., Ontario.
L. Zapf is with the Electronic Calibration Center, National
u of Standards, Boulder, Colo.

changes resulting from its shipment could be detected

and, perhaps, separated from the interlaboratory com
parison.
The purposes of this paper are 1) to describe the in
strument used for the international comparison of cali
brations; 2) to discuss separately in some detail the
calibration techniques of each laboratory; and 3) to
compare the results of calibrations at the three labora
tories at 400 and 1000 Hz.
Description of the Test Divider
The inductive voltage divider used for this compari
son of calibrations is a seven-decade commercial device
[2]. It consists of seven multitapped windings on four
toroidal tape-wound cores of very high magnetic perme
ability. Each winding of the top two decades (10_l and
10~5 per step of the input voltage) is wound on a sep
arate core and encased in individual magnetic shields.
The third and fourth decades share one core, and the
last three decades share another magnetic core. The
windings are connected in a modified Kelvin-Varley
arrangement and provide a resolution of 1X10-7. The
useful frequency range claimed by the manufacturer is
50 to 10 000 Hz. The maximum input voltage is 0.35
(operating frequency) or 350 volts, whichever is smaller
NBS-W Test Method
At the National Bureau of Standards, Washington,
D. C, the test inductive voltage divider was calibrated
by direct comparison with a capacitance divider, made
up of three-terminal capacitors whose relative values
were obtained by intercomparison in a bridge having
transformer-type ratio arms.
Bridge Circuits
With the test inductive voltage divider input con
nected across one arm of the bridge circuit (as shown
schematically in Fig. 1, but simplified by not indicating
the shielding arrangements), balance was achieved by
adjusting the output taps (Ni and iV2) of the inductive
voltage dividers which were connected across the 10
percent taps ( Vb and Vc) of the ratio arm transformer.
Ci consisted of an appropriate number (A7) of units
of capacitors of the same nominal values. N depended
on the setting of the dials (D) on the test divider. C>
was larger than Cj by a factor of one decade or more,
depending on the ratio of the test divider. The in-phase
voltage balance was achieved by insertion of a current

371-124

into the detector junction of the bridge through ca

pacitor Cs. This current usually was less than 0.001
percent of the current through capacitor J^i" &. The
phase differences between the bridge arms were com
pensated by insertion of a quadrature current of appro
priate magnitude into the detector junction of the bridge
through resistor R and current divider Ci/(Ci + Ct).
These techniques had been fully described by Mc
Gregor, et al. [3].
Connection of G was then moved from D to Vb, and
the individual units Ct, making up ?an Cs were com
pared to C\ as shown in Fig. 2. Thus the capacitance
divider ratio was determined accurately by a substitu
tion method in which the exact values of G and C need
not be known. However, these must be very stable
capacitors of negligibly small voltage coefficient. As in
the previous arrangement, the in-phase and quadra
ture differences between the bridge arms were com
pensated by insertion of currents of appropriate phase
and magnitude.
All of the above steps were straightforward and could
be done in a symmetrical fashion (comparison of ca
pacitors, calibration of test divider, and recomparison)
in less than one hour per decade by an experienced
operator.
Although small three-terminal capacitors were much
easier to calibrate than small two-terminal capacitors
because of the virtual elimination of connection prob
lems, a particularly troublesome source of error arose
when the capacitance to ground was large. Cutkosky
and Thompson have shown that if the capacitor is
treated as a four-terminal network with a high permea
bility core or coaxial choke linking one ground lead with
one active electrode lead, this small but significant error
can be eliminated [4], [5].
This technique has been used throughout the NBS-W
calibrations reported in this paper. Figure 3 shows the
capacitor comparison circuit of Fig. 2 with high permea
bility cores inserted into each bridge arm and the de
tector leads.
Theoretical Relations
The equations for the voltage-ratio corrections and
phase-angle errors were developed as follows: by defini
tion, the voltage ratio of a transformer or divider is the
ratio of the secondary or output terminal voltage Kout
to the primary or input terminal voltage V, so that

Fig. 1. Voltage divider calibration circuit.

Fig. 2. Capacitance comparison circuit.

TO TEST DIVIDER

Fig. 3. Comparison circuit shown with high-permeability cm.

Figure 1 shows the general circuit for the calibtM*
of a voltage divider. For the calibration of the fir?'
10_1 decade, we arrived at the following relations:

Vout
= DU+a+jP)
V in

C, - 10C,
Va = VQ{\ + 0 + jO) or the reference voltage

where D is the nominal ratio or setting of the dials and

a is the zero-burden1 ratio correction derived from the
measured quantities. /3 is the difference in phase angle
between the input and output voltage phasors of the
divider.
1 Zero-burden corrections were evaluated by observations with
double the capacitance burden across the output terminals.

V
Vb = (1 + ab

V
Vc = ~ -~ (I + a. +-

Vt = - Va(l + ad+ jfit).

D = N/\0 setting of dial on the 10_1 decade, A'beinj^
integral number from 1 to 10. The balance equations
the network is ii+ti+*i+*a = 0. We have

372-125

t/wG + G,)[Z>(1 + a+jfi] + ViNtiaCt + ft)

^c, + l)o.
ce on, and & < 1 X lO"' and ab, ae, and & <5 X 10-
1 Ni and Nt readings are usually small, the terms inving the products of these figures are insignificant
1, therefore, neglected. G's shown in the above equal are the conductances associated with the capacitors
I Cr represents the equivalent shunt capacitance of
stor R. From the network of Fig. 2, a second balance
ation is derived. By solving the two equations simuleously, omitting insignificant terms, and separating
id real terms, we have
a = at + Si
NC
(at, - Z ay)
A'C,(C, + C4)

0 = 06 +

c,
(X>i'- at.).
uNRC^C, + O

ilar equations could be developed for the calibration

le lower decades. It will be noted that the in-phase
nee expression contained a term representing the
t of the equivalent shunt capacitance Cr across the
tor R, used for quadrature balance. In any practical
, this Cr would never be more than a few parts in
and, thus by careful selection of the resistance elet and shielding used, it was easily made negligible.
ription of Apparatus
le usefulness of the bridges described in Figs. 1 and
>ends to a great extent on the quality of the ratiotransformer, and particularly on the 10 to 10 and
> 1 ratios of the secondary windings. The method
at NBS-VV for the precision measurement of these
s has been fully described by Cutkosky [6]. The
former used in the bridges was constructed in a
ar manner to the ones described by McGregor, et
nd Dunn [3], [7].
e dividers, used to achieve the balance of the bridge
sertions of in-phase and quadrature currents into
letector junction, were high quality commercial
-decade inductive voltage dividers. Their small
ge ratio and phase-angle errors could be neglected
e following reasons: 1) the dividers were connected
e lO percent taps of the ratio-arm transformer;
; capacitance values of capacitors G and C% would
j usted so that the setting of A/2 would be very near

zero; and 3) the loss angles of three-terminal capacitors

were found to be so small that the use of a current
divider CS/(C+C4) was necessary in order that the
setting of Ni be made large enough to be useful.
The uniplanar-suspension capacitors (Fig. 4), a modi
fication of a design by Clothier, were used for the capaci
tance divider [8]. They were of a coaxial cylindrical
construction with dry air as dielectric between the active
electrodes. Each one of the active electrodes was sup
ported by three inch diameter sapphire balls placed
120 apart in a massive brass mounting block. The
maximum adjustment of the threaded trimmer elec
trode could produce only a small change in total capaci
tance, and coarse trimming was accomplished by a
lathe cut across the open end of one of the electrodes.
Eleven capacitors were buried in the same mounting
block, of which a decade was made up of ten capacitors
that could be separately connected or summed. Each
capacitor could be adjusted with a little care and
brought to the same value to within a few ppm. The
purpose of the block was to achieve a mounting that
was relatively strain free and isothermal, and to achieve
reasonable attenuation of ambient temperature and
pressure changes. The individual units of this decade
have not been readjusted since the initial adjustment
which brought all ten very close to the eleventh in value.
A ratio check has been made from time-to-time with
this decade, and the usual day-to-day variations
amounted to one part in 10'. Since the temperature co
efficient of capacitance of these units was approximately
20 ppm/C, the small variations must mean that
changes in ambient temperature were well attenuated
and that the temperatures within the mounting block
were nearly uniform on each measurement, in spite of
the fact that room temperature moved through about
1C range. It also must mean that the individual units
were relatively stable dimensionally. Decades of 1 pF
per unit and 100 pF per unit were also constructed on
the same principle. Surfaces of all capacitor parts were
gold plated before their final assembly.
The detector circuit consisted of a preamplifier and a
cathode-ray oscillograph. An LC tuned circuit having a
band-pass of about 7 Hz was used ahead of the pre
amplifier. In this way, sharp discrimination against
noise was introduced before the first stage of ampli
fication.
The cathode-ray oscillograph was used to indicate
bridge balance; but the usual elliptical display pattern
was modified as suggested by Clothier [9]. The ampli
fied unbalance voltage from the detector arm of the
bridge is impressed on the vertical plates of the CRO.
On the horizontal plates is a synchronous signal whose
phase can be adjusted so that angular departure of the
major axis of the elliptical display pattern represents
magnitude unbalance and opening of the minor axis
quadrature unbalance. At balance the figure appears to

373-126

and for n- units in parallel (n being a running integer

from 1 to 10).
ELECTRODE
G.r = nG,(\ + axi + asi) + (DGm - DJ
i
- nwCrtfxi + M
C,r = nC,(\ + axi + aai) + (c- - DcJ
i
nGr
+
08xi + 0si)
a>

SIX
1/8" DIA.
ELECTRODE
SAPPHIRE BALLS
Fig. 4. 10 pF uniplanar-suspension capacitor.
be a single line which rocks about the horizontal as a
result of random noise. In Clothier's modification a
repetitive square-wave signal is superimposed on the
horizontal plates, so that the elliptical pattern is
doubled. The square-wave signal can be adjusted until
the right tip of one ellipse just touches the left tip of
the other, and at balance the adjacent ends appear to
move past each other as a result of noise. An average
horizontal position (i.e., balance) can now be estimated
much more accurately than with a single ellipse, and an
unbalance signal observed which is lower by almost an
order of magnitude.
NRC Test Method
At the National Research Council, Ottawa, Canada,
the test inductive voltage divider was also calibrated by
direct comparison with a capacitance divider, using a
transformer ratio bridge whose ratio, with the burden of
the divider and measuring components, is established in
the sequence of balances [7].
Ratio of Capacitance
The procedure begins by establishing known ratios of
the capacitance divider. Eleven nominally equal threeterminal capacitors are assembled as a unit in a mount
ing block, of which a decade is made up of ten that can
be individually connected or summed (the subscript m
designates individual units and n, n units in parallel).
The eleventh capacitor is represented by symbols Gr
and Cr- In Fig. 5, the test divider is connected to the
Fs=l/1 tap of the bridge transformer with a coaxial
cable to establish a burden and to determine the zero
reference balance (Dc, Doo) of the bridge with dials
Dc and Da- Then capacitors Cr and C,r are connected
to the bridge as shown and each Cmr is measured in
terms of Cr- The balance equations are
GmF = G,(l + axi + asi) + (D0m - Da.)
(oCrifixi + Psi)
Cm, = Cr(l + axl + asi) + (DCm - Dc.)
Gr
+ 0xi + fisi),
a)

where axi and aSi represent the in-phase ratio correc

tions of the Fx =1/1 and Fs=l/1 taps of the bri<if-!
transformer, while An and &si are the respective q-idrature corrections. Higher order terms of small quantties are neglected.
Capacitors with nominal values of a decade higher c
lower can also be compared by selecting the appropr.r.e
ratio taps on the bridge transformer [7J.
Calibration of the Test Divider
Figure 6 shows the general circuit diagram for i
calibration of a voltage divider, whose true zero-bunk:
voltage ratio is defined as before as
Vout

e Eo

The errors resulting from additional burdens imp

by the impedance of the coaxial cables and circuit cc
ponents on the outputs of both the test divider as:
bridge transformer must be applied as correction te:_to a and 0 of the bridge balance equation.
The following equations, Da and D(3, w ith all thes;
nificant corrections added to the general bridge balar;:
equation, represent the corrections to the nth tapo:::first dial of the test divider:
n
n
1 f" n
Da = a = (asi "si') H
(Dc
10
10
loc, Lio

374-127

+ ( E Dc - ^ E Dc}} - (Den' - Dcr'j

Cs + Co

nC r
(Dc' - DCn") +
Cs

nCr
(Dc* - Dc*) +
Cs + Cos
(Dc* - Dc**)
C

10

10
-o*l/IO
z>o~ -
i
10 i

- (JW - *W)
/
-oxIO

Cs +
? Cos UW -^.")+
nCf

nCr
Fig. 5. Comparison of capacitors.

(Do.* - Do.*) +
Cs + Cos
c7~

- Oft.**)].

The term n(Dcs Z>c<>)/10 represents the correction

he nominal ratio of Cs to Cf, when n = 10, as deterled from the capacitance ratio measurement and

ie correction to the nominal ratio of the capacitance

der determined previously.
'ceo' is the dial setting for the reference zero balance
he calibration. This zero offset is caused by the immce of the outer conductor of the coaxial cable con
ed from terminal 0 of the test divider to the comi ground point of the bridge as shown in Fig. 6. The
i {DCn' Dceo') represents the apparent or unected a.
he presence of Cs from the output tap through the
ctor to the common ground point effectively shunts
of the test divider and changes the total impedance ;
rn, depending on the dial setting D, it also changes
)urden on the bridge transformer. The magnitude of
effect can be determined by connecting another
citor Cs*(Cs* Cs) between terminal E and the
tor junction point and by connecting Cs between
livider output terminal and the common ground
. The bridge is then balanced, with bridge diallg Den* for each setting/) of the divider dials. When
, there is no additional burden effect and this is
ed as a subsidiary reference point Deo*considering the measurement of the effect of the
;ns due to the components generally associated
the bridge, the fact that the test divider has an
alent output impedance (represented by Ro+juLo
g. 7) which varies with the dial setting D, must
e overlooked. The connections and measurements
described represent the in-phase and quadrature
tions of the divider when burdened by (Cs + Cos),
ffect of this burden is given by
io (zero burden) =

,-411 o - 88 - 25

e Eo
E - E

-(1 + a+y/J)

Fig. 6. Calibration of inductive voltage dividers.

DETECTOR
Fig. 7. Inductive voltage divider with external burden.
and
Ratio (with burden) =

^
^
= (1 + a + jff)
E o 10

e E<>
[1 +
E - 0

- u-io.XCs + Cos)].

When an additional burden Ch is added to Cos, the

ratio becomes
e" - E0
E- Eo

n
(l+a"+jp").
10

These ratio expressions may be combined to give

a = a' +

Cs + Cos
(a' - a").
Ch

It is necessary to make the basic measurements Dc'

and the subsidiary measurements DCn* in order to deter375-128

mine the effective ratio of the bridge transformer at E,

and the additional measurements Z?c" and DCn** for
each setting D, with Ch connected to take account of the
residua] output impedance of the test divider.
The coaxial cable between terminals Fs of the trans
former and E of the divider will increase the effective
leakage impedance of the transformer at E, which is
chosen as one of the reference points of the calibration.
The term (otsi as/) represents the change of the bridge
transformer ratio from Fs = 1(1 +asi+j&si), effective at
the transformer tap, to Fs' = 1(1 +a'+J/Ji')i effective
at terminal E of the divider.
By combining the basic equation of balance with the
subsidiary equations and the initial capacitance ratio
calibrations, a pair of equations (Da and D$) are ob
tained relating the zero burden corrections to a series of
observed bridge balances. All terms in these two equa
tions are measurable quantities except for (asi <*si')
and (jSsi j3si'). Values are assumed for these terms in
the final tabulation which make ai = /3 = 0, as assumed
for the test divider initially.
Similar equations can be developed for the calibration
of the other dials of the divider.
NBS-B Test Method
At the National Bureau of Standards, Boulder, Colo
rado, the test inductive voltage divider was calibrated
by direct comparison with a "standard" divider [10].
The voltage ratios and phase angle errors of the "stan
dard" divider were determined, periodically, by the best
method available.
Early Method
A transformer capacitance bridge method was used
from 1959 until late 1963 for 1000 Hz measurements of
the voltage ratios and phase-angle errors of the "stan
dard" divider. This method was completely described
by Zapf and played no part in this international intercomparison [ll]. Therefore, there is no need to review
it here. However, it did serve as the basis for all previous
calibrations of the test divider, thereby providing an
independent means of checking on the results reported
in this paper. The overall uncertainty in voltage ratio
calibration by this method was believed to be less than
2X10-'.
Present Method
This method, under development since 1961, required
the analysis of complex equivalent circuits and experi
mental verification of several basic assumptions before
it was considered operational late in 1963. It provided
the voltage ratio and phase-angle error data that were
used in the NBS-B contribution to the international
comparison. The method has been described, and it is
necessary only to mention it briefly [14].
Internal loading is the cause of major systematic er
rors in inductive voltage dividers. The effects of internal
loading can be controlled or predicted by a judicious

choice of winding design [12]. Experimentation with

several winding designs, and one in particular, demo:
strated that a graph of errors (at the various taps of i
divider) vs. the nominal voltage ratios fell on an Sshaped curve. As expected, it was found that a parti;,
lar winding design used repeatedly, but with variations
in wire size, number of turns, and core characteristicyielded errors that fell on 5 curves of the same ships
but different amplitude. The task of predicting tfc
shape of the Scurve, with probable accuracy surpass^
that which could be experimentally measured, was ac
complished by solving the network equations for a carfully derived, complex, equivalent circuit represent;
the chosen design [13 ].
To measure the amplitudes of the 5 curves of varies
dividers, a method of complementary measuremer.was developed to accurately measure the true voltas
ratios and phase angles at the nominal voltage rats
0.25. (The complement of a voltage ratio is obtained :
simply reversing the connections to the input o: .
divider.)
To summarize, the method is based on a particifc
design of inductive voltage divider for which the intern,
loading errors are calculable and measurable. Corrects
for these major systematic errors in particular divfe
leaves residual errors of much smaller magnitude. Tte
is both the expectation and experimental evidence to
these residual errors are the result, primarily, of imper
fections of random nature in the construction of e*
divider, and secondarily, of comparatively very soil
imperfections in the theoretical model on which the cal
culations are based. Therefore, by submitting the R
suits from a large number of dividers to statistical pr>
cessing, it is possible to improve upon the determinate'
of the errors. The use of a group of dividers and statis
tical averaging reduces the effect of individual "par. .
lar errors" on the establishment of voltage ratios ai
phase angles.
The method outlined above was used at frequencies
1000 and 400 Hz to calibrate a "standard" inducts
voltage divider. At 1000 Hz ten 20-section dividers r
used for establishing the corrections. At 400 Hz tttfa
20-section dividers were used. The work at 1000 produced results in agreement (a few parts in 10'?
input) with results previously obtained by the cap;
tance bridge method, which is smaller than the dife
ences between results from the NBS-B and either of w
two other laboratories.
The mean of a group of ten or more was used for estab
lishing the NBS-B values, and these were used in
ways, with input connections normal and reverse:
thereby improving the results by a factor of four or t '
over individual dividers measured one way. A sunt
the data taken at each ratio reveals that the stan:-':
deviation from the mean for the residual random er '
at the taps of individual dividers in the group did:
exceed, at any ratio, a voltage ratio of 3.4X10"' tor "
1000 Hz measurements, and 3 1 X 10"' at 400 F-

376-129

TABLE I
1000 Hz
NBS-W
Setting

NBS-B

0 X00000 0
0.900 000 0
0.800 000 0
0.700 000 0
0.600 000 0
0.500 000 0
0.400 000 0
0 .300 000 0
0 .200 000 0
0.100 000 0
0 .0X0 000 0
0.090 000 0
0.080 000 0
0.070 000 0
0 .060 000 0
0.050 000 0
0.040 000 0
0.030 000 0
0.020 000 0
0.010 000 0
0.00X 000 0
0.009 000 0
0.008 000 0
0.007 000 0
0.006 000 0
0.005 000 0
0 004 000 0
0.003 000 0
0.002 000 0
0.001 000 0

Da(10-)
+0.02
+0.14
+0.19
+0.08
+0.03
+0.00
-0.10
-0.26
-0.18
-0.13
-0.15
-0.20
-0.18
-0.18
-0.14
-0.12
-0.11
-0.08
-0.06
-0.02
-0.02
-0.02
-0.02
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01
-0.01

D/3(10-)
0.0
-1.8
-1.6
-2.4
+ 1.2
+4.0
+6.8
+8.1
+8.0
+5.4
+4.5
+3.6
+3.0
+2.7
+2.5
+2.2
+2.0
+ 1.8
+ 1.4
+0.84
+0.78
+0.72
+0.53
+0.43
+0.37
+0.37
+0.35
+0.33
+0.27
+0.16

0 X00000 0
0.900 000 0
0.800 000 0
0.700 000 0
0 .600 000 0
D.50O 000 0
5.400 000 0
3.300 000 0
) 200 000 0
MOO 000 0
( 0X0 000 0
1.090 000 0
1.080 000 0
'.070 000 0
.060 000 0
.050 000 0
.040 000 0
030 000 0
.020 000 0
.010 000 0
.00X000 0
.009 000 0
.008 000 0
.007 000 0
.006 000 0
.005 000 0
004 000 0
003 000 0
002 000 0
001 000 0

0.00
+0.09
+0.12
0.00
0.00
+0.08
+0.04
-0.07
-0.05
-0.05
-0.15
-0.08
-0.08
-0.07
-0.06
-0.05
-0.04
-0.03
-0.02
-0.02
-0.02
-0.04
-0.02
0.00
-0.02
-0.02
-0.02
-0.01
0.00
0.00

0.0
-0.9
-1.6
-0.7
0.0
+ 1.5
+2.4
+3.3
+3.2
+2.1
+2.0
+ 1.7
+ 1.4
+ 1.2
+ 1.1
+ 1.0
+0.88
+0.75
+0.60
+0.34
+0.34
+0.30
+0.24
+0.20
+0.19
+0.17
+0.16
+0.14
+0.11
+0.06

DoUO-')
+0.01
+0.06
+0.10
+0.05
-0.01
-0.06
-0.13
-0.32
-0.20
-0.18
-0.20
-0.17
-0.15
-0.14
-0.11
-0.11
-0.09
-0.08
-0.07
-0.04
-0.05
-0.04
-0.03
-0.02
-0.03
-0.03
-0.03
-0.02
-0.01
-0.01
400 Hz
+0.03
+0.11
+0.15
+0.12
+0.11
+0.12
+0.08
-0.06
0.00
-0.04
-0.12
-0.11
-0.07
-0.08
-0.06
-0.05
-0.04
-0.03
-0.02
0.00
-0.01
0.00
+0.01
0.00
0.00
0.00
+0.01
+0.01
0.00
+0.01

377-130

NRC
D0(lO-)
-0.10
-0.54
-0.16
+ 1.26
+3.42
+6.0
+8.0
+9.0
+8.4
+5.4
+4.7
+4.0
+3.4
+3.1
+2.8
+2.5
+2.2
+ 1.9
+ 1.4
+0.84
+0.76
+0.67
+0.59
+0.55
+0.51
+0.47
+0.44
+0.37
+0.30
+0.20

Da(10-)
-0.04
+0.07
+0.11
+0.08
+0.03
+0.02
-0.03
-0.19
-0.10
-0.11
-0.12
-0.12
-0.11
-0.12
-0.10
-0.09
-0.08
-0.07
-0.05
-0.03
-0.03
-0.04
-0.04
-0.04
-0.05
-0.05
-0.04
-0.04
-0.03
-0.02

D/S 10-)
-0.06
-0.93
-0.97
+0.37
+2.48
+4.90
+7.00
+8.14
+7.85
+5.25
+4.59
+3.94
+3.34
+3.08
+2.67
+2.32
+2.09
+ 1.81
+ 1.40
+0.81
+0.74
+0.63
+0.55
+0.50
+0.45
+0.42
+0.38
+0.33
+0.26
+0.15

0.00
-0.54
-0.48
+0.14
+ 1.02
+2.0
+2.8
+3.3
+3.2
+2.2
+2.1
+ 1.7
+ 1.4
+ 1.3
+ 1.1
+ 1.05
+0.96
+0.84
+0.64
+0.37
+0.37
+0.32
+0.28
+0.25
+0.23
+0.21
+0.20
+0.17
+0.13
+0.08

+0.01
+0.04
+0.07
+0.05
+0.09
+0.14
+0.16
+0.06
+0.12
+0.04
0.00
-0.01
-0.02
-0.02
-0.01
0.00
+0.01
+0.01
+0.01
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00

+0.02
-0.59
-0.63
+0.09
+0.89
+1.91
+2.77
+3.23
+3.16
+2.10
+2.06
+ 1.74
+ 1.48
+ 1.39
+ 1.15
+ 1.03
+0.92
+0.79
+0.60
+0.35
+0.35
+0.30
+0.25
+0.23
+0.20
+0.18
+0.15
+0.14
+0.10
+0.06

Therefore, the standard error of the result was 0.8 X 10-8

or less at 1000 Hz, and 0.7 X 10r* or less at 400 Hz.
In the establishment of voltage ratio, residual (i.e.,
uncorrected) systematic errors greater than + 2X10-8
are believed to be very unlikely. These limits of sys
tematic error are based on subjective considerations.
Thus, the uncertainty in the establishment of voltage
ratio is less than 5X10-8 computed by adding three
times the standard deviation to the limits of systematic
error.
The voltage ratios are preserved, or maintained, by
means of a "standard" inductive voltage divider. The
stability of this "standard" divider has been observed
over a period of years. The divider used for the inter
national comparison was calibrated by comparison with
the "standard." It is believed that a proper allowance
for systematic error in the maintenance of voltage ratio
is + 2X10-*, and for random variations a standard de
viation of 2X10-8. In the comparison with the "stan
dard" divider an additional allowance of + 2X10-8 for
systematic error is warranted, and the standard devia
tion of the comparison method is about 0.7 X10~8.
The three sets of limits of independent systematic
error mentioned above are +2, 2, and 2 (all X10-8).
A realistic combination of these limits might be taken
as +5X10-8. The three standard deviations mentioned
above are 0.8, 2, and 0.7 (all X10-8), and the square
root of the sum of the squares is about 2.3X 10~8. These
figures may be taken as a partially subjective estimate
of the uncertainty in the calibration results at 1000 Hz
on the top decade, describing errors that could be ob
served if the entire experiment were to be repeated by
the same method, but with a different group of 20section dividers constructed to the same specified
design.
Adding the limits of systematic error and three times
the standard deviation yields 1.2X10-7. These limits
express concisely the uncertainty in voltage ratio asso
ciated with the results of the calibration at 1000 Hz at
NBS-B and are strictly applicable only to the top dec
ade. The uncertainties of results on the second and
third decades are about the same as for the first dec
ade; the expected improvement is offset by the necessity
for additional comparison measurements.
Comparison of Calibration Results
Table I lists the voltage ratio corrections (Da) and
phase-angle errors (Dfi) observed for the test divider in
the three national laboratories for 400 and 1000 Hz. All
values were obtained with the case connected to the
low voltage terminal of the input and with an input of
100 volts RMS (except 400 Hz data for NRC, which
was performed at 40 volts input).
Values reported by NBS-W represent the average of
three sets of measurements, which were taken before,
between, and after the measurements taken by the
other two laboratories. The maximum spread was
0.14X10-6 for the ratio correction at 0.5 setting and

1000 Hz. Spread on any*other setting was much smaliei

than 1 X 10-7 and appeared at random. Therefore, it cat
be said that no apparent shift in either ratio or phase
angle was noted after the divider was exposed to the
hazards of travel and use.
The voltage ratio corrections for the lower fojr
decades (10~4 to 10~7) were less than 1 X 10"8. Therefore,
they are not included in this intercomparison of ca.ibration report. The phase-angle errors associated wit*
these lower decades are of secondary importance aad
perhaps of lesser general interest; thus, they are ai*
omitted.
Conclusion
This is the first international comparison of induct re
voltage divider calibration in which United States and
Canadian national laboratories have participated. It a
interesting to note the close agreement in the resuks
obtained by three independent laboratories using com
pletely different methods. The close agreement indicates
that the voltage ratio and phase-angle measurements it
each of the three laboratories are accurate to within
+ 0.1 ppm of input and 1 /(setting of dials)1'1 microradians at 400 and 1000 Hz.
Acknowledgment
Mr. Sze wishes to acknowledge the valuable guidaia
and help of Dr. F. K. Harris, Chief of the Absolute Da
trical Measurements Section, Electricity DiviaoNBS, Washington, D. C.
References
[11 W. C. Sze, "Measurement of voltage ratios at audio frequenr?:
Trans. AIEE (Communication and Electronics), vol. 76, pp. r^449.
[2] Catalog Sheet C-25 (Voltage Divider), Electro Scientific Iocjtries, Inc., Portland, Ore.
[3] M. C. McGregor, et al., "New apparatus at the National E r
of Standards for absolute capacitance measurement." Trans, on Instrumentation, vol. 1-7, pp. 253-261, December
[4] R. D. Cutkosky, "Evaluation of the NBS unit of resisUbased on a computable capacitor," J. Res. NBS, vol. 65A
147-158, May-June 1961.
[5] A. M. Thompson, "AC bridge methods for the measureme-"
three-terminal admittances," IEEE Trans, on Instrumm- ~
and Measurement, vol. IM-13, pp. 189-196, December W4
[6] R. D. Cutkosky and J. Q. Shields, "The precision measurene
of transformer ratios," IRE Trans, on Instrumentation, vm
pp. 243-250, September 1960.
[7] A. F. Dunn, " Determination of an absolute scale of capari'-a.-:Can. J. Phys., vol. 42, pp. 53-69, January 1964.
[8] W. K. Clothier, National Standards Lab., Chippendale, .V.tralia, and F. K. Harris, National Bureau of Standards, W--"
ington, D. C, private communications.
[9) \V. K. Clothier, NSL, Chippendale, Australia, and R
Cutkosky. NBS, Washington, D. C, private communicr*
[10) R. V. Lisle and T. L. Zapf, "Comparison calibration of induvoltage dividers," ISA Trans., vol. 3, no. 3, pp. 238-242. I1*[111 T. L. Zapf, "Voltage ratio measurements wfith a trans
capacitance bridge," J. Res. NBS, vol. 66C, pp. 25-32. Jaui"
March 1962.
[12]
, "The calibration of inductive voltage dividers and -3.-' ^
of their operational characteristics, ISA Trans., voL 1. i>
pp.L.195-201,
[13] T.
Zapf. C.1963.
H. Chinburg, and H. K. Wolf, "Inductive v.-
dividers with calculable relative corrections." IEEE Trz*
Instrumentation and Measurement, vol. IM-12, pp. S*'
September 1963.
[14] T. L. Zapf, " The accurate measurement of voltage rack*
ductive voltage dividers," 1964 Acta IMEKO, ~-t .
Sweden, pp. 317-331.

378-131

U.S. Department of CommerceBureau op Standard*

RESEARCH PAPER RP580
Part of Bureau of Standards Journal of Research, Vol. 11, July 1933
EQUIPMENT FOR TESTING CURRENT TRANSFORMERS
By Francis B. Silsbee, Ray L. Smith. Nyna L. Forman, and John H. Park
ABSTRACT
The equipment and test procedure developed at the Bureau of Standards for
measuring the ratio and phase angle of current transformers up to currents of
12,000 amperes at power frequencies are described in detail. Data are given to
show the accuracy of the standard current transformer used in the higher ranges,
and the effectiveness of the shaping of the heavy-current circuit so as to
minimize errors from stray magnetic fields.

IEEE Transactions on Instrumentation and Measurement

Volume IM-14, Number 4, December 1965

for
BERNADINE L. DUNFEE, senior member, ieee

bstractThe design, construction, and performance of two

y identical, multirange current transformers for operation 400
> 10 kHz and with errors of only a few ppm are discussed. A
al discussion of the parameters that influence the performance
agle-stage transformers and their consideration in effecting
num errors throughout a wide frequency band is included. Con
don of the two standards having consecutive ratios 1 to 1 to
1 and operating at a rated secondary current of 5 amperes is
ibed. Circuit diagrams are used to describe 1) the self-calibra:ircuit for measuring the errors of the 1-to-l ratios and 2) the
I) circuit together with an auxiliary circuit for measuring the

errors of ratios >1. Balance equations are derived in the Appendix.

Measured values are presented; included are the results up to 16
kHz obtained at the National Research Council, Canada, during an
international comparison and described in a companion paper [6].
The effects of polarization at low frequencies are emphasized by util
izing the 400 Hz data together with information obtained in supple
mentary experiments; also the role of winding and interwinding
capacitances and the necessity for a more explicit definition of cur
rent ratio at the higher frequencies is highlighted.

379-1

IEEE Transactions on Instrumentation and Measurement

Volume IM-12, Number 2, September 1963
Inductive Voltage Dividers with Calculable
Relative Corrections
THOMAS L. ZAPF , senior member, ieee, CARL H. CHINBURG , senior member, ieee,
and HARRY K. WOLF , senior member, ieee

SummaryInductive voltage dividers are now used in measure

ment laboratories for the production of audio-frequency voltage
ratios with errors (deviations from turns ratio) of a few parts in ten
million of input. A major component of error arises from the interac
tion between distributed shunt impedances and leakage impedances
in the windings. Correction for this systematic error in dividers of
special design can result in an order-of-magnitude improvement in
accuracy.
A solution using network equations has been obtained for the cor
rections to the relative errors of inductive voltage dividers of specific
design. Earlier theoretical considerations, confirmed by measure

ments of limited accuracy, indicated an S-shaped curve of ratio

error vs nominal ratio, and quadrature component vs nominal ratio.
The results from recent calculations are in agreement with the earlier
measurements and provide much better definition of the true shape
of the S-curve. An algebraic equation has been derived for the limit
ing form of this S-shaped characteristic curve. A resistance analog
of an inductive voltage divider was constructed to represent the
lumped circuit parameters equivalent to the distributed shunt ad
mittances and the winding impedances. Measurements of the gross
errors in the analog have yielded experimental results in excellent
agreement with those calculated from the network equations.

ISA Transactions
Volume 2, Number 3, July 1963

The Calibration of

Voltage

Inductive

Dividers and

Analysis of Their Operational

Characteristics

THOMAS L. ZAPF
Radio Standards Laboratory
National Bureau of Standards
Boulder, Colorado

y A review of recent work in alternating-voltage ratio measurements at the National

Bureau of Standards is followed by a discussion of equipment and methods of measure
ment now under investigation. Improvements in the accuracy of measurement of
alternating-voltage ratio by capacitance bridge methods have reduced the overall
limit of uncertainty in the calibration of voltage ratio of inductive voltage dividers to
less than +2x I0~7. Calibrated inductive voltage dividers have been used as stan
dards for the calibration of other dividers by a comparison method with relatively
little degradation of accuracy. Recent investigations on the design and construction
of inductive voltage dividers have demonstrated the feasibility of compensation of
ratio errors arising from internal loading, and several inductive voltage dividers have
been designed and constructed that have ratio errors less than + 5x I0~8 at a frequency
of 1000 cycles per second.

380-2

High

Voltage

and

Surge

Measurements

Papers
Special shielded resistor for high-voltage d-c measurements, J. H. Park.
Shunts and inductors for surge-current measurements, J. H. Park
Spark-gap flashover measurements for steeply rising voltage impulses,
J. H. Park and H. N. Cones

383
389
413

Abstracts
The measurement of high voltage, F. M. Defandorf
381

424

JOURNAL OF RESEARCH of the National Bureau of StandardsC. Engineering and Instrumentation

Vol. 66C, No. 1, January-March 1962

Special Shielded Resistor for High-Voltage

D-C Measurements
J. H. Park
(September 26, 1961)
A new design for an accurate high-voltage d-c standard resistor has been devised. It
is made up of a large number of individually shielded, one-megohm wire-wound resistors
connected in series and arranged to form a vertical helix between a ground plate and a highvoltage electrode. The individual shields completely enclose each one-megohm resistor and
prevent formation of corona at the surface of the resistance coil no matter how high the
potential of the shield is above ground. The vertical helical configuration with a large
hat," or high-voltage electrode, on top serves to prevent concentration of electric field
and corona formation at the high-potential end of the resistor. A 200-megohm unit, con
structed during 1955 and tried out up to 100 kv in 1956, indicated the design to be free of
corona errors but, for the particular one-megohm resistors used, the variation with tempera
ture was quite large (0.01 percent per C). A 100 megohm unit using low-temperature
coefficient resistors has been recently constructed and tested up to 100 kv.
An experimental method of checking for corona or leakage errors at high values of
voltage was developed. It consisted of accurately comparing the current "in" at the high
voltage end of the resistor with the current "out" at the ground end for several different
values of applied voltage. These measurements together with others performed to check
temperature and leakage errors indicated that the value of resistance for the 100-megohm
unit remains constant to within about 10 ppm for voltages up to 50 kv under ordinary labora
tory conditions. At 100 kv the maximum error (caused by heating) was estimated to be
about 40 ppm.

1. Introduction
Accurate measurements of d-c voltages at values
above 10 v are nearly always dependent upon a
resistive voltage divider which consists of a high
resistance, Rlt in series with a low resistance, R2.
The voltage to be measured is connected across the
series combination with R2 at the ground end. The
divider ratio, (R1 + R2)/R2, is chosen to give about a
1 v drop across R2, which can then be measured
with a null potentiometer. This is the well known
'volt-box" method and tapped resistors or "volt
boxes" are readily available for voltages up to at
least 1,500 v. In extending this method to higher
voltages the only additional difficulty is in designing
a high-voltage resistor with an effective value which
does not change with voltage.
A change in effective resistance with voltage may
be due to any one or a combination of three factors:
(1) heating of the resistance wire due to the PR
lossthe magnitude of this change will depend
upon the temperature coefficient of the entire resis
tor; (2) current leakage through the volume or over
the surface of the insulation used to support the
resistorsuch leakage usually increases with voltage
and in effect decreases the total resistance; (3) corona

discharges which may appear at locations of high

gradient along the resistor as the voltage is
increasedin effect they "leak" part of the resistor
current to ground. The heating effect, factor (1)
can be quite bothersome unless the overall tempera
ture coefficient is very low. However, a nighvoltage resistor is most conveniently made up by
connecting a large number of one-megohm units
in series. Thus, temperature coefficient can be
reduced to a negligible minimum by choosing some
resistors with positive and others with negative
temperature coefficients. This can readily be done
if a large number of one-megohm units made up
of either Karma 1 or Evanohm 1 wire are available
and their temperature coefficients have been meas
ured. Factors (2) and (3), above, cannot be meas
ured and reduced to a low magnitude in any such
simple manner. Thus, the primary design problem
is to keep leakage and corona effects as low as
possible. Also, some experimental method must be
devised to chock for these effects with full voltage
applied to the complete resistor.

i Trade names for special alloys (NICr+Al+Fe) having high resistivity und
low temperature coefficient of resistance.

383 19

2. Helical Resistor Made in 1955

A special helical design to keep corona effects to a
minimum when a large number of 1 -megohm resistors
are connected in series was first devised in January
1955. It was made up of 200 one-megohm wirewound resistor units which had been oDtained on
surplus after World War II. Each unit consisted
of wire wound on a cylindrical spool % in. in diameter
and 1% in. long and connected to screw terminals at
each end of the cylinder. Overlapping brass caps
with ribbon insulation between them were attached
to the screw terminals as indicated in figure 1.
These two brass caps form a complete potential
shield around the unit and definitely fix the voltage
gradient environment immediately surrounding the
resistance wire. Such voltage gradients are deter
mined only by the potential difference between the
two caps which is equal to the IR drop across the
resistor. Thus the possibility of corona appearing
in the air immediately surrounding the resistance
wire is eliminated no matter how high the potential
of the resistor and its shields may be with respect to
ground.
A jig was made to support the individually shielded
units along a circular helical path. They were then
soldered together in this position and mounted on a
lucite tube 4 in. in diameter and 39 in. high as
shown in figure 2. The circular helical shape Keeps
a uniform potential drop per turn along the lucite
tube. The 24 inch diameter copper "hat" placed
on top of the lucite tube serves to intercept electro
static lines of force from objects at ground potential
and prevents concentration of such lines of force
at the top turns of the resistor helix. Thus the
gradient at the outside surface of the individual
resistor caps is very nearly the same for units near
the top as those near the bottom, and the possibility
of corona from these, shield caps is greatly reduced
or eliminated.

Figure 1. Two one-megohm wire-wound resistors, mounted

inside their shield caps, as used in first model of helical
resistor.
Indicates method of connecting resistors In series to form helix.

Figure 2.

The first model of helical resistor.

Made during 1955.

If there were any corona discharge or even "dark

current" from the resistor at some point between
the high-voltage electrode and the low-voltage lead,
it would cause the current "out" at the low-voltage
lead to be less than the current "in" at the hignvoltage electrode. This would introduce a change
in the value of effective resistance which is used 111
voltage measurements. Current leakage through
or over the surface of any insulating supports would
cause a similar type of error. An experimental
check for such errors can be obtained by getting
a simultaneous measurement on the current "in"
and "out" at various voltages up to the highest
which can be used. This was done by adjusting
the applied voltage to give a certain fixed value of
current "in" under steady conditions and measuring
the current "out".
The current "in" was set to a fixed known value by

384 -20

iissing it through a standard resistor and balancing

le IR drop to that of a standard cell by use of a
iflecting mirror galvanometer. The standard resist, standard cell, and galvanometer were placed on
>p of the high-voltage electrode and covered with
3-gal cylindrical can to prevent corona discharges
om them. The resistance of the standard resistor
ted in this circuit was chosen to give the value of
irrent desired. The current "out" was measured
the usual manner by passing it through a standard
sistor and measuring the IR drop with a null
tentiometer. The procedure then was to choose
value of standard resistor to give the required
rrent "in" and adjust this current to give a
.lance on the reflecting galvanometer mounted
side the 3-gal can. When a balance was attained
e current "out" was read on the potentiometer at
e ground end. Any discrepancy between these
o currents indicated either a leakage or corona
ect.
Some special precautions taken during these measements should be noted. The switch closing the
lvanometer circuit inside the 3-gal can must not
closed until the current lias been approximately
justed to the correct value. This requires a
notely operated switch because the 3-gal can is at
*h voltage. A waxed cord was tried first for
crating this switch but found to be unsatisfactory
cause it jarred the galvanometer off scale. A
ay operated by a photoelectric cell circuit which
aid be closed by a beam of light was installed
tide the 3-gal can and found to operate satisfac*ily. It was also found desirable to place extra
er capacitors across the high-voltage supply,
ring high RC time constants on both charge and
charge, thus serving to keep the current steady.
L'urrent measurements were made using 5,000- and
'O0-olim standard resistors in the 3-gal can giving
Itage values of approximately 40 and 100 kv
oss the 200-megohm resistor. At 40 kv the
-rent "in" and current "out" agreed to within
aut 20 ppm which was considered the limit of
mracy for this particular test. At 100 kv the
-eement could be certain only to within about
ppm because of unsteadiness in the supply voltage.
. further attempt to minimize these measurement
ors was made at that time because the expected
uracy of the first model was about 0.01 percent.
Actually the largest error in this model was found
be due to the temperature coefficient of resistance
the 1-megohm resistors. Apparently Nichrome V
e had been used in winding these resistors. The
asured change of resistance with room temperat ure
determined with low voltage applied was +0.01
cent per degree C increase in ambient temperre. The increase of resistance with applied
tage followed the square law as expected, and
s total change from low to rated voltage of 100 kv
3 0.09 percent. Correction curves were derived
changes in applied voltage and room temperature,
was concluded that, with care in applying these
rections, accuracies within 0.02 percent could be
ected for applied voltage up to 50 kv.

3. Improved Helical Resistor Built in 1961

The first model has been used as a standard for
calibrating other high-voltage resistors at the
National Bureau of Standards for several years. In
most cases the accuracy attainable was sufficient,
but in several instances accuracy needs reached 0.01
percent or better. In view of the ever higher accu
racy needs as time goes on, especially in experiments
involving the accurate measurement of atomic and
nuclear constants, it was decided to build another
resistor of the same design using 1-megohm units
with lower temperature coefficients.
In 1960 it was found that due to improvements in
resistor winding techniques and to the development
of special alloy wire of low temperature coefficient,
it was possible to obtain 1-megohm units guaranteed
to have a temperature coefficient less than 5 ppm per
degree C. A large number were ordered, and on
delivery their temperature coefficients were deter
mined by measurements at three or more tempera
tures between 20 and 44 C. They were all well
within the 5 ppm per degree C limitalso some
resistors increased and others decreased with rising
temperature and in general the change was propor
tional to temperature. Thus it was possible to
choose matched pairs of these resistors with nearly
zero average temperature coefficient. By connecting
such matched pairs in series, a high-voltage resistor
with nearly zero change with temperature was made.
To keep corona and leakage errors as low as pos
sible the same spiral type of construction as that in
the first model was used. A new technique for
making the shields and connecting them in a helix
was devised. Each shielded unit contained two
resistors and was assembled from brass parts, ma
chined out of tubing, and a 45 brass elbow, as shown
in figure 3. Polyethylene sheet insulation was used
between the overlapping shields around each resistor
as before. Eight of these units, fitted together.Jconstitute a single turn of the helix with an outside
diameter of 9)i in. They were fastened to a 7-in.diam lucite tube {% in. wall thickness) using one
nylon clamp and screw for each unit. The pitch of
the helix was chosen to prevent any possibility of
corona between adjacent turns.
SEC AA

Figure 3. A two-resislor shielded unit as used in the second

model of helical resistor.

385 -21

First an estimate of required pitch was made based

on a maximum rated voltage for each 1-megohm
resistor of 1,000 v. This gave 16 kilovolts per turn.
A pitch of 2% in. per turn was taken, giving 1% in.
between adjacent brass shields from one turn to the
next. A 1^-turn resistor (maximum rating 24 kv)
was made up using this pitch and its resistance was
measured at various voltages up to 27 kv. This was
done using a wheatstone bridge with the 1955 model
200-megohm helical resistor as the other highvoltage arm. Only gradual uniform changes in
resistance were noted as voltage was increased in
steps from 5 to 27 kv, and they could be accounted
for by heating effects. There was no evidence of
corona effects even at 27 kv, which is 125 percent of
the maximum to be used.
On the basis of these preliminary tests a 100megohm resistor of 6)4 turns was made up with the
same pitch. The lucite tube is I6J2 in. long and
mounted vertically between a brass ground plate and
a high-voltage electrode as shown by the photograph
in figure 4. The electrode on top of the lucite tube
was designed so that another similar resistor could he
placed on top of it. By stacking resistors and con
necting them in scries, voltage ranges in multiples of
100 kv could be obtained. At present only one 100megohm resistor has been constructed because so far
there has been no urgent requirement for very accu
rate measurements at voltages above 50 kv. How
ever, corona tests on the high-voltage electrode were
made up to 200 kv by using two 100 kv power supplies.
The normally grounded brass plate was supported on
ceramic insulators and connected to the minus 100 kv.
The high-voltage electrode was connected to the plus
100 kv. For this test the 100 1-megohm resistors
were removed but the same lucite tube was used to

Figure 4. 100-megohm, 100 kv, helical resistor constructed

in 1961.

separate the electrode from the brass plate. Only

a visual test could be made; thus the setup was put
inside a completely enclosed dark box large enough
to house an observer also. With a total voltage of
200 kv between brass plate and electrode there was
no visible or audible evidence of corona. Since this
was twice the maximum voltage ever to be used with
the resistors in place, it was concluded that there
would be no corona current off the middle shield
electrode when two units are stacked in series. If
more than two units are to be stacked, further corona
tests should be made and possibly larger shield elec
trodes than the one shown in figure 4, which is 22 in.
in diameter, would be required.
4. Accuracy Limits on 1961 Model
Throughout the design and construction of the
1961 model helical resistor, primary consideration
was given to eliminating any possible corona or
leakage errors. Corona tests were made using volt
ages much above the expected maximum rating.
These precautions by themselves do not eliminate
the possibility of corona and leakage errors. The
crucial test for detecting such errors is to measure
current "in" and current "out" by the method al
ready described and used for the 1955 model.
The accuracy attained was considerably improved
over that in the previous tests. In part the higher
accuracy was obtained by using somewhat more
sensitive galvanometers and reducing the effects of
thermal emf's. However, the main improvement
was obtained by stabalizing the d-c voltage supply,
getting finer control, and using greater patience.
Also, by a modified experimental procedure, any
constant measurement errors such as might be in
troduced by the potentiometer, resistance standards,
or standard cells were cancelled. This was done by
taking current "in" and "out" measurements at low
voltage with the 100-megohm resistor shorted, before
and after each high-voltage check. Then if the po
tentiometer reading at balance is the same at high
voltage as at low voltage with the resistor shorted,
no leakage or corona effects are introduced by the
high voltage.
Measurements were made with three different val
ues of current (approximately 0.204, 0.509, and 1.019
ma) giving total voltages across the resistor of about
20.4, 50.9, and 101.9 kv. The odd values result ed
from the use of 5,000-, 2,000-, and 1,000-ohm XBStype resistance standards in conjunction with an un
saturated standard cell for setting current "in." A
5-dial null-type Wolff potentiometer was used to
measure the voltage across a similar standard resistor
connected between the resistor low-voltage lead and
ground, yielding the current "out." The precision
obtainable with stable current and fine control was
considered to be within 10 ppm. The check of cur
rent "in" and "out" with the 100-megohm resistor
shorted gave values agreeing within 10 ppm. With
the 100-inegohni resistor unshorted, requiring high
voltage to balance current "in" and "out", settings
on the Wolff potentiometer were the same within
10 ppm as with the 100-megohm resistor short ed.

386 -22

Voltages above 12 kv were also applied to the 24The accuracy of setting the Wolff potentiometer
with 20 and 50 kv across the resistor was within 10 unit special test resistor in order to estimate possible
ppm, i.e., the same as when the 100-megohm resistor heating errors at higher voltages. After 24 kv had
was shorted. With 100 kv across the resistor this been applied for 2 hrs the shields around the resistor
accuracy was estimated as within 20 ppmthe small units became noticeable hot to the touch and the
total change in resistance was 270 ppm. Based on
increase heing due to greater difficulty in holding cur
the 3 ppm temperature coefficient this would mean
rent nearly constant at the required value.
Current "in" and "out" data were obtained on a 90 C temperature rise for the resistance wire.
several different days over a period of 2 months. Since the final 100-megohm resistor has an estimated
Room temperature varied only from 24 to 25 C and temperature coefficient of 0.4 ppm, as compared to
relative humidity from 37 to 50 percent, ^fost of 3 ppm for the 24-unit special test resistor, its maxi
these data were taken wTith the brass ground plate mum change with 100 kv applied would be only 36
connected to ground and tied to the resistor low- ppm. However, this voltage docs cause considerable
voltage terminal through the standard resistor used heating and should not be held continuously. It was
in measuring current "out" (because that is the way arbitrarily decided that after applying voltages above
the resistor will normally be used). As a further 70 kv for 30 min a cooling period of 30 min should be
check on possible leakage current errors when two or allowed with no voltage applied.
more complete units are connected in series, data
were also taken writh the brass ground plate con
5. Calibration of 100-Megohm Resistor
nected to the resistor low-voltage lead and insulated
from ground. On all of these tests the current "out"
was found to be equal to the current "in" to within
As explained in the preceding section, it has been
experimentally demonstrated that corona and leak
the accuracy of measurement. This shows that co
rona and leakage errors for this helical resistor are age errors for this 100-megolun resistor remain within
less than 10 ppm for voltages up to 50 kv and are not certain limits up to 100 kv. Within these limits
then it is permissible to calibrate or measure its resist
greater than 20 ppm for 100 kv.
The only other source of error is that duo to the ance at low values of voltage and then use this
overall temperature coefficient of resistance of the measured resistance at much higher voltages to com
1 -megohm wire-wound resistors. The temperature pute resistor III drop.
coefficient from 20 to 40 C as measured for each 1One method of calibration was to measure each
megohm resistor used in this resistor was less than 2 1-megohm unit individually by placing it in the
"unknown" arm of a Wheatstone bridge. All other
ppm per degree C. Also resistors with approxi
mately equal positive and negative temperature co
arms of the bridge were made up of sealed XBS-tvpe
efficients were mounted as pairs. The overall tem
standard resistors except for a decade resistor box
perature coefficient is estimated to be less than 0.4 used for fine control by supplying from 0.00 to 10.00
ppm per degree C. Thus no correction for room ohms in 0.01-ohm steps. Values of all resistances
temperature under normal laboratory conditions used in the bridge, except the 1-megohm unit being
need be applied. Heating of the 1 -megohm units tested, were carefully determined by the Resistance
due to current at the higher voltage values, however, and Reactance Section of the National Bureau of
might cause a slight error because the resistors arc Standards. Temperature corrections were applied
totally enclosed and attain higher temperature than when necessary. These measurements were made
they would in a free air space. An estimate of the before the complete resistor was assembled. The
magnitude of such heating errors was obtained as total resistance was taken to be the sum of the
described in the following two paragraphs.
measured values for the 100 units comprising the
During the corona tests evidence of heating errors complete resistor.
A question now arises as to a possible change in
were detected (as already mentioned) when the resist
effective resistance of an individual unit due to its
ance of the 1/2-turn, 24-unit, test sample was mea
sured at various voltages. These 24 units were all being mounted inside its shield caps and their being
mounted in turn on the lucitc tube using nylon
purposely chosen to have large temperature coeffi
cients of the same sign so that temperature rise clamps and screws. Such a change could be caused
could be estimated from change in resistance. The by leakage current either via the lucite tube and
average coefficient for the 24 units was 3 ppm per nylon clamps or via the polyethelenc between shield
degree C. The total change in resistance, from a caps. Normally any leakage current via the lucite
measurement at 9 kv taken immediately after putting tube would have been detected during the current
voltage on, to a measurement at 12 kv after voltage "in" and "out" tests but leakage via the polyethelenc
had been on for about 70 min, was 70 ppm. This would not. To check for a possible error due to
indicates a temperature rise of about 23 C with 0.5 such leakage, the resistances of several matched pairs
kv applied across each 1 -megohm coil. Thus for the of the 1-megohm units were measured after they had
complete 100-megohm resistor it would take 50 kv been mounted inside their shields and on the lucite
to produce a 23 C temperature rise; but since the tube. In every case the resistance thus measured
agreed to within 10 ppm with the sum of the resist
complete resistor is made up of temperature compen
sated pairs, the 50 kv would cause a resistance change ance of the individual units as measured before
of less than 10 ppm.
assembly.
387-23

Another method of calibration was used with the

complete resistor assembled. This was to make
connections at everv fifth brass elbow (every 10th
resistor unit) so that the 100-megohm standard
could be divided into 10 sections of 10-megohm
units. These 10 sections were then connected in
parallel and measured as a 1 -megohm resistor. The
resistances of all 10 sections in series could then be
readily computed from the parallel resistance value,
provided the resistance of each section is equal to
10 megohms (within 0.1 percent). Values were
obtained in this manner on several different davs
and at room temperatures from 23 to 26 C. All
such values agreed with that obtained by summing
individual resistor measurements to within 15 ppm.
6. Summary
A special design of high-voltage d-c resistor made
up of a large number of individually shielded 1megohm wire-wound resistors connected in series
and arranged in helical form, was investigated and
found to be suitable for voltages up to 100 kv.
Measurements on the first helical resistor built in
1955 indicated negligible leakage and corona errors
to within the accuracy of the measurements made at
that time. However the heating errors were consid
erable (0.01 percent per degree C) in this resistor
because of the temperature coefficient of the wirewound 1 -megohm units.

A second helical resistor, of the same design, was

constructed in 1961 using 100 one-megohm, lowtemperature coefficient, wire-wound units. By using
matched pairs of these units having temperature
coefficients equal in magnitude and opposite in sign,
the final 100-megohm resistor was estimated to have
a temperature coefficient not greater than 0.4 ppm
per degree C. Corona and leakage errors for this
resistor as determined using improved measurement
techniques were found to be less than 10 ppm at
50 kv and less than 20 ppm at 100 kv.
This 100-megohm resistor has a special "hat" or
high-voltage electrode designed to give uniform
gradients from it to ground. This would allow other
similar resistors to be stacked on top of it and con
nected in series for higher voltages. Corona tests
on the electrode indicated the feasibility of such
"stacking". Exact limits of accuracy attainable at
voltages above 100 kv have not yet been determined.

The construction of the two high-voltage resistors

described above was carried out almost exclusively
by John L. Mills. Harold N. Cones and John L.
Mills performed most of the experimental measure
ments.

388-24

(Paper 66C1-83)

U. S. Department of Commerce
National Bureau of Standards

Research Paper RP1823

Volume 39, September 1947

Part of the Journal of Research of the National Bureau of Standards

Shunts and Inductors

for Surge-Current

Measurements

By John H. Park
The special requirements that must be fulfilled by a shunt intended to be used in surgecurrent measurements are explained. A tubular shunt with coaxial potential leads that
meets these requirements is described, and factors affecting its design are discussed. A
theoretical derivation of the "skin effect" in this type of shunt at high frequencies is given
in one of the appendices.
The advantages of using a mutual inductor for obtaining oscillograms of the rate of
change of current during a surge are outlined, and several types of mutual inductors developed
especially for this purpose are described. Theoretical derivations, given in the appendices,
indicate that the concentric-tube mutual inductors described in this paper can be used to
measure the high-frequency components of a current surge up to 70 megacycles with less
than 10 percent error.
Several shunts and mutual inductors of the designs described in this paper were con
structed for use in the high voltage laboratory at the National Bureau of Standards. Their
complete description and oscillograms showing results obtained with them are included.
I. Introduction
Surge-current generators are widely used in
high-voltage laboratories to simulate the heavy
current component of a lightning discharge. They
are regularly employed in testing protective de
vices that are intended to withstand lightning dis
charges and in determining the effects of heavycurrent discharges on various materials and ap
paratus. In order to correlate the test results
obtained in various laboratories, it is necessary
to know the magnitude and wave-form of the surge
current used in each test.
A generally accepted method of measuring
heavy-current surges is to insert a shunt in the
discharge circuit and to apply the voltage drop
across this shunt, through a suitable cable, to the
deflecting plates of a cathode-ray-oscillograph
capable of recording the variation of this voltage
with time. The over-all accuracy of this method
is dependent upon the three components used in
the measurement: the cathode-ray-oscillograph,
the cable, and the shunt.
The development of the cathode-ray-oscillograph
to its present state of efficacy has involved years
of work of a large number of experimenters and
theorists. At the present time there are several

types of oscillographs having recording speeds and

response accuracies suitable for surge measure
ments. In general they may be classified as (1)
the high-voltage cold-cathode type with the re
cording film inside the vacuum chamber, (2) the
hot-cathode type with the film inside the vacuum
chamber, and (3) the hot-cathode sealed-tube type
using a fluorescent screen photographed by means
of an external camera. Results reported in the
present paper were obtained using an oscillograph
of the first type; however, with slight modifica
tions in design constants the apparatus to be
described would be equally suitable for use with
other types of oscillographs.
In order to facilitate safety precautions and to
allow one cathode-ray-oscillograph (CRO) to be
used with any one of several surge generators, the
oscillograph is usually located at some distance
from the surge generator and its potential divider
and/or shunt and a cable (40 to 100 ft. long) is used
to connect them to the deflecting plates of the
oscillograph. The purpose of this cable is to
reproduce the voltage drop across the shunt, at
the deflecting plates of the oscillograph without
distortion, and it must not introduce other volt
ages by induction from nearby currents. The use
of coaxial-type cables reduces inductive effects to

389-191

a minimum, nnd when constructed with insulating

material of low dielectric loss, such as polyethylcjh>, coaxial cables arc capable of transmitting all
components of a surge up to at least 100 mega
cycles/sec. over the distances required without
significant attenuation. The terminations of the
cable, at the shunt and at the oscillograph, are
both important.
Various types of shunts have been used for
measuring heavy-current surges [1,2,3],' and long
before surge generators were in general use, the
design of alternating current shunts to insure low
residual inductance was given considerable at
tention [4,5]. Thus, the design of the shunts now
in use at this laboratory for surge-current measure
ments is based on experience with shunts used for
measuring large alternating currents as well as
on subsequent experience in the field of surgecurrent measurement. The present paper will
discuss the design of surge-current shunts, describe
those constructed, and indicate the results ob
tained with them.
This paper will also describe a new design of
mutual inductor for measuring rate of change of
current. Because it tends to magnify the high
frequency components in current waves it has
been useful in their measurement.

should preferably be such that its inductance and

skin effect can be computed for as great a range
in frequency as possible; otherwise the comparison
of various types of shunts would require the actual
construction of a large number with subsequent
experimental comparisons. Two general types
lend themselves to computations: (a) the con
centric tubular and (b) the flat strip. Illustra
tions of these, giving the methods of attaching
potential leads, arc shown in figure 1. Another

II. Factors To Be Considered in the Design

of a Shunt
1. Choice of the Best Type
In choosing the best type of design for a shunt
to be used in measuring surge currents the three
most important requirements are (1) the effective
impedance considered as a 4-tcrminal network
must be constant over as great a range in fre
quency as possible, (2) inductive effects of parts
of the current circuit, other than the shunt, upon
the potential-lead circuit of the shunt should be
a minimum, and (3) it must be possible to connect
the sheath of the cable from the shunt to the
CRO, to ground at or near the shunt without in
troducing induced voltages in the shunt potential
circuit (this is desirable to minimize the flow of
ground currents in the sheath of the cable which
might cause induced voltage at the CRO plates).
In order to determine compliance with require
ment number one, the configuration of the shunt
i figures in brockets indicate Iho literature references at the end of this
pnix-r.

Fic.unn 1. Three, types of shunts whose inductance and skin

effect can be computed.
a, Cnncentric-luhc shunt; b, flat-strip shunt; c, tint strip shunt with poten
tial leads arranged to Rive minimum inductance.
type of shunt used by Bellaschi |l] in surge meas
urements consists of several twisted "hairpin"
loops of resistance wire connected in parallel i t
the current terminals. This type of shunt bus
been found suitable for most surge-current mea iurements, but its inductance and skin effect can
not be accurately computed from theoretic il
formulas. An approximate theoretical computa
tion by Brownlee [2] indicates that the twisted
loop shunt would probably have a somcwh: t
larger time constant than a flat strip shunt.
Theoretical formulas for computing the induc
tance (assuming uniform distribution of curren
and skin effect (for a limited frequency range) of
the concentric tubular and flat strip shunts a e

390-192

readily available.2 3 As suggested by Silsbee 4 a

strip shunt with potential leads attached as shown
in figure 1 , c, could be designed so as to have zero
effective inductance by a suitable choice of values
of li and It. However, this arrangement of po
tential leads prevents the shunt from complying
with requirement (3) above, as will be explained
later, so it is not considered in the following com
parison of coaxial tubular vs. flat-strip shunts.
To compare the constancy of effective imped
ance as frequency changes, for the tubular and
flat-strip shunts, two typical designs of shunt were
assumed and their inductance and skin effect
computed from the formulas referred to above.
A concentric tubular shunt (fig. 1, a) in which the
resistance material was a CuNi alloy (p=22.9xl0-6
ohm cm) tube whose outside diameter was 0.25 in.
and wall thickness was 0.008 in. was chosen as one
example because such a tube was readily available
for use in constructing a shunt. For comparison
purposes a strip shunt was chosen whose thickness
was the same as the wall thickness of the tube,
whose width was the same as the circumference
of the tube, and whose separation between strips
was taken to be 0.01 in. (The shunt length does
not affect these computations provided it is several
times the largest transverse dimension.) The
time constants for the shunts as obtained using
the theoretical formulas assuming uniform dis
tribution of current were (1) for tubular shunt
L\R= 0.0365xl0"6; and (2) for strip shunt
L//? = 0.218xl0-6. The skin effects atj=l mega
cycle as obtained from formulas given by Silsbee
were (1) for tubular shunt R/Rde=0.959; and (2)
for the strip shunt R/Rdc=0.955. It must be
remembered that these results do not give an
accurate evaluation of the constancy of effective
impedance at frequencies over 1 megacycle,
because at such frequencies the skin effect causes
a nonuniform distribution of current density over
the thickness of the tube or strip, for which the
theoretical formulas used in the above computa
tion do not holdno published theoretical formu
las good at these high frequencies were known to
the author. However, the above results should
serve their purpose as a comparison between the
two types of shunts. The tubular shunt has a
much lower time constant than the strip shunt,
' ^ce p. 400 of reference [4] for inductance of tubular shunt.
See p. 81 of reference [5] for inductance of a flat-strip shunt; p. 80 for skin
effect of flat strip, and p. 91 for skin effect of a tubular shunt.
1 See p. 81 of reference [5|.

315-411 O - 68 - 26

which means that its reactance will be negligible

over a higher frequency range. The skin effect
at 1 megacycle appears to be the same for both
shunts, but the formula used in obtaining the
skin effect for the strip shunt neglects edge
effects that are known to considerably increase 6
the skin effect at higher frequencies. Thus it
would appear that the tubular shunt, as it has no
edge effects, would have less change in effective
resistance at the higher frequencies than the strip
shunt. A decrease in both inductance and skin
effect could be obtained by using a thinner strip
or a thinner walled tube, but this cannot be carried
to extremes because of the heat capacity require
ments of the shunt, which will be discussed later;
any improvement made by this means in the strip
shunt could probably also be made in the tubular
shunt, though possibly not to the same extent.
Thus, it would appear that either a coaxial tubular
shunt or a flat strip shunt would satisfy require
ment (1) (stated above), but that the tubular
design would be preferable.
As indicated above, the readily available theo
retical formulas are not capable of yielding the
effective 4-terminal impedance of a strip or tubular
shunt at frequencies of 1 megacycle or higher.
Due to the axial symmetry of the tubular shunt,
it is possible to develop a theoretical formula for
calculating the impedance of such a shunt, taking
into account the nonuniform distribution of cur
rent over the thickness of the tube (skin effect),
which will hold for very high frequencies. The
development of this theoretical formula (taken
from hitherto unpublished notes of F. B. Silsbee)
is given in appendix 1, and it permits the imped
ance of any tubular shunt to be computed for
frequencies up to 109 or higher. The develop
ment of a similar formula for strip shunts would
be much more difficult because of the lack of axial
symmetry.
The inductive effects of current-carrying parts
other than the shunt upon the potential lead cir
cuit of the shunt (requirement 2), are dependent
upon two factors: (a) the mutual inductance
of the current carrying parts upon any loop in
the potential circuit of the shunt, and (b) the
maximum rate of change of current with time
in the current-carrying part.
The maximum rate of change of current occurs
on the front of the current wave, and it is some Sec reference [5], p. 86 and 87.

391-193

times taken to be the peak vnluc of current divided

by the time to peak. This is actually the average
rate of change for the front of the wave, and any
detailed study of the front of a current surge re
quires a more accurate basis for obtaining the
maximum rate of change of current. Theoretical
ly, the maximum rate of change of current for a
capacitance discharging through an inductance
and resistance occurs at the instant the circuit is
closed and is equal to the voltage, E, to which the
condenser was charged divided by the inductance,
L, in the discharge circuit. The total inductance
of the discharge circuit of a surge-current generator
can be as low as 2 /th, and with a maximum voltage
of 100 kv the maximum rate of change of current
would be 5xl010 amp per second. Such a rate of
change of current in the main current circuit would
induce 50 v in the potential circuit of the shunt if
the mutual inductance, M, between the two cir
cuits wore 0.001 ph. As a "pick-up loop" in the
potential lead circuit of only 1 sq cm area, placed
very close to a current-currying part, may have a
mutual inductance of 0.002 ph (giving a peak in
duced voltage of 100 v), the extreme importance
of minimizing mutual inductance between the
potential circuit of a shunt and all current-carrying
parts becomes apparent. If the surge generator
discharge circuit is fairly long so that its total in
ductance is more than 2 /xh, the maximum rate of
change of current will not be correspondingly re
duced, because the stray capacitance across the
discharge circuit will in effect short out the added
inductance at the instant the surge is being
initiated.
Referring to figure 1, a, the potential circuit of
a tubular shunt may be considered as an extension
of the coaxial cable from the CRO, ending in a
direct short circuit from the central conductor to
the sheath. Due to the axial symmetry of this
arrangement the mutual inductance between the
potential circuit and any current-carrying parts is
the lowest obtainable. Even when the sheath of
the cable is connected to ground and ground
currents flow in the sheath of the cable and in the
outer current return tube of the shunt itself, these
ground currents will not induce voltage in the
potential circuit of the shunt provided the density
of such currents is symmetrical around the axis
of the shunt. End effects or concentration of
current along one side of the shunt axis can be
largely eliminated by extending the two current-

carrying tubes of the shunt several diameters

beyond the end of the central conductor that
serves as a potential lead.
To reduce mutual inductance in the potential
circuit of a strip shunt the potential leads can he
taken off the strip as a central conductor and
coaxial tube as shown in figure 1, b. There still
remains the small loop formed by the resistance
ribbon itself which is in the potential circuit of
the shunt and offers the possibility of inductive
pick-up from such currents as might flow in the
sheath of the cable and one of the current leads
up to the potential circuit. There is no way of
separating the potential circuit from end effects
arising from such currents in a strip shunt.
Although a strip shunt could be built with very
low inductive pick-up between its potential circuit
and current-carrying parts, the tubular shunt is
the only design offering the possibility of entirely
eliminating such inductive effects.
In order to understand the importance of being
able to ground the sheath of the cable going from
the Hhunt to the (!|{() at or near the shunt (roquircmcnt (.*{) above), it is necessary to consider
the entire surge-current generator circuit. A
schematic diagram of a surge-current generator is
given in figure 2. The heavy lines show the path
of the surge current from capacitor, C, through the
tripping gap, test specimen, measuring shunt and
the necessary connecting leads. No ground con
nections are necessary to complete the heavycurrent discharge circuit, but this circuit is usually
tied to the ground or the floor of the laboratory at
some point between the low-voltage terminal of
the capacitors and the test specimen in order to
complete the charging circuit and to definitely fix
the potentials of various parts of the discharge
circuit with respect to ground. The self-in
ductance of the conductors forming the discharge
circuit will depend upon their cross-section, length,
and arrangement. Assuming the conductors to
be of round cross section %. in. in diameter, their
inductance is 0.01 n h per cm length for a return
at a considerable distance. For the maximum rate
of change of current 5X1010 amp per second, this
means the maximum voltage difference between
two points 1 cm apart on this discharge circuit is
500 v. Thus, if the discharge circuit were grounded
at point g in figure 2, a shunt located only 10 cm
from point g would be 5,000 v above ground as the
discharge is initiated. This voltage will tend to

392-194

charging lead

ripping lead

i
1c
JURE 2. Wiring diagram of surge-current generator showng method of connecting shunts, their potential lead
irrangement, and several possible ground connections.
ike current flow in the sheath of the cable from
j shunt to the CRO, through the ground con
ation at the CRO and back to point g through
i laboratory grounding system. Due to the
ay capacitance (C, in fig. 2) from the lower ball
the tripping gap and parts of the discharge
cuit to ground, this voltage will be of a high
quency oscillatory type occurring just as the
tin current discharge is initiated and any curits caused by it will induce voltages of the same
iracter in any circuit coupled with them. In
ler to minimize such inductive effects at the
tO it is necessary that the cable from the shunt
the CRO have its sheath grounded at or near
1 shunt. In order to reduce the effect of stray
jacitance (C, in fig. 2) it may be desirable to
>und other points in the discharge circuit, such
g, in addition to the sheath of the cable. Thus,
\ best type of shunt would be one which would
mv any or all of these ground connections to be
de and still have minimum pick-up due to
luced voltages from ground currents.
In figure 2, all of the three types of shunts
istrated in figure 1 are represented as being
inected in the discharge circuit of a surge

generator in order to help explain the effect of

ground currents on each. The ground currents
most likely to cause inductive pick-up are those
due to the high frequency oscillations arising
from the stray capacitance, C to ground. If a
type C strip shunt were used, a consideration of
the various possible ground connections in addition
to the cable sheath ground at the CRO, leads to
the following conclusions: (1) a single additional
ground connection to the cable sheath at c would
not be suitable, because then ground currents
from C, would flow in one potential lead of the
shunt and induce high voltages in the loop formed
by the potential leads; (2) a single additional
ground connection at g would still permit cablesheath currents, which flow to ground at the
CRO end of the cable, to flow through one of the
shunt potential leads (actually this is better than
no ground at g); (3) two additional grounds, one
at g and one at c would introduce a combination
of the two above effects without eliminating
either; (4) a single additional ground at d would
reduce the inductive effects of ground currents
but would not eliminate them entirely.
Thus the use of a shunt of this type would pre
clude the experimentally desirable flexibility in
location of ground connections.
If a type B strip shunt were used, the same con
clusions given for the type C would hold except
all inductive effects would be much less because,
as may be seen from the diagram, the arrangement
of potential leads insures a lower mutual induc
tance between the potential circuit and currentcarrying parts.
If a tubular, type A, shunt were used the ground
current for any combination of ground connec
tions would be symmetrically distributed around
the axis of the coaxial potential circuit, thus,
theoretically at least, eliminating any pick-up due
to these ground currents.
The. above considerations indicate that the co
axial tubular design is superior to other designs
in surge-current measurement work because (1) it
should be more constant in impedance over a wide
range in frequency, (2) it can be constructed to
insure minimum inductive pick-up from currentcarrying parts of the surge generator, and (3) it
offers the greatest freedom in location of ground
connections at the surge generator. In view of
these advantages it was decided to investigate

393-195

theoretically ns well as experimentally the ad

vantages and limitations of concentric tubular
shunts for surge current measurements, realizing
that some of the derived relations would be appli
cable to strip-type shunts.
2. Determination of Cross-Sectional Area
and Length
The cross-sectional area and length of a shunt
arc related as follows:
V=IA
R=Pl/A
where V is the volume of the resistance material, I
its effective length, A the cross-sectional area, R
the resistance, and p the volume resistivity.
Solving for A and I

approximately this cross-sectional area as the

thinner the wall the slower the change in effective
impedance as the frequency is increased (see
appendix 1). The limit on wall thinness is set by
magnet ic stresses, to be discussed later.
If for practical reasons the values actually chosen
for I and A differ markedly from the approximate
values computed from cq 1 and 2 (this will neces
sarily be true if the preliminary value of / is too
long for convenience or if tubing having approxi
mately the estimated cross-sectional area is not
available) the allowable energy input should be
computed for the chosen values. This allowable
energy input together with the resistance of the
shunt and the maximum energy available from the
surge generator (% CE2) determine a minimum
effective resistance for the surge-generator dis
charge circuit. Except in cases where an oscilla
tory discharge with extremely low damping is
required, this minimum effective resistance will be
exceeded.
3. Forces on a Tubular Shunt

Tlic resistivity, p, is a constant of (he shunt mate

rial, and it is well to choose a material having as
high a resistivity as possible. The resistance of
the shunt is determined by the maximum current
to be measured and the maximum voltage that
can be recorded at the CRO. The volume of the
shunt resistance material is fixed by the largest
amount of energy that must be absorbed by it dur
ing a single discharge of the surge generator. A
reasonable temperature rise of say 100 C allows
a thermal input of 100 times the specific heat times
the mass of the shunt material. For most metals
the specific heat is about 0.1 and density is 9, so in
general the maximum allowable thermal input to
the shunt would be 90 times V, in calories for V in
cubic centimeters. The maximum energy fed into
the shunt during a single discharge is equal to the
energy initially stored in the surge generator capac
itors (1/2 C E2), times the resistance of the shunt
divided by the minimum effective resistance of the
surge-generator circuit. As this minimum effec
tive resistance is usually at least twice that of the
shunt, the maximum shunt input energy may be
taken as 1/4 C E2. When the volume of shunt
material computed from this energy input is in
serted in the above equations, approximate values
of / and A for the shunt are determined. It is best
to choose the thinnest wall tubing available having

When a heavy surge current is passed through

a tube made of resistance material, it will he. sub
jected to forces from two sources: (J) the. action
of the magnetic field produced by the current upon
the current itself and (2) the thermal expansion of
the tube arising from its sudden change in tempera
ture when the current is applied. The magnetic
force acts as a pressure from the outside tending
to collapse the tube. A formula for computing
this pressure is derived in appendix 2. It depends
upon the square of the current through the tube
and upon its cross-sectional dimensions. Except
for extremely high currents most thin-wall tubes
available for use in a shunt would probably with
stand the pressure arising from the magnetic field;
however, in the design of any shunt, and especially
in the case of a shunt made with extra-thin-wall
tubing, this pressure should be computed in order
to make sure that the tube will not collapse.
The sudden change in shunt temperature when
the discharge occurs introduces a stress in the
tube whose approximate value can be computed.
As the shunt tube itself is made of material having
much higher resistivity than other current-earn
ing parts of the shunt and surge generator circuit,
it will momentarily attain a higher temperature
than its supporting parts during each current
discharge. This sudden change in temperature

394-196

;nds to increase the length of the tube, but it

ccurs so fast the tube does not have time to
tpand, and the tube is momentarily put in comression. The maximum compressive stress may
e computed from the strain (total temperature
mnge T, times coefficient of linear expansion, 8)
id the stress-strain curve for the shunt material,
he tube will then behave somewhat like a spring
itially compressed and suddenly released; i. e.,
aves of stress will proceed along the tube, with
ie velocity of sound, being reflected at the ends
id traveling back as tensile stress. The nature
id magnitude of these stresses could be computed
r any given tube if the exact masses connected
the two ends of the tube were known ; however,
e maximum stress would not be expected to
eatly exceed the initial compressive stress that
ay easily be computed as noted above. Assumg the shunt material to' be perfectly elastic and
at it will not be stressed beyond its clastic limit,
e maximum stress will be
S=ET5,

(3)

lere E is the modulus of elasticity for the shunt

aterial. If a value for S is taken equal to the
iximum allowable stress (just below yield point)
r the material being used, this equation gives an
owable temperature rise for the shunt. For
>st materials this allowable temperature rise
about 100 C. This value could probably be
ceeded because of the high ductility of most
distance alloys; thus if a minimum resistance is
sired in the surge generator discharge circuit
taximum energy in the shunt) it would be adiable to determine the maximum allowable
nperature rise or energy input experimentally.
This could be done by placing a sample of tubing
the same material and cross-sectional dimenns as the shunt tube in the discharge circuit of
surge-current generator. The energy absorbed
the sample may be computed by multiplying
5 energy initially stored in the capacitors
E72) by the ratio of the resistance of the sample
the total effective resistance of the complete
ge generator discharge circuit including the
nple tube. The total effective resistance of the
ge generator can be deduced by using a shunt,
ose resistance tube has a cross-sectional area at
st five times that of the sample tube, in the
charge circuit and taking oscillograms of the
charge current. From measurements of fre

quency and logarithmic decrement on these oscil

lograms, the inductance and effective resistance of
the complete surge-generator discharge circuit can
be computed. Thus, by starting with the surge
generator capacitors charged to a fairly low vol
tage, E, and taking repeated discharges through
the sample tube at successively higher values of E
the energy input at which the tube is broken or
deformed may be determined. The safe allowable
temperature rise for a tube of this particular
material and size can then be stated.
The experimental procedure just described was
carried out for a K6-in. diameter CuNi alloy tube
whose wall thickness was 0.039 cm and length 33
cm. No evidence of tube failure except discolora
tion due to the high temperature was noted for
energy inputs below 160 joules/centimeter (instan
taneous temperature rise of 400 C) . However, after
the first discharge at this energy input a slight bend
in the tube was noted as though it were failing as a
column in compression. After each discharge of
successively higher energy input the bending was
more pronounced and after a discharge at 268 j/
cm (670 C temperature rise) in addition to the
bending, decided crinkling and crushing were
noted. The crushing was probably caused by
magnetic forces acting at an instant when the
tube was weakened because of its high tempera
ture. These values of energy input exceed what
might be considered a desirable design value.
The safe design value should be one that would
not stress the tube beyond its elastic limit. This
was determined in another experiment by measur
ing the length of the tube before and after each
discharge. After two discharges, the energy input
of each being equal to 60 j/cm (150 C tempera
ture rise), no permanent change in length was
noted. For energy inputs of 100 j/cm (260 C
temperature rise) or greater an increase in length
of about 0.2 mm (0.6 percent) was noted after
each discharge. Thus, it was concluded that the
maximum allowable temperature rise for a shunt
made of this material should be about 150 C.
This is only slightly above the maximum value
(133 C) computed from the stress due to thermal
expansion.
4. Inductance and Capacitance Between Current
Terminals
Because of the method of attaching potential
leads, the 4-terminal impedance of a tubular shunt

395-197

is constant over ;i wide range of frequency. 1 lowcvor, at very high rates of change of current (highfrequency components of Hie surge), the distrib
uted capacitance and inductance of the enrrent
circuit will cause variations in the root-meansquarc values of these high-frequency components
along the length of the shunt. That is, the in
stantaneous value of current along the length may
differ from the value at the current terminals.
This variation in the instantaneous value of
current along the length of the shunt is, of course,
entirely negligihle at lower frequencies, but in
order to determine the frequency at which this
effect becomes appreciable a solution for the cur
rent and voltage along the length of the tube,
assuming uniformly distributed inductance and
capacitance only, was obtained (in a manner sim
ilar to that used in appendices 3 and 4). The
current at the potential lead end of the shunt,
1 1, was found to be

'7
The magnitude of this voltage does not afreet Hie
accuracy of the shunt, but it should be evaluated
to indicate the insulation required between cur
rent terminals. For shunts up to 100 cm long
the quantity in the bracket of cq 7 reduces to
unity (within 5 percent) for frequencies up to 18
megacycles. Thus this voltage may bo taken as
equal to juHI or Hill/(It. The total inductance
of the current circuit for any tubular shunt, II,
may be computed from its dimensions by use of
eq 5, and the rate of change of current dl/dt
depends upon the particular surge generator heinp
used and its discharge circuit. For a typical
shunt, 100 cm long, this inductance is about
0.14X10-4 henries, and for the maximum rate of
change of current assumed above (5X 1010 amp/sec)
the voltage between current terminals is 7,000 v.
III. Description of Two Shunts

lc
/,=;
"cos uljLC

(4)

where
/e=current at current terminals; /^dis
tributed inductance per unit length of
shunt; C= distributed capacitance per
unit length of shunt; w=2ir/ (/ in cycles
per second); /=length of current circuit
in centimeters.
For concentric tubes whose wall thickness is small
compared to their diameter, the inductance and
capacitance per unit length are:
L=2 log, b/aX 10- henries

(5)

(7=^X 10-" farads

"log, b/a'

(6)

and

where a is the outer diameter of the inner tube,

and b is the inner diameter of the outer tube.
Multiplying eq 5 by cq 6 gives Z,C=1.110X10-81.
Thus the quantity -yjLC in eq 4 is a constant, and
the variation in current along the length of the
tubes depends only upon the length, I, of the tubes
and the frequency of the current being measured.
For shunts up to 100 cm long, the maximum varia
tion in current is less than 10 percent for frequen
cies up to 21.5 megacycles.
The voltage between the current terminals was
found to be

There arc two surge generators being used in

this laboratory at the present time. One is n
2,000 kv surgc-voltago generator capable of
giving 30,000 amp on short circuit. The other is
a 100 kv lOfif surge-current generator capable of
discharging a maximum of 200,000 amp. It was
Table 1. Constants of two tubular shunts

Four-terminal resistance
Over-all length
LenKt h of resistance tube.
Diameter of outside tube
Diameter of resistance tunc
Wall thickness of resistance tube.
Cross-sect innnl area of resistance tube
Resistance tube material
Resistivity of resistance tube (micro-ohm
em).
Ratio of a-c impedance to d-c resistance,
at:
/0.25 megacycle
/=o.M) megacycle
/=1.0 megacycles
/=2.0 nicpncycles
/=5.0 megacycles
Allowable temperature risecomputed
from stress due to thermal expansion.
Allowable enemy input
Current limit due to mnpnelic force
Total inductance of current circuit
Maximum voltage drop at current ter
minalsL <fi/df for </{/<U5xlO amp/soc

396-198

Shunt A

Shunt n

0.04 ohm
39 In
2ISJU in
5* in
H In
0.008 in. (0.0203
cm).
0.0302 sq cm....
Ocrman silver..

0.0048 ohm
38M in.
:V u in.
2 in.
1 in.
0.0246 in. (0.06
cm).
0.400 sq cm
80% Cu, J0%
Ni

23.33-

1.00..
1.00..
0.09..
0.90..
0.86..

OA
0.92
0.7G
0.48
0.15

133C
1320 joules .
87,000 amp.
0.202 m*----

133C
14,500 Joules
223,000 amp
0.106* *

10,000 V

397-198a

398-198b

thought that at least two shunts would be needed,

one for currents from 10,000 to 50,000 amp and
one for currents from 50,000 to 200,000 amp.
As the CRO was capable of giving satisfactory
records over the range from 200 to 2,000 v when
used with a suitable potential divider at the de
flecting plates, resistance values of 0.04 and 0.005
ohm were chosen for the two shunts that have been
constructed. Cross-sectional drawings of the
shunts are shown in figures 3 and 4. Their
physical and computed constants are given in
table 1. The resistance tubes were selected from
a small stock of readily available thin-wall tubes,
because at the time it was not feasible to wait for
special tubes to be drawn.
The ratios of a-c impedance to d-c resistance
given in table 1 were computed from the theoreti
cal formulae derived in appendix 1. At fre
quencies above 5 megacycles for the 0.04 ohm and
1 megacycle for the 0.0048 ohm shunt, this ratio
is considerably less than unity. This fact is of
no great concern in the measurement of the present
standard impulse-current test waves for which the
fundamental-frequency component is of the order
of 100 kc or less; but the fact should not be over
looked that the magnitude of higher-frequency
components, probably present in such test waves,
may be considerably reduced on a comparative
basis. Other types of shunts, as they have some
inductive coupling between their potential-lead
circuit and current^carrying parts of the discharge
circuit, may give an indication of these higher
frequency superposed currents. However, this
indication will depend, not upon the resistance of
the shunt, but upon the unknown value of in
ductive coupling, and thus will give an erroneous
idea of their magnitude. A method of accurately
determining the magnitude of high-frequency
components consists of using a mutual inductor,
having a known value of coupling with the current
circuit, in place of a shunt. Mutual inductors
have been constructed and are used at this labo
ratory. They will be discussed in section IV,
The allowable energy input for these shunts was
computed from the allowable temperature rise (ob
tained from the maximum stress due to thermal
expansion) and the mass of each shunt (actual
values are given in table 1). These values are use
ful in computing the minimum resistance that can
be used in the discharge circuit of a surge generator
without overheating these shunts. As an example:

The surge voltage generator in this laboratory

requires the dissipation of 33,300 j for each dis
charge when charged to full-rated voltage. Thus,
when shunt A is used with this generator, the total
resistance of the discharge circuit must be at least
33,300 divided by 1,320 times the shunt resistance
(0.04 ohm) or 1.01 ohms. A similar computation
for shunt B gives 0.107 ohm. As the minimum dis
charge resistance of this surge voltage generator is
6 ohms, cither shunt may be used with any dis
charge circuit. The surge-current generator in this
laboratory requires the dissipation of 50,000 joules
when charged to full-rated voltage, thus the total
resistance of the discharge circuit must be at least
1.52 ohms if shunt A is to be used or 0.0165 ohm if
shunt B is to be used. As the minimum discharge
resistance of this surge-current generator is 0.03
ohm, shunt B may be used with any discharge
circuit, but shunt A should only be used when a
resistance of at least 1 .5 ohms is added (a resistance
of about this value is needed to give a critically
damped discharge).
The current limitations on these shunts due to
magnetic forces were estimated by the formula
given in appendix 2. These limitations are above
the values of current expected to be used with
these shunts. The inductances of the current cir
cuits of these shunts were computed by using the
formula given in eq. 5. The voltage drops com
puted from these inductances and the maximum
expected rate of change of current indicate the
desirability of having adequate insulation between
the current terminals to prevent flashover.
IV. Mutual Inductors for Measuring Rates
of Chanae of Current
1. General Considerations
For many experiments involving heavy-current
surges, it is desirable to know the wave form of
the rate of change of current, for example: (1) in
studying the voltage induced in a circuit coupled
with the heavy current discharge circuit and (2)
in studying the voltage difference arising from
inductance between two points on a conductor or a
system of conductors carrying the heavy-surge
current. The rate of change of current can be
computed from an oscillogram of the current
obtained by using a shunt, but this involves the
inaccuracy of measuring the slope of the current
wave at a number of points and plotting a curve

399-199

from those measured vnlucs of slope. In addition,

the high-frequency components of the current
wave form ns they appear in the oscillogram ob
tained with a shunt may be. highly distorted or
attenuated and give an erroneous value of rate of
change of current, especially at the point of
maximum rate of change. A more accurate record
of rate of change of current can be obtained with
much less effort by inserting the primary of a
mutual inductor in the heavy-current discharge
circuit in place of the shunt and connecting the
secondary of tin's inductor to the cable to the (TKO.
The record obtained on the CUO is then equal to
M di/dt. If M (the. mutual inductance of the
inductor) is known, an actual oscillogram of the
rate of change of current is obtained.
A mutual inductor, to be considered satisfactory
for this purpose, must fulfill the following re
quirements: (1) Its mutual inductance must be
of suitable value to give the proper voltago at the
CItO (from 500 to 2,000 in this case). For the
expected rates of change of current (from 1 to 5
times 10 10 amp per second) this amounts to about
0.05 fih. This value of mutual inductance must be
definite and should bo computable from the
physical dimensions of the inductor.
(2) The primary circuit of the inductor must be
capable of carrying the heavy current discharge
and should add a minimum impedance to the
discharge circuit.
(3) The secondary circuit should have minimum
coupling to all current carrying parts except the
primary of the inductor.
(4) The method of connecting the secondary
circuit to the cable going to the CRO should be
such that the sheath of this cable can be connected
to ground near the surge current generator without
introducing coupling between ground currents
in the sheath and the secondary circuit.
(5) The secondary circuit should have a mini
mum self-inductance. This may be explained
as follows: The voltage to be measured at the
CRO (M dl/di) is impressed upon a circuit con
taining the self-inductance, L, of the secondary
circuit and the surge impedance, R, of the cable
to the CRO. It is the voltage across the surge
impedance, R, that is recorded at the CRO and its
value is
F ^Mdljdt
Smo=-L*
1+J'R

(8)

in order to have Reno equal to Af d/fdt for t he lii^li

frequency components of /, it is apparent that tlic
self-inductance, L, of the mutual inductor second
ary circuit should bo kept as low as possiMe.
As L cannot be reduced to zero, the highest
frequency components of / arc subject to a phase
displacement and attenuation, the magnitudes of
which can be computed from the above equation
for any assumed value of w=(2x/).
A consideration of the various convent ion typos
of mutual inductors leads to (lie conclusion Unit
none of them will fulfill the above requirements
satisfactorily. The most promising type appeared
to be. one consisting of a straight single conductor
primary insulated from, and running through, the
central axis of a toroidal coil secondary of sufficient
turns to provide adequate voltage at the CRO.
It was thought that by introducing shields and
connecting them to the primary circuit at an
appropriate point with respect to the ground
connection, an inductor of this type might fulfill
most of the requirements listed above. A pre
liminary design of an inductor of this type was
carried out, but the self-inductance of its second
ary circuit had to be at least 2 ph in order to give
the desired value of mutual inductance. When
this value of inductance and 50 ohms for the CRO
cable surge impedance, R, arc inserted in eq. S,
it is found that the voltage at the CRO is only
0.9 of M -jj for a frequency of 1 .8 megacycles, for
higher frequencies the attenuation would be much
greater. The use of a mutual inductor of this
type was, therefore, abandoned.
2. Mutual Inductors of Special Design
The first attempt to design a special mutual
inductor to meet all of the requirements listed
above resulted in the inductor shown in figure 5.
The primary circuit is the straight length of No.
10 copper wire (with the return circuit at least
2 ft away, its effect can be neglected). The
secondary circuit consists of the "D" formed by
part of the No. 10 wire and the No. 16 wire. This
arrangement permits the magnetic flux linking the
secondary "D" to be computed as that due to an
infinitely long, straight conductor, neglecting the
effect of the return lead. The self-inductance of
the secondary circuit is that due. to tho single "D''
loop and can also bo computod from its dimensions.

400-200

I 1 scale
1 1 ' ' j
0
S
10 cm

I ' > ' I ' 1 ' I

0 12 3 4 incht
Figure 5.D-type mutual inductor used to obtain records of
rate of change of surge currents.
M =0.0986
An inductor of this design, with dimensions cor
responding to the scale in figure 5 was constructed.
Its mutual inductance was computed to be 0.0986
uh and the self-inductance of its secondary circuit
0.25 /ih. This value of self-inductance is con
siderably less than that of the toroidal secondary
circuit mentioned earlier, and as a consequence
frequencies up to 15 megacycles will be recorded
with only 10 percent reduction in magnitude. The
secondary circuit is connected to the CRO cable
by coaxial leads, thus no voltage will be induced
in it from ground currents flowing in the cable
sheath. A mutual inductor of this typo appears
to fulfill the requirements of surge current testing
fairly well. The one constructed in this labora
tory has been of considerable value in obtaining
records of the rate of change of surge currents.
Its outstanding advantage lies in simplicity of
construction.
The "D"-type mutual inductor has some draw
backs: (1) it is not completely unaffected by
magnetic fields produced by other current-carrying
parts of the surge generator circuit and (2) the
self-inductance of its secondary circuit is about
three times its mutual inductance (as a minimum
it should be possible to make an inductor whose
self-inductance is as low as its mutual inductance).
Another mutual inductor has been designed in
which the secondary circuit consists of a tube
arranged coaxially with the primary conductor in
order to overcome the small defects inherent in
the "D"-type inductor.
A type I coaxial tube mutual inductor is shown
in figure 6. Tube A constitutes the primary cir

cuit of this inductor (the surge-current return lead

is again assumed to be located at least 24 in. away
so its effect can be neglected). The secondary
circuit of this inductor consists of the following
parts: (1) tube, A, from a to b, (2) a radial ring
soldered on its inside to and around tube A at b
and soldered on its outside to and around the
inside of tube, B, at c, (3) tube B from c to d, and
(4) the four-wire radial "spider" connection from
d on tube B, going through but insulated from
tube, A, and soldered to the central conductor of
the potential leads at/. The secondary terminals
are thus at a and / and are brought out through
the concentric leads to the fitting for the CRO
cable. How well this design of mutual inductor
fulfills each of the requirements listed above may
best be seen by considering each requirement, as
follows: (1) The mutual inductance can be com
puted as that due to the magnetic flux produced
by current in a long straight conductor (tube A)
integrated over the region a b c d a between the
tubes. The mutual inductance thus derived in
terms of the tube dimensions indicated on figure
6 is
M=2l log, ^ 10~' henry (for I in cm).
There is a small correction to this value of induc
tance due to the thickness of tubes A and B,
which can be computed by the method explained
on page 397 of reference [4]. For ri=0.25 in.
and ra=l in., the length, I, of tube B should
be about 7 in. to give a mutual inductance of
0.05 nh. (2) Because of the method of bringing
out the secondary circuit leads used in this design,
tube A must be at least 0.5 in. in diameter, and it
should be made of a good conducting material such
as copper. The wall thickness of this tube is
relatively unimportant, and any ordinary thin
wall copper tube can be used that is adequate to
carry the heavy current discharge without damage.
(3) The secondary circuit except for the ring and
spider are coaxial conductors and thus have mini
mum coupling to all currents not flowing in the
inductor. Likewise the radial ring and spider
conductors are symmetrical about the axis which
results in minimum coupling. (4) Also, because of
the coaxial arrangement of the secondary circuit
and leads to the CRO cable, any ground currents
flowing in the cable sheath will have minimum

401-201

coupling with (he secondary circuit nnd loads.

(5) The self-inductance of the secondary circuit
can be computed by the method explained on
page 397 of reference [4], nnd it is found to ho
equal to the mutual inductance except for a small
term arising from the finite thickness of tube, Ji.
The self-indiK'tanee of (he spider alone is less than
0.007) /ill. Thus this design gives the minimum
self-induclanee that can be obtained in a straight
forward manner.
An approximate value of the voltage impressed
on the Cl{() cable can bo obtained from o<|. N, but
this equation neglects the capacitance between
tube, A. and tube, Ji, and will not' hold for the
higher frequencies. A more accurate method for
computing the value of this voltage is to consider
the capacitance and inductance of the secondary
circuit to bo uniformly distributed along its
length and solve for the voltage at the CKO end
of this circuit by the methods normally used
for circuits of uniformly distributed constants.
This is done in appendix 3, and as shown there for
a typical mutual inductor designed for surgecurrent measurements, the voltage impressed on
the CRO cable will be equal to M dl/dt (within
10 percent) for all frequencies up to 65 mcgacyclos/scc.
The type I tubular mutual inductor appears to
satisfy the requirements for surge-current measure
ment quite well. One was constructed in this
laboratory having the approximate dimensions
indicated in figure G. This type of inductor is
entirely satisfactory if the surge-generator dis
charge circuit is fairly long and can be arranged so
that the inductor forms the final part of the return
lead to ground. In case the minimum length of
discharge circuit is required, the length of this
inductor may add an appreciable inductance to
the discharge circuit and it may also not bo
feasible to keep the return lead 24 in. away from
the axis of' the inductor. For such cases, a
slightly different design of tubular inductor (type
II) as indicated in figure 7 is suggested. It is the
same as the type I inductor except that a return
circuit is provided for the current by a larger tube,
coaxial with the smaller tubes. This design brings
the two current terminals close together and
eliminates the problem of the location of the
inductor with respect to other ports of the surgegenerator circuit. It allows the inductor to be

used m a discharge circuit of minimum lon^h

with the addition of the least possible induoiumv
to the discharge circuit. The outer current n>turn tube of the type II inductor also acts as un
electrostatic shield for its secondary circuit.
Because of the distributed capacitance between
the outer-current return tube and the coaxial
secondary circuit tube, currents and volljigos in
(he secondary circuit will not be I lie same for the
type II inductor as for the type I. Thus a com
putation of the voltage impressed on the ('KO
cable for a type 1 1 inductor must take into account,
the distributed capacitance mid inductance of I lit'
circuit formed by the outer tula' of the sifondiiry
circuit and the return current lube in addition la
that of the secondary circuit itself. This is done
in appendix 4, and while the results appear cj nite
different from those obtained for the type I in
ductor in appendix 3, they indicate about the same
upper limit in frequency (70 megacycles/sec).
Each of the three types of mutual inductors
described above should prove useful for sur^rcurrent measurements. The "D" mutual induc
tor is very simple to construct and will measure
superposed frequencies up to 15 megacycles/sec
with less than 10 percent error. The tubular
inductors are somewhat more complicated to con
struct but extend the frequency range to 70
megacycles/sec. The type I tubular inductor is
suitable when the surge-generator discharge path
is fairly long and the type II when a minimum
length of discharge path is required. Inductors
of both the "D"-typc and the type I tubular
have been constructed and proved very useful in
this laboratory.
V. Experimental Results
The shunts and mutual inductors described
above have been used in this laboratory for meas
uring current surges both from a surgc-voltago
generator and a surge-current generator and for
various arrangements of discharge circuits. The
oscillograms in figure 8 arc typical of the results
obtained. They arc records of the discharge cur
rent and rate of change of current for a surgecurrent generator consisting of 40 l-/if, 50-kv,
capacitors connected in series-parallel so as to give
a total capacitance of 10 /if and a voltage rating
of 100 kv. The discharge circuit consisted of a

402-2 02

c
d
Figure 8.CRO records of a critically damped discharge from a surge-current generator.
a. Current wave-form obtained by use of tubular shunt . .1 . b, early part of current wave-form on a faster CRO sweep, crate of change of current obtained
by use of the mutual inductor shown in figure 6; d, early part of rate of change on a faster CKO sweep.
three-ball triggering gap; a 1-ohm resistor made
up of 24 globar type B ceramic resistor units, as
sembled in the form of a cage (to minimize in
ductance) having sets of two 12-in. units in series,
connected in parallel so that this assembly was
24 in. long; about 5 ft of copper busbar; and the
shunt or mutual inductor. The total inductance
of this generator and its discharge circuit is 2.5 ph
and the 1-ohm resistor gives very close to critical
damping. For the oscillograms shown in figure
8 the generator was charged to 75 kv and gave a
maximum discharge current of 55,000 amp, the
maximum rate of change of current being 2.5xl010
amp/sec.
The oscillogram in figure 8, a, shows the varia
tion of current with time for the entire surge.
Figure 8, b, shows the initial rise of the current on
an expanded time scale (faster CRO sweep).
Figure 8, c, shows the variation in the rate of
change of current for the entire surge, and figure
8, d, shows the initial rate of change of current

using a faster sweep The current records are

quite smooth and just as anticipated from theory,
as are the rate of change of current records
except for the superposed high-frequency oscilla
tions occurring just as the surge is initiated.
These superposed oscillations have a frequency of
about 40 megacycles and are probably caused by
stray capacitance across parts of the discharge cir
cuit, such as from one ball of the tripping gap to
ground. If this stray capacitance is across a sec
tion of the discharge circuit having an inductance
of 2 /ih, it would require only about 10 /i/if to pro
duce the 40 megacycle oscillations. It is practically
impossible to eliminate such stray capacitances.
In the discharge circuit used in obtaining the oscillo
grams of figure 8, stray capacitances were reduced
to as low a value as possible. For discharge cir
cuits involving fairly large apparatus as a test
specimen, stray capacitances will be larger, result
ing in a superposed oscillation of lower frequency.
The slight upward curvature on the zero line

Surge Current Shunts and Inductors

403-203

of oscillograms in figures 8, b and 8, d at the start

of the sweep is introduced by tbc oscillograph and
should not be considered as part of the surge
record. The source of this "hook" in the CRO
zero line is a small horizontal deflection given to
the beam by the Norindcr relay plates as voltage
is initially applied to them. This deflection
slightly alters the CRO sweep rate at its beginning
but should, by itself, introduce no vertical de
flection. However, since tho focusing coil is
located below the Norinder relay plates, this
deflection puts the beam slightly off the axis of the
focusing coil which then produces the small
vertical deflection.
The oscillograms in figure 8 illustrate the ad
vantages of using a mutual inductor in addition
to a shunt when measurements of a current surge
arc being made. The superposed oscillations can
not be detected on the current record but can
readily be seen and measured on the rate-ofchangc-of-currcnt record. Their actual magnitudes
in amperes can be computed from this record, and
from measurements on the oscillogram of figure
8, d they were found to have a maximum ampli
tude of about 25 amp. Oscillations of this
magnitude would be almost impossible to detect
on a current record even though the attenuation
of the shunt for these higher frequencies did not
mask them out. The rate of change of current
records are also of great value in studying the
effects of induced voltages arising from the
current surge. They further provide an accurate
measurement for the time to maximum current
(zero rate of change).
VI. References
111 P. L. Bellaschi, Heavy surge currentsgeneration and
measurement, Electrical Eng. 53, 86 (Jan. 1034).

VII. Appendices
1. Effective Impedance of a Tubular Shunt as a
Function of Frequency 6
Let the radii of the inner and outer tubes be a, b, a', and
6' as indicated in figure 0. Assuming the tubes to be
coaxial and perfectly symmetrical about their axis the
current density in cither tube at a given radius, r, from
the axis will be a constant for any given instant, so let
tr = current density in inner tube at radius r,+ if into
paper
ir = current density in outer tube at radius r,+ if into
paper

Figure 9.Cross-sectional view of a tubular shunt.

Then
/ = total instaneous current in inner tube
r*=
b
<= y2*i,T&T ^
/' = total instantaneous current in outer tube
Je
A'
2rir'rdr
'
*
^
Also, let:

12] T. Brownlee, discussion of Bellaschi's paper, Heavy

surge currentsgeneration and measurement, Elec
trical Eng. 53, 481 (Mar. 1934).

e=emf drop per unit length of inner tube

e'=emf drop per unit length of outer tube
//,= magnetic field at r taken + if clockwise
p = resistivity of inner tube material
p' = resistivity of outer tube material
r= total magnetic flux outside of radius r
(all permeabilities assumed = 1).

13] N. Rohats, A surge-current generator, General Electric

Rev. 37, 296 (June 1934).
14] F. B. Silsbee, A study of the inductance of four-terminal
resistance standards, Bui BS 13, 375 (1916) S281.
15] F. B. Silsbee, Notes on the design of four-terminal
resistance standards for alternating current, BS J.
Research 4, 73 (1930). RP133.

Note.The cgs magnetic system of units is used

throughout this derivation and all dimensions are io
centimeters.
Thti derivation is taken from hitherto unpublished notes ot F. D. Sflstw.

404-204

As all currents flow along lines parallel to the tube axis,

e and ' are independent of r, so

If t, =/(*'-<

(9)
and eq 19 and 20 become
(.'-nV-l-'

(10)

and
de_
dt'r ^di
, d d<tv=n
dr_p dr
br u

For the inner tube

dx

(13)

where

For r>6', +,=0.

(21)

(22)

(12)

dr _p or + d< dr

/=

+7

(11)

4v and K'1jut
4x the solution of
where K*=ju>
p
P
eq 21 is
tr=Zo(Xr) = C1J0(Xr) + C2AT0(Kr)
(23)

2rirrdr and I'.=j*. 2ri,rdr

Differentiating eq 13 with respect to r

where J0 and Af0 are Bessel's functions of order 0 and of

the first and second kind. G and (\ are constants of
integration. Similarly the solution to eq 22 is
i'r=Zo'(K'r) = C'lJ0(K'r) + C',K(K'r)

Similarly for the outer tube,

(24)

The total current is

(15)
/ = 2JVrrfr=2xJ^0(Kr)dr

and

(25)

and from the general relation

$'--^--ii>j>-H-+ir'**
(16)
since /= I ( 2ri'rrdr

JxZo(ax)dx=^xZl(ax)

(26)

7=^[62,(K6)-aZ,(Xo)]

(27)

it follows that
Inserting eq 14 in eq 11 and eq 16 in eq 12,
di, 4 d fr .
(17)
Similarly
(18)
I'
Differentiating again with respect to r
r dr+dr

p r dt

[b'Z[(K'b') -a'Z[( K'a') ]

(28)

Likewise
U

(19)
(20)

405-205

2r [iZ,(Xi)-aZ,(Ko)]
,
/:=^[xZ;(K'x)-a'Z;(K'o')]

(29)
(30)

To obtain values for the *,'s substitute cq 29 and 30 in eq 13 and 15.

tegrations
*r=A-2 L-Z0(foOjr

For the inner tube, after performing thc;t

1 log -+2/ log j +^r,^-Z0(A'x)J<>

-g,

log

;oi

For the outer tube

*r=2/ log -+ ^r,j -Z0(/C'*) J r

-^t

1 log -

The voltage drop in the outer tube is

After substituting and simplifying

'/

log &' + WZKA'a') log 6' + Z;(/C'6')-log r [^?"+ AVZ^KV)] |

Since ' is not a function of r, tho coefficient of log r in cq 3.1 must vanish, this gives
Z;(KV) = ~> 1

(Ml

This is ono relation fixing the coefficients in Z</. The other relation is in cq 28. Substituting cq 34 in eq 28 gives
Z;(K'6')=0.

(35)

e=pZ(Kb) + KapZx{Ka) log b-KapZ^Ka) log r+2/j log ~+p'lZi(K'b')-Z^K'a')]+p'K'a'Z'l(LK'a') log ^

(3!

Voltage drop in inner tube is, similarly

Coefficient of terms involving r must be zero, therefore

ZdKa)=0

(37)

Putting eq 37 in eq 27
Z,(X6)-^
Inserting eq 37 and 34 in eq 36 gives
e=pZ<,(Kb)-2Ijo, log +p'[Z'(A"&') -Z0'(A"o')]
The effective four-terminal impedance for a shunt with the potential lead inside the inner tube is
d
e-dt*'
e
t

406-206

(39)

From eq 31, 34 and 37

^=pZ(K&) -pZo(lfo) +p'[Zi(K'6') -Zi(KV)]-jy/ log 4
Subtracting eq 40 from 39

d *
i tv~\ _pZ0(Ka)
di ,=p
'
7

From eq 24 and 34
z;(KV)=2^/=c;j;(/ra')+c;Ar;(KV)
From eq 24 and 35
ZJ(K'6') = 0= C'JWV) + C^,(K'6')
From eq 23 and 37
Z,(Xo) =0= CJdKa) + CjN^Ka)
From eq 23 and 38
Z1(K6)=^=C1/,(K6)+C2iV1(/C61
Solving eq 42, 43, 44, and 45 for the coefficients
K'l Nl(K'b')
Ll ~2xo'
D'
r,_K'I JAK'V)
Cj -2xa'
D' '
where D' =JAK'a,)Nl(K'b')-Ji(K'b')Nl{K'a'),
n

KI Nt(Ka)

r_KI MKa)
Ci~2ib~.D~'
where D=J,(Ka)NAKb) -J^Kb)N,(ka).
From eq 23
Z0(Ka) = dJo(Ka) + CN^Ka)
Using this and putting in values of Ci and Cj from eq 46 and 47 eq 41 becomes
_pK rN0(Ka)J, (Kg) - N,(Kg) Jo(Kg) 1
*~2rb L^i(Ka)iVi(K&) -./,(/&) tf,(Ka) J
Using the general relation
Nl>(Kg)Jl(Ka)-Ni(Kg)J,(Kg) = :~-a
eq 48 becomes
P
1
r1abJ,(Kg)Nx(.Kb)-Ji(Kb)Nl(Kg)

315-411 O - 68 - 27

407-207

Using the following general relations

/,(Xa)"VS [/<*> in(lCo-j) + Q1(Ko) cos(ifo-^)]
Nt(Kb)=J^s [-P,(Kb) cos(jtt-f) + Q,(K6) sin(if6-^)] .

letting 6 a=d, and putting in the following approximate values of Pi and Q,

P,(/fa) = 1+72&-,

the denominator of eq 49 becomes, for Ki=jmi

,*JS^|[(l+,n,XwnjM Wssb] *5f -*+

K'i^X^-('ii&X3Kh3lM
Most of the terms inside the brackets in cq 50 arc negligible if the frequency is much above 00 cycles per second. L
/= 1,000 c/s, />^40,000 (cgs units), (6-a)^0.06 cm, and b = a=l (approximately), this equation simplifies to
n ^= d-i)

(51)

md
Letting

and putting eq 51 back in eq 49

z-

pS r(sin 4 cosh 4+cos 4 sinh 4) +j (cos 4 sinh 4 sin j cosh 4)~J

'ird^Jab L
(sin * cosh 6)'+ (cos 4 sinh 4)'
J

where
p=rcsistivity of shunt material in cgs units
a = inner radius of shunt tube in cm
6= outer radius of spread tube in cm
d= thickness of shunt-tube wall in cm
u>=2r times the frequency of the current through the shunt.
z= vector impedance of shunt for values of above quantities used.
The d-e resistance of the tube in absohms, for d<6 is

* 2xdva6
and eq 52 becomes
z=jj . f(sin 4 cosh 4+cos 4 sinh 4) +j (cos 4 sinh 4 sin 4 cosh 4)"|
L
(sin j cosh 4)'+ (cos 4 sinh 4)1
J
Equation 54 is very nearly correct for any thin-wall tube when the frequency is above 1,000 cycles per second.

408-208

Freq.

in

KC

0.0048

shunt

500
1000 .
, 2500
. 50OO . . iO 000
Freq.
in
KC for
o.o4n
shunt
Ficubk 10.Curves showing ratios of a-c resistance R, reactance uL, and impedance Z, to d-c resistance Rjc, plotted as ordinates against the parameter, S, as abscissa.
Abscissa scales in frequency are also shown for two tubular shunts. The solid-line curves were obtained from the formula derived in appendix 1. The
dashed-line curves from formulas known to be applicable only at lower frequencies.
Yo-

For lower frequencies, the skin effect formula as given

on page 91 of Research Paper No. 133 (see reference 5) is

The current density inside the metal of the tube is

._
I
l_x(6'-a>)'

=^(1-w>-(1-iJs4)

<55>
The total current in the shell of thickness dr is

The inductance at lower frequencies, where the current is

assumed to be uniformly distributed as given on p. 400 of
Scientific Paper S281 (see reference 4) is, in abhenries per
centimeter,
l
d

ir = i2-rrdr = i?^-i
o2 o1 dr.
The total radial force (directed inward) on this shell of
current, due to the magnetic field H, is
F,=irHr

For d<<6 this gives

Rdc

"3

(56)

Actual values of R/Rdc, wL/Rj*, and 'z'JRdc as functions

of I are shown by the curves in figure 10. The solid lines
represent values obtained from eq 54 and the broken lines
values from eq 55 and 56. For i< 1, the broken and solid
curves coincide, but for S> 1 the values obtained from eq 55
and 56 exhibit rapidly increasing errors as would be
expected from the assumptions made in the derivation of
these equations.
Frequency scales are also given on these curves for the
two shunts in use at this laboratory.
2. Magnetic Force Due to Current in a Tube
Referring to figure 9, the magnetic field (Hr) at any
point r distance from the axis of the tube, where a<r>6,
for a total current, /, in abamperes uniformly distributed
over the cross section of the tube, is
r>-aa 2/
W-o* r

in dynes per centimeter length of tube. The total force

on the entire tube per centimeter of length is
ft
Cb r2 a1
Performing the integration

This force expressed as a pressure per square centimeter is

which simplifies to
p_2/T 1 + a1
3x6 U+o^^+o^J
which is the pressure in dynes per square centimeter.

409-209

The collapsing pressure for thin-wall tubing as obtained

from A. E. II. Dove's formula 7 is
2E ( t \
P.' "l-m* \Dj '
K is tlic modulus of elasticity and m is Poisson's ratio for
the tube material; I is the wall thickness of the tube and
I) is its diameter. If E is expressed in dynes per square
centimeter, the collapsing pressure l'c will also be in dynes
per square centimeter. Equating this to the magnetically
induced pressure due to current, /, as obtained above
and solving for /, we obtain the current in abamperes at
which the tube will collapse
7 = 1.085

e.= K, sinh [xVC^ + A',]

(59)

i. = / +jKt-yjjf; cosh [uxJCrfT, + K>}

(00 !

To obtain values for the arbitrary constants Kx and A'2l

use relations at x=0 and x=l. At x=0, e, =0 and from
cq 59
0=A', sinh [0+A'2).
As Ki cannot be zero, A'a=0.
At x=i, e,=i|/i. and from cq 59

c
1
i ;
I i

The solution of these differential equations gives equations

for i, and e, containing two arbitrary constants:

(b-o)'(b+a)'.
b*{b + 2a)

3. Effective Inductance of a Type I ConcentricTube Mutual Inductor as a Function of

Frequency
In a ty|>c I concentric-tube mutual inductor, shown
schematically in figure 11, the current, I, whoso rate of

.1 b

C| = capacitance per centimeter length between the inner

and outer tubes,
e, = tho potential difference between the inner and outer
tubes at the distance, x, from the end of the outer
tube,
t', = thc current in the outer tube at the distance x from
the end of the outer tube.
the differential equations are
diM=juCiCadx
and
de,=juMi(I i,)dx.
(58)

o pp'flenti Sfg
."iimcdonei
o> Cob
^
1
coble

FlOURK 11.Schematic drawing of a type I concentric tube

mutual inductor.
change is to be measured, passes through the inner tube
and the return circuit for this current is assumed to be
far enough away (about 24 in.) so that its effect can be
neglected. The secondary circuit of this mutual inductor
consists of the length of the inner tube from a to b. the
radial ring from b to c, the outer tube from c to d, and the
four-wire "spider" from d to a. It is considered to be
completed by the resistance, It, which represents the surge
impedance of the cable to the CRO. Assuming that (1) all
currents (lowing in these tubes are uniformly distributed
around the central axis, (2) end effects may be neglected,
and (3) shunting capacitance from the outer tube to out
side conductors may be neglected, differential equations
can be written giving the relations of the currents in the
tubes to the voltage difference between the tubes at any
longitudinal distance, x, from the end of the outer tube.
Using symbols defined as follows:
M|= mutual inductance per centimeter length of the inner
tube, between it and the secondary circuit. Its
value is very nearly the same as the self-inductance
of the secondary circuit if the wall thickness of
each tube is small.
' Taken from Marks' Handbook (or Mechanical Engineers, Fourth Edi
tion, p. 4M

i,R
sinh wiVC|il/i
Putting values of K\ and Kt in eq 60
iiR
cosh ux-JCtMi
sinh ZVC;
at x = l

To get a relation between the voltage impressed on the

CRO cable (ti/J) and the rate of change of the current
being measured (dl/dt), multiply eq 61 by juAfil and take
R out of the brackets

As M\l=M, the total mutual inductance, the above

equation becomes
. .tOiU I =jw/. j
mr
nrj
M it,anh ul-jMiCi +j-,r

(62)

In this equation iiR is the voltage impressed on the cable

going to the CRO and thus is equal to the voltage recorded
by the CRO. The value of M is constant and may be
computed from the dimensions of the mutual inductor.
The right-hand side of this equation is the rate of chance
of the current, /, being measured. The quantity in the
brackets is unity for all but the higher frequencies. Thus,
for most surge work the rate of change of the current being
measured is obtained by dividing the voltage recorded at
the CRO by the constant M. In order to determine the
upper limit of frequency for which this relation holds, the
magnitude of the vector quantity in the bracket must be

410-210

evaluated as a function of frequency. When the hyper

bolic tangent is replaced by the first two terms in its series
expansion the magnitude of this quantity becomes:
3 +2W
As shown by eq 5 and 6 in the main part of this paper, the
product M\Ci = 1.1 IX 10~21 and is the same for any pair of
coaxial tubes provided they have thin walls. I is the
length of the secondary circuit tube in centimeters. M is
the total mutual inductance of the inductor in henries and
can be computed from its dimensions as already shown in
the main part of this paper. R is the surge impedance in
ohms of the cable connected to the potential terminals of
the inductor and going to the CRO where it is terminated
by a resistance equal to R.
For the type I concentric-tube mutual inductor con
structed and used at this laboratory f = 1 7.8 cm and A/ =
O.Oo^h. The surge impedance of the CRO cable, R, is 50
ohms. For these particular values, the magnitude of the
quantity in the brackets of eq 62 varies with frequency as
follows:
Frequency
c/sec.
30 X 10
40
50
60
70
80...

1.
1.
1.
1.
1.
1.

02
03,
06
08,
11,
154

For this example of a type I concentric-tube mutual

inductor, the rate of change of current as measured at the
CRO will be correct to within 10 percent for all frequencies
up to about 65 megacycles.

change is to be measured passes through the inner tube

(radius n) and returns through the outer tube (radius r3).
With this arrangement the current terminals are close
together and the self-inductance of the primary current
circuit is fixed and quite small. The secondary circuit of
this mutual inductor is the same as that of the type I
inductor. Making the same assumptions as stated in
appendix 3, differential equations can be written giving
the relations between currents in the three tubes and
voltage differences between the tubes at any longitudinal
distance, x, from the end of the middle tube (radius r,).
Using symbols defined as follows:
A/i = mutual inductance per centimeter length of the
inner tube, between it and the secondary
circuit. This is the same as the self-inductance
of the secondary circuit if the wall thickness
of each tube is small,
L2 = self-inductance per cm of the circuit formed by
the middle and outer tubes,
C, = capacitance per cm between the inner and
middle tubes,
C2 = capacitance per cm between the middle and
outer tubes,
e, = the potential difference between the inner and
middle tubes at distance, x, from the end of the
middle tube,
e^=the potential difference between the middle and
outer tubes at distance, x, from the end of the
middle tube,
i=that part of the current flowing at x in the middle
tube whose return path is in the inner tube,
t'^=that part of the current flowing at x in the middle
tube whose return path is in the outer tube,
Lq= self-inductance of the circuit formed by the inner
and outer tubes from a to /. The differential
equations are

4. Effective Inductance of a Type II ConcentricTube Mutual Inductor as a Function of

Frequency

dix =juCixdx

(63)

di',=juCtejix

(64)

dex=juMt (I i,)dx

(65)

de',= jaLi(I i',)dx.

(66)

The solutions of eq 63 and 65 are

In a type II concentric-tube mutual inductor as shown
schematically in figure 12 the current, /, whose rate of

e, = Ki sinh [ax-)/C,Mi + Kt]

(67)

U=I+jK, -y/^ cosh [wx VC.M. + K,)

(68)

and
I .
fc,. ir'i.
c*

c
1

p' 1r'

4
'-'

i
1
o

"it
T

The solutions of eq 64 and 66 are

J
~ Tocable
CRO

e't Kj sin [ux1\!L2C2+ Kt]

(69)

i'.=IjKi-^^ cs [uxy1LiC2 + Kt],

(70)

The arbitrary constants JCi, K?, K3, and Kt in the above

equations can be determined by using the terminal condi
tions at 1 = 0 and i = /.
At x = 0, ez = 0, and i, =0. So from eq 67
Figure 12.Schematic drawing of a lypell concentric lube
mutual inductor

411-211

0=Kl sinh 10 + K2],

but Ki cannot be zero, therefore A'j 0.

Also, from cq 70

At x=Z:e.=/i(i| tV) and e,'=/<(iu')+jmLt(lif).

0=7/A'jy^j cos A\

So from cq 07, after putting in value of A',

giving
sinh ul*JCtM,
Also, from eq 00, after putting in value of A',
tan K{= tan ul-jL2Ci+

>V2
/ cos uI^CjLq

Using above values of A'i and A'j cq 08 at z=/ gives

W+/K(W) V
Using above values of A'i and Kt eq 70 at x=l gives

tanh^/Vra

7=7-

("2;
cos djCLj ' ' V Ln tan wl V^iCj [B(t| O +jwio (/ *D)
Subtracting eq 72 from eq 71 and assuming t'(' to be very small compared to / in the juLt term, a value for R (*i ), the
voltage across the CRO cable, is obtained
wl^fM,Ci COS ul V^Cj
,, , /Cj . . nr-fi , vM\l
, /-,=
' , ..7=}r!-*-wMil -t/ r s,n ul\LiC,+j - cos V^Cj
tanh utyA/|Ci
V

(i,-h)R
1 +a.Lo -y/^ sin <->WtC~i
For concentric-tubular construction, if the wall thickness of the tubes is neglected the product of capacitance and
inductance per cm is a constant for any two tubes. So let yMiCi=VLiCi=<i=3.33X 10~". Also let M\l = M, the toUl
mutual inductance, then the above equation becomes
~ual cos wal
IC, .
, , .<>*[
,"
=jM7
1 +<jLo "y j~ s'n
By expanding the first term in the numerator of the quantity in the brackets in series form and neglecting higher
order terms, the equation becomes
1

wM

sin <Ml+J -g- COS IMI

=jwl

CI

1 +wLo -^j^j 8>n <*d

This equation is similar to eq 02 of appendix 3 for a type I
mutual inductor. Thus, to determine the upper limit of
frequency for which a type II inductor can be used, the
magnitude of the vector quantity in the brackets of eq 73
must be evaluated as a function of frequency.
A type II mutual inductor having the following dimen
sions (see fig. 12):
ri = 0.5 in.
r2=1.0 in.
ra=2 in.
1 = 7 in.
distance from a to /=2.5 in.
is used as an example. By use of eq 5 and 6 in the main
part of this paper values for Mi Lj, C2, and Lo may be
computed from these dimensions. Then, assuming the
CRO cable surge impedance, R, to be 50 ohms, the mag
nitude of the quantity in the bracket of eq 73 varies with
frequency as follows:

Frequency
c/sec
10X101
30
50...
00
70
SO
90
100

0.
.
.
.
.
.
.
.

998
982
952
932
908
885
861
835

For this example of a type II concentric-tube mutuil

inductor the rate of change of current as measured at the
CRO will be correct to within 10 percent for all frequence
up to 70 megacycles.
Washington, March 28, 1947.

412-212

JOURNAL OF RESEARCH of the National Bureau of StandardsC. Engineering and Instrumentation

Vol. 66C, No. 3, July-September 1962
Spark-Gap Flashover Measurements for SteeplyRising Voltage Impulses
J. H. Park and H. N. Cones
(March 19, 1962)
Two designs of thin ribbon resistors have been devised which are suitable for high-volt
age surges and have very low time constants (2X 10-8 sec). They were used in making up
dividers for measuring linearly rising chopped impulses with peak voltages up to 300 kilovolts
and times to sparkover from 0.03 to 50 microseconds. Errors in divider ratio due to residual
inductance were found by computation to be less than 1 percent. Stray capacitance errors
were kept low by making total divider resistance 1,000 ohms or less. By a combination of
computation and experimentation, capacitance errors were deduced to be not greater than 1
percent for times to sparkover 0.1 microsecond or greater.
A large number of oscillograph records were obtained of spark-gap flashover voltage
with linearly rising voltage impulses at rates of rise up to 10,000 kilovolts per microsecond.
From these oscillograms data were derived giving a relation between rate of rise (or rise time)
and flashover voltage for (1) 12.5-cm-diameter spheres spaced 6 cm apart, (2) 25-cm-diameter
spheres spaced 6 cm, and (3) 6.5-inch-diameter uniform field electrodes spaced 5 cm. Volttime curves showing these relations were plotted. It is recommended that the curve for
25-cm-diameter spheres be used as a reference standard for interlaboratory comparison of
measurement methods.
2. Requirements for Setting Up Volt-Time
Standards

1. Introduction
The measurement of wave form and peak value of
a steeply rising voltage impulse is of considerable
importance in the testing of high-voltage equipment
such as insulators, transformers, and lightning
arresters. Such measurements are usually made by
impressing the voltage impulse on the high-voltage
terminal of a divider consisting of resistors, capaci
tors, or a combination of both, and connecting the
low side of the divider through a coaxial cable to the
deflecting system of a cathode ray oscillograph
(CRO). For measuring the peak value of full-wave
impulses or waves chopped on the tail, the accuracy
of dividers is well established, and measurement
methods can readily be checked using standardized
sphere-gap breakdown tables [1, 2, 3]
However,
the sphere-gap tables cannot be used for times to
sparkover less than 2 /isec, because for such short
times, the sparkover 2 voltage of sphere gaps in
creases as time to sparkover decreases giving a volttime curve. This was noted in 1935 by Bellaschi
and Teague [41. Hagenguth [5] has shown that rod
gaps ana insulator strings also exhibit a volt-time
effect.
Several years ago, in AIEE Conference Paper No.
57-215, the present authors suggested the use of
sphere-gap volt-time curves as reference standards
for checking measurement methods at times to
sparkover less than 2 usee. Since that time work
reported from other laboratories [6, 7, 8, 9, 10] has
indicated that such a standard could be useful. The
purpose of the present paper is to serve as a per
manent published record of the work done several
years ago and to present results of later work in
this field at the National Bureau of Standards.
' Figures in brackets indicate the literature references at the end of this paper.
' The terms sparkover and flashover are assumed to have the same meaning.

Before volt-time curves can be considered for

adoption as part of a standard, experimental data
must be obtained and published by various labora
tories working in the high-voltage field. Conditions
under which the data for such curves are obtained
must be decided upon, and they should be such
that they can be readily duplicated in all labora
tories. Ordinary laboratory conditions of tempera
ture, pressure, and humidity should be satisfactory,
provided corrections for relative air density are
applied. All comparisons can then be made at
760 mm of Hg pressure and either 20 or 25 C
temperature. No applicable corrections for humid
ity are available, but records should be kept to see
if any correlation between humidity and flashover
voltage can be detected.
Another condition which must be controlled is
the location of the spheres with respect to floor,
leads, divider, surge generator, and any other nearby
objects. International Electrotechnical Commis
sion Publication 52 entitled "Recommendations for
Voltage Measurements by Means of Sphere-Caps
(One Sphere Earthed)" contains very dennite speci
fications for clearance distances when spheres are
used for 60-cycle or full-wave voltages. It would
seem logical, at first, to adopt these same speci
fications for front-chopped wave measurements.
Also, the effect of changes in such distances should
be determined.
In addition to the above, it is necessary to specify
wave shape of the applied voltage before chopping.
In fact, in all steep-front testing it is extrcmely
important to be able to define waveform exactly
and in detail so that results obtained in different
laboratories can be accurately compared. This is

413-197

true not only for chocking measurement methods,

but also for comparing lest data on high-voltage
equipment. Thus, even though other means are
found for chocking measurement methods in various
laboratories, the task of obtaining an accurately
definable wave front must be completed before any
steep-front chopped-wave tests will give results
which can be repealed in other laboratories.
The presently accepted methods of defining wave
front and rate of rise use either the 10 and 90 or
30 and 90 percent voltage points together with some
restriction on superposed oscillations. Such defi
nitions pin down only two points on the wave front
and give no possibility of fixing tolerances. How
ever, if a linearly rising ware front is defined as one
which rises at a constant rate from SO to 100 percent3
of jlashover (or breakdown) voltage, any departure
from the defined wavejorm can be determined. Al
though linearly rising waves, free from oscillations
and curvature, are not always readily obtained in
practice, they can in all cases be attained to within
the required tolerance with a little patience in
choosing and adjusting circuit parameters. Since
they offer the only possibility of setting up a standard
steep-front wave shape which can be accurately
duplicated in various laboratories, it is suggested
that linearly rising waves be adopted as the standard.
This does require setting up an allowable tolerance
from an exactly constant rate of rise and devising
a practical method for measuring rate of rise at
all points from 30 to 100 percent of flashover voltage.
Both of these problems are discussed and a solution
suggested in appendix I.
The remainder of this paper is devoted to (1) a
description of the methods used at the NBS to
obtain points on the volt-time curves, (2) an analysis
of the accuracy obtained, and (3) a presentation of
the results.
3. Impulse Generator Circuit Arrangements
for Obtaining a Linearly Rising Voltage
The circuit arrangement and general precautions
needed to obtain a linearly rising voltage will prob
ably vary considerably for different laboratories.
The impulse generator used in the present work
has a nominal rating of 2,000 kv when its twenty
)i-ti[ capacitor units are connected in series. A
modified connection of the 20 units, putting two sec
tions (10 units in- series for each) in parallel on dis
charge, was used in the present work, giving a rated
discharge voltage of 1,000 kv. The breakdown volt
ages of the gaps used varied approximately from
140 to 300 kv depending on the rate of voltage
rise. Thus, breakdown could always be made to
take place on the linearly rising portion of the
applied voltage, i.e., much before the knee of the
exponential curve.
> (Waveform 1m?1ow the 3fl percent voltape point is not considered to he of
importance in front-chopped testing ljceause here the phenomenon l>einp studied
is dielectric failure for an applied voltape surpe of short duration, and the mecha
nism probably is not initiated until the voltape magnitude approaches the fullwave breakdown value.)

SURGE GEN

Figure 1. Schematic diagram of impulse generator and

discharge circuit used to obtain linearly rising voltages.
A schematic diagram of the impulse generator and
discharge circuit arrangement for obtaining %rarious
rates of voltage rise is shown in figure 1. Rate of
voltage rise was controlled by (1) adjusting the
generator charging voltage and/or (2) changing the
time constant of the R-C circuit made up of "R,"
and "(,'." To keep rate of voltage rise linear up to
breakdown, it was found that the charging voltage
had to be maintained above about 50 percent of
rated value. Thus large changes in rate of rise were
made by using wide variation in /?, and C Com
binations of paper-oil capacitors connected in series
were used to give fixed values of C as follows: 0.06.
0.006, 0.003, 0.002
(for steepest rates of rise
f=0). /i's was made up of double wire-wound card
resistors 2 in. wide and 12 in. long having values of 5,
15, 35, or 125 ohms each. They were used in series
combinations giving values from 35 to 875 ohms.
Because of residual inductance in the paper-oil
capacitors, f1 resonates at about 1.3 Mc/s and
produces oscillations on the rising front of the
voltage across C. The lumped inductance "L"
and capacitance "C0" together with "R0" act as a
filter to give a smoothly rising voltage at the sphere
gap. The inductor L consisted of 100 turns of
polyethylene-insulated wire on a 2-in. diain bakelite
tube, giving an inductance of 50 ph. The capacitor
C\ consist ed of 3 aluminum hemispheres 1 ni in diani,
placed with their curved surfaces down and sup
ported by porcelain-pedestal insulators so that the
curved surfaces were kept about 8 in. above the
conducting floor. The resistor Rn consisted of the
same type of cards as used for Rs, its value being
held constant at 105 ohms (seven 15-ohm cards in
series). Results were obtained with the test gaps
at various distances from the aluminum hemispheres
in order to study proximity effects.
4. Measurement-Methods and Dividers
Resistance dividers in conjunction with a coldcathode CRO [11] were used for all measurements.
Several different divider and gap arrangements (as
indicated in figs. 2, 3, 4, 5, and 6) were tried to
determine possible effects that the location of nearby
conductors might have on either gap flashover or

414-198

42"

GAP 20' FROM C0

55"
1
-r

Figure 4.

iure 2.

VoU-lime cwve for li.S-cm spheres spaced 6 cm,

upper sphere negative.
II voltages corrected to 760 mm of Hg and 25 C. Solid curve obtained with
tl in. bushed curve with -A 16 in. Insert shows sphere gap and divider
lgemcnt.
1,000 ohms.

Volt-time curve fm uniform field gap, spacing 6 cm,

upper electrode negative.
All voltages corrected to 760 mm of Hg and 25 C. Insert shows gap and divider
arrangement. Has- 1,000 ohms. Points obtnined with ^4-20 in. and A = 40
in. all fall on same curve.

rCOM TO
\ CM

34--

-I

\
3" CRO BUS ON FLOOR

j be 3.

Volt-lime curve for S5-cm spheres spaced 6 cm,

upper sphere negative.
voltages corrected to 760 mm of Hg and 25 C. Insert shows sphere gap
iivider arrangement. ^.1 = 1,000 ohms. Other sphere gap and divider
foments also gave points falling on this curve.

Figure 5. Sphere gap and divider arrangements using RHs

= 600 ohms (5-100 ohm ribbon-wound cards) with 12.6-cm
spheres spaced 6 cm.
(Upper) A. Similar to figure 2 except Pus 500.
(Lower) B. With divider very close to spheres.

der response. A 60-ft length of polyethylene

sial cable (RG 8/TJ) was used for the connection
a divider low side to the CRO. The method of
ninating this cable at the oscillograph and
>rmining its attenuation correction have been
*ribed in a previous publication [12]. A check
possible stray pickup voltage in the measuring
uit was made by using a special cable fitting

between the divider low side and the cable to the

oscillograph. This fitting offers no discontinuity or
asymmetry in the cable sheath, but it disconnects the
center conductor of the oscillograph cable from the
divider low side and connects it to the cable sheath.
With this fitting in place there should be no signal
deflection recorded on the oscillograph when the
generator is discharged and a voltage applied to the

415-199

FlOURE 6. Spheie gap and divider arrangement with 25-cm

/spheres spaced 6 cm.
Spheres close to floor. i?//.il,000 ohms.

Figure 7. Douhle-wound card resistor cast in epoxy.

Karma ribbon 1/1S Inch wide anil 0.001 inch thick with 0.001 inch Teflon between
layers. Time constant, /./A =0.(102 X 10"4.
divider. Any departures from a straight zero line
would indicate pickup in the measuring circuits,
probably due to ground currents in the cable sheath,
CRO case, etc. Methods of connecting the cable
to the oscillograph and arranging ground connections
were devised and necessary changes made until this
test indicated negligible pickup.
The divider high side (I?H, of figs. 1 to 6) consisted
of five or ten 100-ohm special Karma * ribbon
resistor units. Each unit (see fig. 7) has an overall
length of 6% in. and consists of two oppositely
wound layers of ribbon (Ke in. wide and 0.001 in.
thick) on a lucite card (1% in. wide, and }U in. thick).
Slots were milled in the edges of the lucite cards to
keep the turns of ribbon uniformly spaced. The
first layer of ribbon was enameled; but since the
enamel did not completely cover the edges, a sheet
of teflon (0.001 in. thick and \% in. wide) was laid
over each side of the lucite card before starting to
wind the second layer. A brass mold was specially
constructed in which the double wound lucite cards
were cast in epoxy resin (epon 815). All air bubbles
were removed by placing the mold in a bell jar while
the epoxy was still liquid and subjecting it alternately
to low pressure (about 1 cm of Hg) and then room
pressure (for at least three cycles). The resin was
cured by baking overnight at about 55 C.
* Karma is the trade name for an alloy (Ni 73 percent, Cr 20 percent-f-Al+Fe)
having a hlph resistivity and low temperature coefficient.

The purpose of using this special construction

was to get a resistor of minimum inductance which
would withstand a high momentary voltage. Min
imum inductance was obtained by using two oppo
sitely wound layers connected in parallel and occupy
ing as nearly as possible the same physical space.
Sparking between turns and layers was prevented
by (1) very uniform spacing and (2) casting in a
high dielectric strength resin. The actual resistance
and reactance of these units were measured at 0.5,
2.5, 10, and 30 Mc/s using a high-frequency bridge
(at low voltage) . Resistance did not change appreci
ably over this frequency range, and values of re
actance were such that they could be represented
by an inductance of 0.2 p.h per unit. Thus, these
ribbon-wound cards for the divider high side have
a time constant {L/It) of 2X10"9 sec, which is
much less than the minimum figure (10X10-9 sec)
usually given for noninductive high-voltage wirewound resistors.
In order to determine how much voltage could be
applied momentarily to these resistors. linearly
rising chopped waves of increasing peak value were
applied. For chopping times of about 0.7 ^sec each
resistor withstood a peak voltage of 60 kv. repeated
at least 80 times with no apparent damage or change
in resistance. For full waves or waves chopped on
their tail such high values of peak voltage could not
be applied because of the heating effects they might
produce in the ribbon material. The instantaneous
temperature of the ribbon is determined mainly
by the energy [ffodt] dissipated in the ribbon per
shot. Full wave tests on the ribbon resistors in
dicated that each unit can withstand up to 300
joules on a single shot without damage provided
sufficient cooling time is allowed between shots.
The momentary resistance change due to this heat
ing was estimated to be less than 0.5 percent because
of the low temperature coefficient of Karma. If
shots are repeated at 1-min intervals or less, the
temperature buildup might cause failure. When
ever a series of shots is to be made, it is advisable
to check resistor temperature occasionally between
shots.
A set of seven resistors for use as the divider
low side (RL, in figs. 1 to 6) was made with values
from 2.60 to 15.8 ohms. This was done so that the
CRO peak deflection could always he set , by choosing
the proper I'L at some value between 65 and 100
percent of full scale. A cable terminating resistor
with taps could not be used at the CRO because
of the special method used for cable termination [12].
Each low-side resistor consisted of elements made
up of Karma ribbon (% in. wide and 0.001 in.
thick) bent back and forth on itself every 2 in. of
its length. All bends were pressed flat m a vise
Small strips of mylar insulation (0.001 in. thick '
were placed between adjacent 2-in. lengths of ribbon.
All folds were then held tightly together between
bakelite blocks and cast in epoxy for permanence.
Each complete low-side resistor consisted of one
two, or three of these elements symmetrically placed
as closely as possible around a coaxial chassis con

416- 200

ector mounted on a 3-in. wide copper busbar.

11 elements were connected in parallel and to the
Mitral coaxial terminal. The coaxial terminal
rovided the connection from the divider low side
> the CRO cable. Measurements on these flat
bbon resistors, using a high frequency bridge and
equencies up to 30 Mc/s, indicated that their
sistance remains essentially constant up to that
equency. Also their reactance can be represented
\ a fixed inductance. The time constant of each
w-side resistor was measured and found to be
>out 2.5 X1(T' sec.
5. Measurement Accuracy
The main sources of error after eliminating stray
ckup, insuring proper cable termination, and cor
ding for cable attenuation are: (1) the oscilloaph and (2) the divider. Repeated checks on the
flection sensitivity of the cold-cathode CRO used
the experimental work indicated that for deflec>ns greater than 65 percent of full scale the accucy of CRO voltage measurements was within
1 percent. Sweep calibrations obtained by coneting the output of a signal generator (1.5 to 18
c/s) to the deflecting plates, indicated that time
sasurements could be made to within 2 percent
total swreep time. Sweeps witb total duration
>m 0.3 to 40 fisec were used. Their linearity was
termined using the method described in appendix
1 , and in measuring linearity of voltage rise only
ose sweep sections found to be linear to within 10
rcent were used.
Divider errors, which ordinarily introduce the
lin source of uncertainty in steep-front measure'nts, depend mostly upon a combination of two
ects: (1) stray capacitance from parts of the diler to nearby conducting surfaces and (2) residual
luctance in the divider elements. If these two
ects are considered separately, a computation can
made of the error introduced by each, based on
asured or estimated values. Inductance errors
re kept low in the present work by using special
bon wound resistors as described in the preceding
tion. Actual values of time constant, as measured
these resistors, were inserted in the formula
ived in appendix 8.2 to obtain computed values
error due to stray inductance. This gave an
ar of 0.5 percent for a rise time of 0.1 jisec and less
)r for longer rise times.
in exact computation of stray capacitance errors
not be so easily deduced because stray capacitance
ends on the location of the divider with respect
ill nearby conductors. For the divider arrangclts used (see figs. 2 to 6), all conductors except
laboratory- floor and the high voltage lead to the
ider were kept at sufficient distance to not affect
der capacitance. The error due to stray capacice from a vertical divider to the grounded floor
ie may be expressed as a time lag t'=RC/6, for a
arly rising voltage, as shown by Bockman and
ten-Cavallius [13]. In this formula R is the
,1 divider resistance and C is the total measured
stimated capacitance to ground. The percent

error decreases as the time to measured value in

creases. Computed values are given in the following
table for both the 1,000-ohm (see figs. 2, 3, 4, and 6)
and 500-ohm (see fig. 5) divider used in the present
work:
Estimated
Total
divider divider capaci
resistance
tance to
ground
Oh mj
1,000
500

Picofaraiin
22
11

Percent error for rise times

V
0.05
usee
0.003f>7
.(KXKH7

0.1 fMtC
-7.1
-1.8

-3.7
-0.9

0.5 )
-0.7
-0.2

The total stray capacitance error would include not

only the component due to capacitance from divider
to floor, as listed in the above table, but also a com
ponent of opposite sign due to capacitance from
divider to its high voltage lead. Thus the total
stray capacitance errors would be somewhat lower
than those given in the above table.
A comparison of experimental results obtained
with the 1,000 and 500-ohm dividers affords a basis
for estimating the cancellation effect of the two
stray capacitance components. As seen in the
above table, flashover voltage measurements ob
tained with the 1,000-ohm divider would be ex
pected to be several percent lower than those ob
tained with the 500-ohm divider unless the
capacitance from the divider to its high voltage lead
compensated for part of the capacitance from divider
to ground. Measurements indicated that both
dividers gave the same results for all rise times down
to about 0.05 ixsoc. This indicated that, in the
present work, errors due to stray capacitance for
rise times greater than 0.1 /isec were probably not
greater than 1 percent.
The above consideration of various possible
errors indicates that the values of voltage here re
ported are accurate to within 1.5 percent for rise
times greater than 0.1 /isec.
6. Experimental Results
Linearly rising voltages with various rates of riso
obtained as explained in section 3 were used. Wave
forms were considered to be linearly rising if the
"true variation in rate of voltage rise" (S'kr) as
defined and explained in appendix 8.1 was less than
20 percent. "Rates of voltage rise" were determined
by the procedure described in appendix 8.1. Riso
times or times to flashover were then obtained by
dividing peak voltage at flashover by "rate of voltage
rise." Typical examples of the impluse waveforms
used are shown in figure 8. In all cases a negative
polarity impluse was applied to the high-voltage
electrode. Peak voltages were taken as those
measured at the first sudden break in the smoothly
rising trace. Other peaks occurring later were
assumed to be due to reflection between sphere gap
and divider. This is most clearly indicated in figure
8K for which the divider was very close to the spheres.
Sufficient records, similar to those shown in figure 8,
were taken so that complete volt-time curves could
be plotted for three different electrodes: (1) 12.5-cm

417-201

10

15 20

fj. sec
0.2

0.1

0.3

0.1

02

0
G

0.2

0.4

0.6

0.8

Figure 8.

Typical samples of oscillograms: linearly rising voltages.

Rise time
A
B
C
D
E
F
O
H
K

t*sec
27
2.7
0.8(1
.13
.093
.050
.037
.032
.046

Peak

For divider
setup see

Rhs

Fin. 4
Fin. 2
Fit!. 4
Fig. 3
Fie. 2
Fie. 6
Fig. 2
Fig. 5A
Fig. 5B

Oft TO
1,000
1,000
1.000
1.000
1.000
1,000
1,000
500
500

kr
135
164
171
243
230
277
266
264
262

418-202

P"
20"
34"
20"
42"
34"
17"
34"
34"

diam spheres spaced 6 cm, (2) 25-cm diam spheres

spaced 6 cm, and (3) 6.5-in. (16.5 cm) diam uniform
field electrodes spaced 5 cm.
The solid line in figure 2 is the volt-time curve for
12.5-cm spheres spaced 6 cm apart obtained with the
divider arrangement, also shown in figure 2, and A
=34 in. The proximity of high voltage conductors
of fairly large surface areas, such as parts of the
generator or the 1 m hemispheres used as "C0", were
found to affect flashover voltage. To completely
eliminate such effects, the gap was kept 20 ft from
any generator or discharge circuit parts while ob
taining this curve. It is an average drawn through
about 200 plotted points. Each plotted point was
obtained from four CRO records repeated at 30-sec
intervals. The rates of voltage rise were always
nearly identical for each of the four traces in a set.
The four values of peak voltage usually agreed to
within 1 percent. Plotted points were obtained for
various rates of voltage rise over a period of several
months and under various laboratory conditions:
(1) ambient temperature 22 to 26 C, (2) atmospheric
pressure 748 to 762 mm of Hg, and (3) relative hu
midity 42 to 68 percent. All voltages were corrected
for relative air density to 760 mm of Hg pressure
and 25 C. [1,2]. No correlation between humidity
and flashover voltage at various rates of rise could
be deduced, possibly because the actual range in
humidity was not very great. Over 98 percent of
the plotted points fell within 2 percent of the
solid curve in figure 2.
Data were also obtained (1) with the horizontal
distance between sphere gap and divider changed
from 55 in., as shown in figure 2, to 20 in. and (2)
using a 500-ohm divider arrangement as shown in
figure 5A (A =34 in.). Points plotted from these
data also fell along the solid curve.
When the distance from the floor to the sphere gap
was changed, points did not fall along the same
curve. A complete set of data for various rates
of voltage rise was obtained for yl=16 in. (see figs. 2
and 5). The dashed curve plotted using these data
is 2 to 3 percent lower than the curve for A = 34 in. as
seen in figure 2.
Some data were also taken using the 500-ohm
divider high side and the gap-divider arrangement
of figure 5B. Here the lead from sphere to divider
was kept as short as possible which accounts for the
sharp break in the CRO record at flashover with no
immediate oscillations (see fig. 8K). The oscillations
occurring at flashover for some of the other divider
arrangements (see figs. 8G and 8H) are probably due
to reflection in the lead from sphere to divider.
Flashover voltage was always taken as the. first
break in the linearly rising trace. For the divider
arrangement of figure 5B the spheres are close to the
floor which tends to make flashover voltage low, but
the divider and its lead are high voltage conductors
fairly close to the sparking point which tend to in
crease flashover voltage. The observed points fell
fairly close to the soild curve in figure 2 showing that
these effects tended to cancel each other.
For the purpose of finding a gap arrangement less
dependent upon the location of the electrodes with

respect to the floor and other nearby conducting

surfaces, it was decided to also try 25-cm spheres
spaced 6 cm apart. Using the divider arrangement
shown by the insert in figure 3, the volt-time curve
plotted in figure 3 was obtained. Procedures used
were the same as those already described for 12.5-cm
spheres. Proximity effects were studied by (1)
changing to the sphere gap and divider arrangement
shown in figure 6 (spheres much closer to floor); (2)
moving the sphere gap and divider setup closer to the
surge generator and the 1-m hemispheres used as C0
(fig. 1), i.e., with the spheres 13 ft from the generator
and 3 ft from the nearest hemisphere; and (3) going
back to the sphere gap and divider setup shown in
figure 3 but with sphere gap close to generator and
hemispheres. Data obtained at various rates of
voltage rise under all of the above conditions gave
points falling along the volt-time curve plotted in
figure 3. Thus, as was expected, 25-cm spheres
spaced 6 cm apart afford a more satisfactory volttime curve for use as a reference standard than 12.5cm spheres spaced 6 cm.
For 60 c/s and full wave impulses, uniform-field
gaps have been found to give more repeatable
flashover voltages than sphere gaps [14, 15], It was,
therefore, considered advisable to try such gaps on
front of wave flashover using linearly rising surges
in the same manner as already described for sphere
gaps. One pair of 6.5-in diam uniform field elec
trodes, rated at 140 kv peak, was constructed from
solid brass according to the specification given by
Bruce [14]. Values of flashover voltage were
obtained at an electrode spacing of 5 cm with various
rates of voltage rise. From plotted points using this
data the volt-time curve in figure 4 was drawn.
Proximity effects were determined by getting data
with sparking point both 20 and 40 in. above the
floor and with the gap near the generator and
hemispheres (3 ft from nearest hemisphere) and at a
much greater distance (15 ft from nearest hemi
sphere). All of these data gave points falling along
the curve in figure 4, which indicated that proximity
effects were much less than for 12.5-cm spheres
spaced 6 cm but about the same as for 25-cm spheres
spaced 6 cm. The scattering of the points about
the curve was very nearly the same for uniformfield gaps as for 25 or 12.5-cm diameter spheres.
The main disadvantage of uniform field gaps was
found to be the difficulty encountered in getting the
electrode axes sufficiently well alined.
All of the data used for plotting the curves in
figures 2, 3, and 4 were obtained with a small needle
of Co60 (equivalent to 0.5 mr per hour at lm) placed
inside the upper electrode. This was done because
Bruce [8] found that full-wave surge breakdown data
on spheres had the least scatter when a small amount
of radium salt was placed inside one of the spheres.
In order to find out if Bruce's conclusion held for
breakdown on a linearly rising surge voltage,
experimental data were obtained at several different
rates of voltage rise with the Co60 removed. The
points plotted from these data also fell along the
curves, and the repeats from shot to shot were the
same as with the Co60 inside the sphere except that
419 203

ftt slow rates of voltage rise (time to breakdown 8

to 10 itsec) repeats from shot to shot were not quite
as good without the Co60. The good repeats obtained
without Co60 can be accounted for by radiation from
other gaps in the surge generator circuit. When a
12-in. diam cardboard tube was placed around the
measuring sphere gap, cutting off this radiation,
the plotted points all fell to the right and above the
curves; also repeat shots for a fixed rate of voltage
rise yielded breakdown voltages differing as much as
10 percent from an average value. These results
indicate that a certain minimum ion density between
the spheres is needed to obtain rcpeatable data, but
probably radiation from other gaps in the surge
generator circuit is sufficient if the measuring gap is
not shielded from them. For very low rates of
voltage rise (times to breakdown, say, greater than
5 /*see) the repeatability was definitely not as good
as for the higher rates; and it was found advisable
to use Co60. An open carbon arc placed '.i ft from
the sphere gap was also tried and found to give
almost as good repeatability as the Co6" for low rates
of voltage rise.
7. Conclusions
Special ribbon resistors with low and nearly
matched time constants were used to make up the
divider high and low sides. From a combination of
experimental results and computations based on these
resistors, it has been estimated that the results re
ported are accurate within 1.5 percent.
Data obtained using 12.5-cm spheres spaced 6 cm
indicated a decrease in flashover voltage of 2 to 3
percent when the distance from laboratory floor to
lowest point on upper sphere was decreased from 34
to 16 in. Variations were also noted when high volt
age parts of the surge generator and discharge
circuit, which have large surface areas, were placed
within a radius of 20 ft from the sphere gap. There
fore, when volt-time curves for 12.5-cm spheres
spaced 6 cm (from various laboratories) are being
compared, the exact configuration of all conductors
within a radius of 20 ft of the spheres should be
considered.
For the uniform field gap, proximity effects were
found to be negligible at distances greater than 4 ft
from the sparking point. Also variations in distance
of sparking point above the floor had little orno effect .
The repeatability of results from shot to shot and
day to day were found to be about the same for the
uniform field gap as for sphere gaps. However, ad
justing (he mounting of the uniform field electrode
so that the electrode axes were in accurate alinement
was found to be a fairly time-consuming operation.
Also uniform field electrodes are not readily obtain
able but must be specially made. Because of these
disadvantages and since uniform field gaps appar
ently give no marked decrease in scattering of flashover voltage for a linearly rising impulse, it was
concluded that they should not be recommended for
setting up a standard volt-time curve.

For 25-cm spheres spaced 6 cm, proximity fir.

were found to be negligible at distances greater!
4 ft from the sparking point, and the variatk :
distance of sparking point above the floor micn
noticeable change in flashover voltage. Suehsp:
are probably available in most high voltage laborsries or they can easily be obtained. Therefore
volt-time curve in figure 3, for 25-cm spheres spa
6 cm, is considered best suited as a reference starus:
for use in checking the accuracy of steep-front r
pulse measurements at various laboratories.
8. Appendix
8. 1 . Definitions for Use in Measuring and Specfc:
Linearly Rising Impulses
A CRO with a linear sweep would be most ca
vemen t for accurate measurement and toleir
specification because a uniformly rising voltage wo.
be recorded as a straight line, and any depirrcoidd be readily detected. If the sweep is not lir
to evaluate uniformity of rate of rise exactly,the <>.;
age trace must be converted to its equivalent r
a linear time base by a computing or ltmh
method. Although most oscillograph sweep? :
not exactly linear over their entire range, this rst: '
complicated conversion can be eliminated ui " cases by devising a method for evaluating
linearity. Then practical limits in sweep nonl;irjity can be specified, and particular sweeps or >k' '
of sweeps which exceed these limits should nor
used.
a. Determination of Sweep Nonlinearity
A sweep calibration is obtained by taking an al
lograph record with a suitable high-freqw:
signal applied to the deflecting system, ft
obtained from measurements on this reeorii <
then be expressed as a plot of time in microsKf
against sweep deflection in inches or centiniC'f
as illustrated in figure 9 for a typical CRO (<*
For a linear sweep this plot would, of course.*'
straight line. Any departure from lineariu
indicated by curvature of the plot. The der
of nonlinearity between any two sweep dtfit(h>'*
and a\, corresponding to time ts and f2, is deter
as follows: draw a straight line tangent to iht?
at a\ (see fig. 9) and call its intersection wit
di ordinate, t\. Sweep nonlinearity in perm.:. is here defined as
&sl=~,
f2 * l 100
for the sweep interval from /, to t2. 8si is J?r
for r, > t[. In general, 5SL may be either pes J
or negative; but usually the same sign is maintc
throughout any one sweep, and its mapiii*' 1
less the shorter the interval f2 fi. This defc
makes it possible to specify the degree of n<:
arity allowable for various types of measure
and in some cases to apply corrections for '
linearity.

420-;204

SWEEP OEFLECTION, cm
cure 9. Sweep calibration curve showing method for measur
ing sweep nonlineirity, S31,

SWEEP DEFLECT ION, M sec

Definitions of Rate of Voltage Rise, Time to Flashover, and

Specification of Tolerances
For a linearly rising waveform, the results of
tiere gap sparkover data can be presented as a
>t of sparkover voltage against either rate of
Itage rise or time to sparkover. Since sparkover
Itage depends mainly on rate of rise just prior to
irkover,5 rate of voltage rise seems the more
ical independent variable. However, volt-time
ves are already a familiar concept in dielectric
nkdown; and since measurement accuracy can
;t be expressed as a function of rise time, it is
night desirable to use time to sparkover as the
ependent variable. The definition of "time to
,rkover" presents no problem if the voltage rise
exactly linear, but truly linear waveforms are
dly ever attained.
t becomes necessary, therefore, to consider
ctly how "time to sparkover" should be deterled.
for very short times to sparkover, in order to get
eatable results it is necessary that tlte rate of
(age rise be very nearly constant over the period
ime "just prior to sparkover." It is this rate of
that determines the magnitude of sparkover
tttpe- Thus, an equivalent or quasi "time to
rfcover" in defined as peak voltage divided by late
he term "Just prior to sparkover" Is intended to mean that portion of the
;c rise N'tween the voltage at which sparkover would occur [or a 60-cycle
je
the actual
occurs
particu",% Mini
shape
of the value
appliedof voltage
voltage.at Itwhich
wouldsparkover
correspond
to for
the the
section
bei eto tinil tm in figures 1'IA and 10B.

SWEEP DEFLECTION. M sec

Figure 10. Draurings of oscillog> ants showing method for
measuring rale of voltage rise anil variation in rate of tise,
is*.
(Upper) A. Voltage Impulse with very short time to flashover.
(Lower) B. Voltage impulse with longer time to flashover.
oj voltage rise just prior to breakdown. To measure
this rate of rise, a straight edge is lined up with the
section of voltage record from about 80 to 100 per
cent of the peak or sparkover voltage. For times
to sparkover less than 1 /xsec nearly all acceptable
waveforms will fall along the straight edge in this
section of the voltage record as indicated in figure
10A. If there is curvature in this section, which is
more likely for times to sparkover greater than 1
jisec, the straight edge should be placed tangent to
the curve at sparkover voltage as shown in figure
10B. The slope of the straight edge in terms of the

421- 205

voltage and time coordinates of the trace is the rate

of voltage rise. If the sweep is linear, this slope can
be obtained by dividing peak voltage, em, by the
time interval between the zero intercept of the
straight edge and peak voltage point, (tmQ, as
shown in figure 10A. Accuracy requirements are
specified by saying that the sweep nonlinearity,
hSL, must be less than 10 percent in the interval used
to measure slope, tmU- Unless this interval is a
fairly small fraction of the whole sweep, the 10
percent limit is quite likely to be exceeded. The
slope can usually still be obtained to the required
accuracy by using a shorter interval such as (tm U)
[see fig. 10A] and a smaller corresponding change in
voltage, (em e,). These measurements, of course,
depend upon the particular sweep being used. It
is probably easier to discard voltage traces occurring
over sweep intervals with high values of 8Sl than to
apply a correction to rate of rise which could be
determined from sweep nonlinearity.
In addition to measuring rate of voltage rise just
prior to sparkover by the method described, it is also
necessary to determine whether rate of rise varies
by a significant amount between 30 and 100 percent
of sparkover voltage. Experimental data obtained
with different wave shapes at the National Bureau
of Standards have indicated that as long as the rate
of voltage rise is constant from the 60-cycle peak
sparkover value of voltage up to actual sparkover,
it is not important that it also remain constant
below the 60-cycle peak sparkover value. Thus,
between the 30 percent voltage point and 60-cycle
sparkover voltage it is only necessary to ascertain
that the voltage rises approximately linearly and has
no abrupt discontinuities. However, in the range
between 60-cycle peak sparkover voltage and actual
sparkover, it is necessary to define a "variation in
rate of rise" and state tolerances. To do this, em
(see fig. 10A) is taken as the point on the voltage
trace corresponding to 60-cycle peak sparkover volt
age. A vertical line through em intersects the sloping
line drawn for measuring rate of rise just prior to
flashover at e'm. Then "variation in rate of voltage
rise in percent," 6RR, is defined as
5jw= Jo=j. 100.

This is a true measure of change in rate of rise from

point (t60, fm) to point (em,tm) provided the sweep is
linear in this range. If the sweep is not linear, con
stancy of rate of rise would be indicated if Srr=Ssl
where both 8RR and 6SL arc determined for the same
time interval tmfM, and their signs are found using
the definitions stated. Then the true variation in rate
of rise, d'RR, can be defined as

A tolerance is specified by saying that the true vari

ation in rate of rise, 8'RR, between the 60-cycle, peak
sparkover voltage and actual flashover voltage shall
not be greater than 20 percent.

The voltage-time trace shown in figure 10A is

typical of those obtained for times to sparkover up
to about 1 fisec. For longer times, the curvature
of the trace is more likely to be reversed in sign as
shown in figure 10B. For very long times to flashover the actual voltage at sparkover approaches thf
60-cycle (full-wave) value, and the tolerances a?
specified, by using em eM, allow considerable curva
ture just prior to flashover. This is not objection
able because flashover voltage changes very slowly
with rise time for times greater than 10 ^sec. To
insure minimum spread of observed values in this
range, it is necessary to irradiate the gap by using
an ultraviolet lamp or some radioactive material
placed inside one sphere.
c. Summary of Definitions and Tolerances
The following definitions and methods for ex
pressing tolerances are proposed for use in obtaiinne
sphere-gap volt-time curves from cathode ray
oscillograph records.
Sweep nonlinearity between times t\ and tt of an
oscillograph record is defined in percentage as
100
where tx is determined as indicated in figure 9.
Uncorrected variation in rate oj voltage rise for a t'RO
chopped wave voltage trace is defined as the cliamre
in rate of rise from point (eM, t^) to point (f, f,
and is given in percentage as
s
^60
oRR=-

^60
/Art
jr i1UU

where e'm is determined as indicated in figure 10A

or 10B.
True variation in rate of voltage rise is defined as
8RR=SRR &SL
where SRR and SSL are determined using the same tirr.>
interval.
Linearly rising wavejront is defined as one rising at
constant rate from 30 to 100 percent of flashovr;
voltage.
Allowable tolerances can be fixed by requinE
that ^,,^20 percent, and that between the 3d
percent voltage point and 60-cycle sparkover point
the voltage shall rise approximately linearly withm1
abrupt discontinuities.
Rate oj voltage, rise is defined as the slope of tV.i
voltage rise trace just prior to flashover. It is ob
tained by measuring the slope of a straight e<i^
which has been either (1) lined up with the sectiiT.
of voltage record from SO to 100 percent of sparkovf
voltage as indicated in figure 10A or (2) pla*
tangent to the voltage trace at the sparkover poo'
as indicated in figure 10B. The sweep time interui
used in measuring this slope should be chosen so
so that 8SL^ 10 percent.
Time toflashover is obtained by dividing peak voltt'at flashover by rate of voltage rise as defined aboft

422 -206

3.2. Inductance Errors of a Resistance Divider

Stray Capacitance Neglected

The measured value of high side voltage is then

^4rm[t+it'a'-1)(n-w2)}

Suppose a linearly rising voltage e(t) = mt 1, is

.pplied across the divider, and
Ri = divider high side resistance
R2= divider low side resistance
11= total divider resistance
Lt = divider high side inductance
i/2=divider low side inductance
= total divider inductance
R/L=a.
'lie divider low side resistance is assumed to be less
dan 10 ohms, thus the CRO cable connected in
iirallel with it can be neglected in a consideration
f inductance errors.
The operational equation for finding current
trough the divider is

(3>

The error in measured value due to inductance is

._e(0-Cm
e{t)

(4)
t
Since a can be easily- be made equal to 10s or greater,
the quantity (1
is approximately equal to
unity for all times greater than 0.05 usee. Thus,
the error becomes
5=\R lij

(R+pL)i=mt 1.
It approaches zero for -k=jy which is true if -jr=jyli l\2
l\\ 1C2
Therefore, the error can be reduced to a negligible
value by making the divider low side time constant
nearly equal to that of the higher side.

The Laplace transform of this equation is

m
(R+sL)I(s)=f
>lving for I(s)
I(s) =

9. References

1
m
L s2(s+a)

(1)

To transform I(s) to a function of time, let

Also, the inverse transforms of/,(s) and /(s) are

(<) and F(t) respectively, thus
l-tF(t) =
d by a well-known theorem in operational calculus

dr.
Jo

\_fter performing the integration, this expression

J?i(t) can be used with eq (1) to obtain
(2)
Jsing this equation the divider low side voltage,
di
e2=i R2-\- L2
be computed as

3X5-4U O - 66 - 28

[1] American standard for measurement of voltage in di

electric tests, AIEE Standard No. 4, ASA C68.1,
(1953).
[2] Recommendations for voltage measurement by means of
sphere-gaps (one sphere earthed), International Electrotechnical Commission Publication 53, (I960) (Ob
tainable from ASA Headquarters).
[3] Methods for measurement of voltage with sphere gaps,
British Standards InstitutionBritish Standard 358,
(1960) 24 pp.
[4] P. L. Bellaschi and W. L. Teague, Sphere-gap char
acteristics on very short impulses, Elec. J. 32, 120
(1935).
[5] J. H. Hagenguth, Short-time sparkover of gaps, Elect.
Engr. 56, 67 (1937).
[6] C. J. Miller, Jr., and J. F. Wittibschlager, Measure
ments of steep-front waves with an isolated screen
room installation, AIEE Communication and Elec
tronics Paper 58-18, 262 (1958) .
[7] F. C. Creed, The measurement of impulse waves chopped
on the front CIGRE Paper 320 (1958).
[8] N. Hylten-Cavallius, High voltage measuring devices
and measuring errors, ASEA Technical Memorandum
TM 9184 (1958).
[9] Unpublished report by P. R. Howard of the National
Physical Laboratory at CIGRE Meetings in 1958.
[10] Informal results from D. L. Whitehead of Westinghouse
Electric Corporation in Pittsburgh.
[11] John H. Park, A fifty-fold momentary beam intensifica
tion for a high-voltage cold-cathode oscillograph, J.
Research NBS 47, 87 (1951) RP2231.
[12] J. H. Park, Surge measurement errors introduced by
coaxial cables, Communications and Electronics,
343, (1958), AIEE Trans. Paper 58-110.
[13] Marius Bochman and Nils Hylten-Cavallius, Errors in
measuring surge voltage bv oscillograph, ASEA
Ludvika Sweden, (1946).
[14] F. M. Bruce, Calibration of uniform field spark gaps for
high voltage measurements at power frequencies,
J. IEE (London) 94, pt II No. 38, 138 (1947).
[15] F. M. Bruce, High voltage spark discharges, Endeavour
XII, No. 50, 61 (1954).
(Paper 66C5-96)

423-207

Journal of The Washington Academy of Sciences

Volume 38, Number 2, February 1948

The Measurement of High Voltage

F. M. Defandorf
This paper outlines the basic principles used in the precise
measurement of direct-, alternating-, surge-, and pulse-voltages,
100 kilovolts and higher. The use of potential dividers, potential
transformers, electrostatic instruments, generating voltmeters, air
spark-gaps , and other devices in the measurement of high voltage
is presented. A discussion of measuring high voltage by the deflec
tion of free-moving charged particles and a description of the
High Voltage Scale is also included.
Editors
August 28, 1967

424-1

Dielectric

and

Magnetic

Measurements

Papers
Standard tests for electrical properties, A. H. Scott
Two-terminal dielectric measurements up to 6X108 Hz, M. G. Broadhurst and A. J. Bur
Basic magnetic quantities and the measurement of the magnetic proper
ties of materials, R. S. Sanford and I. L. Cooter
The calibration of permanent magnet standards, I. L. Cooter

427
431
439
477

Abstracts
Electrical testing, A. H. Scott
An ultra low frequency bridge for dielectric measurements, D. J.
Scheiber
.
.
Low frequency dielectric behavior, W. P. Harris
425

485
485
486

JOURNAL

A comprehensive and useful review of various standards

available in the field of polymer electrical testing.

Standard

Electrical

About the Author

Dr. Arnold H. Scott
was born in Sutherland,
Iowa. He took his
undergraduate work at the
University of Mississippi and
obtained his PhD degree from
Johns-Hopkins University
in 1 928. He has been with the
National Bureau of Standards
nee 1 924 working in the field
of dielectric measurements of
lectrical insulating materials.
He is currently Chairman of
ASTM Committee D-9 and
ASA Committee C-59, both
dealing with electrical
insulating materials. He was
the U.S. Delegate to various
meetings of Technical
Committee No. 1 5 of I EC
from 1 952 to 1 960. He is a
member of the Washington
Academy of Sciences and the
New York Academy of
Sciences.
*E JOURNAL, NOVEMBER,

Tests

for

Properties

Arnold H. Scott,
Dielectrics Section, Electricity Division,
National Bureau of Standards, Washington 25, D. C.

{lectrical properties usually con

sidered to be important when plastics
are used for electrical insulation are:
1. Volume and Surface Resistivi
ties
2. Permittivity (Dielectric Con
stant) and Dissipation Factor
3. Dielectric Strength
4. Arc Resistance or Tracking.
It is generally agreed that there is no
such thing as "good" or "poor" elec
trical propertiesonly properties that
indicate different degrees of suitabil

ity for specific applications. There

fore, before any electrical measure
ments are undertaken, the purpose of
the testing and the end use or appli
cation for which the plastic is to be
used should be known.
Electrical properties, particularly
permittivity and dissipation factor,
are often studied in relation to the
chemical structure and physical char
acteristics of the material. There is
an extensive literature on this subject
but this is beyond the scope of the
present paper which is primarily

concerned with the description and

use of test mediods.
Significance of Tests
Volume and Surface Resistivity
The exact values of volume and sur
face resistivity ordinarily are not very
important because they are usually
several orders of magnitude greater
than required for properly insulating
the electrical circuits involved. How
ever in certain applications such as
insulation for very low current meas
uring devices and for computer com
ponents, very high resistivities are
required. Accurate values are then
needed for the selection of the proper
material and for design purposes.
Also volume resistivity is often used
in production control. Changes in
composition or amounts of impurities
are usually reflected in changes of
volume resistivity before they are re
flected in changes in other properties.
Permittivity and Dissipation Factor
Many plastics are used today in
applications which require an ac
curate knowledge of their permittiv
ity and loss. Various types of im
provement in the composition and
fabrication of plastics have tended to
lower the dissipation factor, and have
increased the need for more sensitiv
ity in the measurement of this char
acteristic. Dissipation factor is often
used as a check on the constancy of
quality in the production of plastic
materials.
Dielectric
Strength Dielectric
strength can be used as a relative
test for material inspection or quality
control, or as a convenient prelimi
nary test to determine whether a
material merits further consideration.
A comprehensive statement of signifi
cance appears as Appendix I of ASTM
D 149.
Arc ResistancePlastic materials
are sometimes used under conditions
where flashovers occasionally occur
between metal parts attached to or
supported by the plastic material.
The flashovers and resulting arcs
may eventually damage the insulat
ing material by tracking or erosion
to the extent that it is no longer use
ful as an insulation. The arc resist
ance test is an attempt to evaluate
materials with regard to their ability
to resist tracking or erosion.
Standards Activity
Much time has been spent in de
veloping adequate test methods for
determining these properties. Im
proved techniques of measurement
have made it necessary to revise these
methods from time to time. Several
are under revision at the present
time.
Standard methods for determining

these properties have been prepared

by Committee D-9 on Electrical In
sulating Materials of ASTM. Inter
nationally this work is performed by
Technical Committee No. 15 on In
sulating Materials of the Interna
tional Electrotechnical Commission.
ASTM Committee D 20 on Plastics
looks to Committee D-9 for electrical
test methods for plastics. Likewise
Technical Committee No. 61 on
Plastics of the International Standards
Organization looks to IEC Technical
Committee No. 15 for international
electrical test methods for plastics.
Volume and Surface Resistivity
The basic methods of test for in
sulation resistance and volume and
surface resistivities are given in ASTM
D 257-61, Methods of Test for Elec
trical Resistance of Insulating Mate
rials. The comparable international
document is IEC Publication 93,
Recommended Methods of Test for
Volume and Surface Resistivities of
Electrical Insulating Materials.* Pub
lication 93 does not include the test
method for insulation resistance but
otherwise is essentially the same as
ASTM D 257. A separate document
for insulation resistance test methods
is now being balloted in IEC. Except
for a few variations in electrode sys
tems, this document is essentially the
same as that part of ASTM D 257
which deals with insulation resist
ance.
Electrodes ArrangementsFor in
sulation resistance measurements al
most any shape of electrode can be
used provided it can be applied to or
embedded in the specimen. However
a value of insulation resistance has
strict applicability only when the
test specimen and electrodes have the
same form as required in actual use
because insulation resistance includes
both volume and surface resistance.
To determine separately the vol
ume and surface resistances (and thus
resistivities) it is necessary to use a
guard electrode system. When prop
erly used this shunts the unwanted
currents around the current measur
ing instrument or bridge detector.
Precautions must be taken to see
that the resistance between a meas
uring electrode and the guard elec
trode is high compared to that of the
input of the current measuring in
strument or the pertinent bridge
element. Otherwise, part of the cur
rent which is to be measured by the
voltmeter-ammeter method will be
shunted past the instrument and will
lead to erroneous results. In the
bridge, the bridge ratio will be up
set.
The most popular arrangement of
electrodes for plate or sheet materials
428-1376

is the parallel guard-ring type, some

times called the "bulls eye" pattern.
W. G. Amey (1) has shown that the
most practical dimensions for these
electrodes includes a gap distance
between .the measuring and the guard
electrodes equal to twice the thick
ness of the specimen. This permits
the same electrodes to be used for
volume and surface resistance meas
urements. The effective area for the
measuring electrodes extends almost
to the center of the gap and ASTM
D 257 gives a correction for the dif
ference between the area to the
center of the gap and the effective
area which can be applied if desired.
However this correction is usually
too small to be significant in view of
the fact that duplicate measurements
of resistance on dielectric specimens
will seldom agree within 10% and
the disagreement is usually 100% or
more. Except in unusual cases the
order of magnitude of the resistance
is all that is required.
Electrode MaterialsVarious mate
rials can be used for the contact elec
trodes which are applied to the
specimens. Conductive silver paint is
very popular because it permits mois
ture penetration to the specimen dur
ing conditioning. However the sol
vent in the paint may affect the
specimen. This can be especially
troublesome for plastic materials. A
study is now being made of conduc
tive rubber electrodes which can be
pressed against the specimen. When
these can be used it saves consider
able time in sample preparation.
Other materials such as evaporated
gold or tinfoil are used for electrodes
but they have the disadvantage of
obstructing the movement of mois
ture into the specimen during con
ditioning.
ElectrometersNeither ASTM D
257 nor IEC Publication 93 include
the use of electrometer type current
measuring instruments although they
do not exclude them. Neither do they
discuss the problems which may be
encountered in their use. Vacuum
tube electrometers are now available
which when used in the voltmeterammeter method will detect con
ductances (reciprocal resistances) as
low as 10"" mho and which are small,
light weight and readily portable.
The present purity of some plasties
makes even greater sensitivity desir
able if the conductances involved are
to be measured with reasonable ac
curacy. Vibrating reed electrometers
are somewhat more sensitive than
vacuum tube electrometers but are
heavier and are more cumbersome to
Copies of IEC and ISO documents may >j
obtained from the American Standards Associa
tion.

use. All electrometers have high in

put resistances (up to 10" ohms or
more at highest sensitivities) and
create problems of measurement if
careful precautions are not taken. The
leakage resistance between the meas
uring electrode and guard or ground
in the voltmeter-ammeter method
must be at least two orders of mag
nitude greater than the input resist
ance of the electrometer if tie ground
is connected to one side of the elec
trometer. Otherwise some of the cur
rent which it is desired to measure
will be shunted past the electrometer
and cause an erroneous measurement
to be made. If the ground is attached
to one side of the specimen the prob
lem is greater because then the leak
age to ground must be at least 2
orders of magnitude greater than the
resistance being measured (2).
Cap WidthThere is some ten
dency, when volume resistance meas
urements only are required, to make
the gap between the measuring and
guard electrodes small so that the
effective area can be more accurately
obtained. However a narrow gap
may cause very large errors when
electrometers are used in the volt
meter-ammeter method because the
surface resistance between these
electrodes may become less than the
input resistance of the current meas
uring instrument and only a small
fraction of the desired current is
measured. This is especially apt to
happen when measurements are made
with the specimen in a high humid
ity. When the integration method is
used with the vibrating reed elec
trometer to get maximum sensitivity,
decaying currents in the insulating
supports induce charges in the meas
uring circuit which may be an ap
preciable fraction of the charge flow
ing through the specimen. Thus
measurements requiring long times
of charge integration may and usu
ally do give quite erroneous results.
Permittivity (Dielectric Constant)
and Dissipation Factor
The general Standard for measure
ment of permittivity and loss in the
U.S.A., is ASTM D 150, A C Capaci
tance, Dielectric Constant, and Loss
Characteristics of Electrical Insulat
ing Materials, which is now under
active revision. A comparable inter
national document has been prepared
by IEC TC/15 and this has been dis
tributed as a Secretariat Document
for comments from the various Na
tional Committees. Standards which
have been prepared for special plastic
materials and special techniques are:
ASTM D 1673, Dielectric Constant
and Dissipation Factor of Expanded
Cellular Plastics used for Electrical
Insulation; ASTM D 669, Dissipation

Factor and Dielectric Constant paral

lel with Laminations of Laminated
Sheet and Plate Insulating Materials;
and ASTM D 1531, Dielectric Con
stant and Dissipation Factor of Poly
ethylene by Liquid Displacement
Procedure.
Guard-ring ElectrodesThe basic
and classical method of accurately
determining the permittivity of di
electric materials is the use of parallel
guard-ring electrodes. Highly con
ducting contact (or specimen) elec
trodes (such as tinfoil, evaporated
gold etc.) are applied to the opposite
sides of a sheet or plate of the mate
rial. The electrodes on one side are
arranged with a narrow gap between
the guarded electrode and the guard
electrode. Aside from an accurate
measurement of the capacitance, ac
curate measurements of area and
thickness must be made. Accurate
area measurements require symmetri
cal guarded electrodes and very small
and uniform gaps between the
guarded and guard electrodes. These
are often difficult to obtain with the
uniformity required for the desired
precision. Also accurate thickness
measurements generally require that
the faces be quite flat and that of the
specimens be uniform in thickness
requiring special preparation of the
specimen either by precision molding
or by grinding or machining the sur
faces. A concept of the magnitude of
the problem can be obtained by con
sidering that if an accuracy of 0.1%
in permittivity is desired, the thick
ness of a specimen which is 0.3 cm
thick must be determined to 0.0003
cm (3 microns) or better. This is im
possible with micrometers unless the
surfaces are nearly optical flat and
the material is hard. In addition, the
effective diameter of a guarded elec
trode which has a diameter of 5 cm
must be determined to 0.002 cm or
better. It is difficult to cut a foil elec
trode with this uniformity and the
optical problems of determining the
exact edge of the electrode are great.
However, an accuracy of 0.5% in
permittivity can generally be obtained
when reasonable care is taken in pre
paring the specimen and making the
various measurements.
The guard-ring method has the
disadvantage that measurements of
capacitance and dissipation factor
must be made on three terminal or
guarded bridges. This usually means
that cumbersome and time consum
ing bridge balancing is required and
such bridges usually will not ordi
narily operate accurately above 100
kc/s. This disadvantage is being par
tially overcome by the use of trans
former bridges where the capaci
tances to ground cause negligible
429-1377

errors. However, commercial bridges

of this type will not at present oper
ate much above 10 kc/s with a sen
sitivity comparable to that of a
Schering bridge with a Wagner
earthing arm.
Micrometer-electrode Holder
A two-terminal technique which
gives almost as good accuracy as the
guarded-electrode technique is the
use of the micrometer-electrode
holder. This can be used at frequen
cies up to 100 Mc/s. This requires a
specimen with a uniform diameter,
with flat, parallel faces and with the
side perpendicular to the faces. The
edges must be sharp, with no round
ing off. The specimen should have a
diameter which is smaller than that
of the holder electrodes by at least
twice the thickness of the specimen.
Measurements are often made with
specimens the same size as the holder
electrodes but the value obtained in
this case is in error by 0.2% to 0.5%
depending upon the permittivity and
thickness of the specimen. Correc
tions to be applied in this case have
not been determined. The precision
with which the spacing between the
holder electrodes can be measured
enters into the accuracy of the
method. The micrometer used to
make this measurement should not
be used to move the electrode to
adjust the spacing, but only to sense
its position. A disadvantage of this
method is that the leakage over the
side of the specimen enters into the
dissipation factor measurement. The
leakage distance is equal to the thick
ness of the specimen and the contri
bution of the leakage can be quite
large especially if measurements are
made in a humid atmosphere.
Edge Correction Method
To decrease the effect of surface
leakage, contact electrodes can be
applied to a specimen in such a way
that the specimen extends beyond
the electrodes. Instructions for com
puting the edge corrections required
for specified electrode arrangements
are given in both ASTM D 150 and
the proposed IEC document. Another
correction is required, that for the
ground capacitance of the ungrounded
electrode. This ground capacitance is
a function of the area of the un
grounded electrode and its distance
from the ground plane and surround
ing objects. This capacitance has not
been adequately determined although
a formula for its computation is given
in ASTM D 150. The best procedure
is to mount the specimen so that it
is as far from the ground plane and
other objects as is practicable. Be
cause of the uncertainties in the de
termination of the edge and ground

capacitances errors in tlie permittiv

ity determination may range from
one to five percent.
Liquid Displacement MethodA
two-lenninal method which is both
accurate and rapid under limited con
ditions is the liquid displacement
procedure (ASTM D 1531). This
eliminates the need for contact elec
trodes on the specimen. If the speci
men is in the form of a sheet or thin
plate and if a liquid can be found
whose permittivity is known and is
close to that of the specimen, the
thickness need not be known with
great accuracy to obtain a very ac
curate measurement of the permittiv
ity of the specimen. However the
liquid must not absorb into the speci
men or otherwise affect it appreciably
during the measurement. Absorbed
liquids even in minute (barely weighablc) quantities can sometimes ap
preciably affect the electrical prop
erties. With care this method can be
used up to 100 Mc/s. If the permit
tivity is close to that of the specimen
this method will yield values of per
mittivity which are in error by no
more than 0.1%.
Air-Cap MethodAnother method
of determining the dielectric con
stant without the necessity of apply
ing contact electrodes is the use of
an air gap in series with the speci
men. This method requires a cell
with guard-ring type electrodes and
thus 3 terminal measurements are
necessary which limits the frequency
range. Also the electrode spacing and
the thickness of the specimen must
be determined with greater accuracy
because the error in the dielectric
constant is greater than either the
error in the spacing determination or
die thickness of the specimen for this
type of measurement.
Dissipation FactorThe determi
nation of dissipation factor for plas
tics is important because the neat
generated in the plastic per cubic
centimeter is directly proportional to
the loss index (formerly loss factor),
which is the permittivity times the
dissipation factor In many applica
tions this heat loss and thus the loss
index must be kept as low as pos
sible. Some plastics have such low
loss indexes that values as low as
0.00005 must be measured to charac
terize the material. Under these cir
cumstances losses in the specimen
holder or cell become important.
Losses in the insulation supporting
the insulated electrode of the holder
and even oxide layers on the surfaces
of the electrodes can produce serious
errors in the measurement of the loss
index of such a specimen. At high
frequencies series resistance losses in
the contact electrodes may become

appreciable. To remedy this, the dis

sipation factor is often measured
separately from the dielectric con
stant with no contact electrodes ap
plied to the specimen. Otherwise
special effort must be made to see
that the contact electrode is thick
(conductive) enough so that the re
sistance from any point of contact to
any part of the area makes a neglig
ible contribution to the loss measure
ment.
Dielectric Strength Test
Dielectric breakdown tests usually
measure the breakdown voltage at
weak spots in the dielectric caused
by dielectric defects of various kinds.
Solid commercial electrical insulating
materials generally contain dielectric
defects and therefore intrinsic dielec
tric strength cannot be determined
from breakdown tests. Furthermore
cumulative heating may develop in
local paths within the material and
breakdown may then occur because
of thermal instability.
The general test for dielectric
strength in the U. S. is ASTM D 149,
Dielectric Breakdown Voltage and
Dielectric Strength of Electrical In
sulating Materials at Commercial
Power Frequencies. A similar docu
ment is being prepared by IEC TC/
15. However disagreement over the
types of electrodes which should be
used has not yet been resolved. There
is a group which insists that for sheet
and plate materials the electrode
system should be a disk (2.5 or 5 cm
in diameter) opposite a larger flat
metal plate, rather than two oppos
ing disks of the same diameter which
is favored by the U. S. and others.
The outcome of this argument will of
necessity influence standards in the
U. S. Although certain electrode sys
tems are specified in ASTM D 149,
special electrode systems are specified
in some U. S. specifications for elec
trical insulating materials.
A study is being made of dielectric
breakdown at higher frequencies.
Breakdown at the higher frequencies
may be more informative for plastics
than that at 60 c/s. It is hoped that a
standard method for higher frequen
cies will soon become available.
A standard which is useful in test
ing the quality of thin plastic mate
rials is ASTM D 1389, Tentative
Method for Dielectric Proof-Voltage
Testing of Thin Solid Electrical In
sulating Materials. Most thin plastic
materials have occasional dielectric
defects, such as fine pinholes and
perhaps fibers. This test serves to in
dicate the frequency of occurrence
of such defects.
Arc Resistance or Tracking
A method of test which has been
430-1378

devised to determine the relativeability of various insulating materials

to withstand the deteriorating effects
of low current arcs is ASTM D 49.5,
Standard Method of Test for IlighVoltaec, Low-Current Arc Resistance
of Solid Electrical Insulating Mate
rials. This is a relative type of test
and occasionally docs not correlate
with results from actual use. How
ever, it is a very useful test in gen
eral for the limited range of condi
tions specified in the test. The test
arcs are caused by overvoltage be
tween the electrodes under dry con
ditions.
The International Elcctrotcchnical
Commission has issued a method of
test for determining the relative abil
ity of materials to resist the deterior
ating effects of small arcs under
specified moist conditions. It is pub
lished as IEC Publication 112, Rec
ommended Method for Determining
the Comparative Tracking Index of
Solid Insulating Materials under Moist
Conditions. This test can only make
a relative comparison but it is gen
erally felt that this test correlates
with life service in the Scandinavian
countries, Holland and in areas ex
posed to moisture. In this test a po
tential insufficient to cause arcing is
applied between electrodes placed
on the material under test. Drops of
water having a specified salt content
are allowed to fall on the material
surface between the electrodes. Small
arcs are formed between droplets on
the surface and this leads to deterio
ration of the surface.
Various methods of testing tracking
resistance under moist conditions are
being studied and tested in the Unite*!
States. A test called the Method of
Test for Dust-and-Fog Tracking and
Erosion Resistance of Electrical In
sulating Materials is presently being
balloted in ASTM Committee D-9.
This could be useful in testing plastics
for use under special conditions where
surface contamination and moisture
are problems.
Literature References
1. Amey, W. C, and Hamburger,
Ferdinand, Jr., A Method of Eval
uating the Surface and Volume Re
sistance of Solid Dielectric Mate
rials, Proceedings of the American
Society of Testing Materials. Vol.
49, 1949.
2. Scott, Arnold H., Insulation Re
sistance Measurements, Fourth
Electrical Insulation Conference,
The Shoreham Hotel, Washington,
D. C, AIEE T-137-52, Page 115,
Feb. 19-20, 1962.
the end
Edited by
Dr. Charles F. Ferraro
FWC Corporation

JOURNAL OF RESEARCH of the National Bureau of StandardsC. Engineering and Instrumentation

Vol. 69C, No. 3, July-September 1965

Two-Terminal Dielectric Measurements Up to 6 x 108 Hz

Martin G. Broadhurst and Anthony J. Bur
(February 23, 1965)
A two-terminal dielectric specimen holder has been constructed and used to make
dielectric constant and loss measurements on a single disk specimen at room temperature
over a frequency range from 10~2 to 6X 109 Hz. The measurement procedures are outlined
and a detailed analysis of the working equations and measurement errors is presented.

1 . Introduction

of the measurement techniques used here are related

to techniques described in a previous publication
from this laboratory [3].

It is often important for the study of dielectric

properties that measurements be made on a single
specimen using a minimum amount of the material 2. Design and Construction of Specimen
over as wide a range of temperatures and frequencies
Holder
as possible. Low-frequency measurements can con
veniently be made on a single disk-shaped specimen,
A schematic diagram of the specimen holder is
but at higher frequencies conventional equipment
shown in figure 1. The case, cap, and electrodes are
requires the use of cylindrical or rectangular speci
mens of various sizes to fit different microwave made of brass, gold-plated to reduce surface losses
guides and cavities. If the dielectric properties of a at high frequency. The lower electrode, which is
material are sensitive to differences in specimen operated at high potential, is supported and insulated
preparation, then the use of different specimens to from the case by a fused silica ring, which in turn is
cover the desired frequency range could lead to supported by the bottom rim on the holder case.
results which seriously misrepresent the frequency The lower electrode and silica ring are held fixed by
dependence of the dielectric constant and loss even small beads of epoxy cement. The upper electrode
though the individual measurements may themselves is connected securely to a brass pin, finely threaded
on the upper end to accommodate the spacing nut.
be quite accurate.
The two-terminal parallel-plate dielectric speci
men holder described in this paper was designed to
make measurements on single disk-shaped specimens
(about 1 to 2 g of material) over as wide a range of
frequencies as possible. Minimum specimen and
holder size together with simplicity of construction
and operation have been emphasized in order that
the cell could easily be fitted with an insulating
jacket and operated at temperatures from 200 to
-!-200 C. The variable temperature operation of
the holder will be reported in a later paper.
In this paper, we will discuss the construction of
the holder currently in operation at room tempera
ture, and develop the equations for its use with the
low-frequency Scheiber bridge [l],1 the General
Radio Mode'ls 716C Schering Bridge and 1615A
Transformer Bridge, and the Boonton Radio Corp
oration Models 260A, 280A, and 190A (^-meters.
Dielectric constant and loss data obtained with the
holder for a disk of commercial poly(methyl methacrylate) over a frequency range of 10"2 to 6X 108 Hz
will be presented to illustrate the use of the holder.
The experimental errors will be analyzed and dis
cussed in detail. Use of the holder to measure liquid
specimens is described in a separate paper [2]. Some
Figure 1. A schematic diagram of the two-terminal dielectric
specimen holder.
' Figures in brackets Indicate the literature references at the end of this paper.

431-165

The compression spring holds the spacing nut against

the shoulder on the holder cap and rotation of the
spacing nut raises or lowers the electrode and hence
regulates the spacing. The spacing between the
electrodes is determined by measuring the position
of a steel ball, mounted at the top of the upper
electrode assembly, with a micrometer barrel mounted
rigidly to the holder cap. Electrical connection to
the upper (ground potential) electrode is made by
soldering four flexible gold foil strips between the
upper electrode and case. Such an arrangement was
preferred to conventional metal bellows mounting
because of the former's shorter electrical path.
Electrical connection to the holder is made through
two connector caps soldered to the case and lower
electrode, which fit snugly over pins suitably attached
to the measuring equipment. The holder electrodes
are 2.54 cm in diameter. The micrometer measures
10 n per division and readings are estimated to 1 n
(1 m=10-6 m).
3. Theory of Measurements
Each dielectric constant and loss determination
involves a specimen in and specimen out measure
ment of the holder impedance at the measurement
frequency, and two calibration measurements of the
empty holder capacitance at some convenient audio
frequency. The calibration measurements can either
be done separately and compiled in the form of a
curve of measured capacitance versus electrode
spacing which is then applicable to all measurement
frequencies, or they can be performed at the time of
each determination with increased accuracy.
Figure 2 shows a schematic diagram of the twoterminal holder for the two measurement frequency
conditions, together with the assumed equivalent,
circuit and the effective series resistance and capaci
tance of the cell. The symbols used in figure 2 are
explained in table 1. The two audio-frequency
calibration measurements can be represented in a

Figure 2. A schematic representation of the two-terminal

holder with (a) and without (b) a specimen and the corre
sponding assumed equivalent circuits.
The symbols arc defined in table 1.

way similar to figure 2b except that the effects of

lead inductance can be ignored. When the measure
ment frequency is itself in the audio range then the
general measurement procedure outlined here can
be simplified as is described in the following section.
Table 1. Description of symbols used in text
d= t'tt ! = equivalent parallel capacitance of the
specimen.
A, = equivalent parallel vacuum capacitance
C,= to
*
of the specimen.
/f=equivalcnt parallel resistance of the specimen.
C,= eo -V = capacitance of the airgap between the
ti *
specimen and electrode.
A
C2=to = capacitance berween adjacent areas A, of
(j
the electrodes. C, corresponds to t|.
Ce=the capacitance reading of the standard capacitor
in a bridge when the cell is connect<?d. Ci
corresponds to (i and CC} to t}.
Ct= capacitance between the remainder of the elec
trodes including that portion outside the area
A, and the edges. C,, corresponds to spacing
ti and Crj to f2.
Cm i = capacitance between the leads when the holder is
in the measurement circuit.
C' = error capacitance due to distortion of the electric
field (see discussion of errors),
fijwi, txi= equivalent resistance and inductance of holder
leads when the holder is in the measurement
circuit.
Ryi, Cji= effective series resistance and capacitance of
holder with specimen in.
Cu= capacitance between the leads when the holder is
in the calibration circuit.
A/=the difference in frequency between the upper and
lower half-power points of an L-C-R series resonant
circuit.
RMm Cm, effective series resistance and capacitance of
holder with specimen out.
/ A, thickness and area of specimen,
(i, /2 = separation of electrodes at settings 1 and 2.
e' = dielectric constant.
o=permittivity of free space.
"=diclectric loss.

Figure 3.

432-166

Ib)
Two simple transformations used to derive the
working equations.

In order to relate the equivalent circuits in figure

2 to their effective measured circuits, the approximate
transformations shown in figure 3 were used. The
approximations will be suitable for most low loss
materials such as many polymers, glasses, and
organic compounds. The validity criteria are shown
in figure 3, and in general if e"/e'<0.1 the trans
formation approximations will introduce less than
1 percent error in the results.
At the measurement frequency, the effective
measured series capacitance and resistance of the
holder with specimen in, CMi and RMt, and with
specimen out, CMo and /?., can be expressed as
follows :

(1)
1 LjttiCm
(2)

e' one finds

ti-t,

(Cc-Cei)

C'
cn-1
(ii)

xi

The last term in brackets is the airgap correction

term whose value is close to unity. It should be
noted that this airgap term will be used in this form
throughout the text. The effect of the error capac
itance C is also small (see discussion of errors) and
hence the value of t' is primarily dependent on the
difference between two empty holder capacitances
measured at audio frequencies and on the sample
dimensions and initial holder spacing.
Since the holder is a part of the measurement
circuit when the specimen is both in and out of the
holder, the lead losses RMt can be eliminated by
combining eqs (2) and (4) , giving,
ARm-Rm^Rm,-^ [__g_J

2 7"He Ar
CMo
Mo -I
1 & IjmV-Ho

(3)

RmoRmi

(4)

(12)

or
. j~C/,,(^+^)J
(13)

where
(5)
and
Cho Cm I ~~\~ Ce- -f- Cz.

(6)

Unfortunately, there is not a convenient way to

make resistance substitutions for the specimen loss
so that the calculation of Rx will depend on the par
ticular methods used at the measurement frequency.
4. Measurement Procedure

One, can write an expression for CT which is inde

pendent of frequency or the type of electrical equip
ment used. If ('m0 is experimentally set equal to
0Mt, eqs (1) and (3) can be equated with the result,
cK[c2+(a2-cri)-c}
"Ot-[Ci+(C.i-C,1)-C]

(7)

"rom the calibration measurements (circuits corre.pond to that of figure 2b ignoring the lead inductnce),

('c2 Cic -\- CtJ-f- d ,

(8)

ne finds
Ct+(Cti-Ctl)=(Cei-Cei+Ci),

(9)

rliich can be substituted into eq (7) to give

Cx=

C,[(Ce3-CCl) + C,-C]
'Cf-[(6V-C'ei) +<?,-<?']

(10)

Writing C and & in the numerator in terms of

leasured quantities and applying the definition of

4.1. 0-Meter Procedure of Measurement

For Q-meter measurements in the frequency range
105 Hz to 2.5X108 Hz, two ^-meters were used, the
Boonton Radio 260A O-meter (100 kHz to 20 MHz)
and the Boonton Radio 190A Q-meter (20 MHz to
250 MHz). The measurement procedure used is as
follows: The holder is mounted on top of the (?-meter
and connected in parallel with the Q-meter capacitor,
C with the shortest possible leads. With the
specimen in the holder, the circuit is brought to
resonance by tuning the Q-meter capacitor, C to
C't. The values of C's, the electrode spacing <1; and
the Q of the circuit with the specimen in the holder,
Qi, are recorded. The specimen is then removed
from the holder and the electrode spacing is de
creased until resonance is achieved while holding
C, fixed at C',. The electrode spacing U and the Q
of the circuit with the specimen out, Q, are
recorded.
The two resonant situations, with the specimen
in and with the specimen out, occur at the sams
frequency and the total capacitance in the resonant
circuits must be the same in both cases. The value
of the total capacitance, CT, is taken as the value of
C, when the Q-meter is tuned to resonance with the
holder removed.

433 -167

For the (9-meter measurement, the circuits of

figure 2 may be viewed in parallel with the C^-meter
capacitance C,. Considering figure 2a, one finds
that if CMt and RMt are placed m parallel with C,
then the transformation of figure 3b can be used
to reduce the circuit to a simple Ij-O-R series circuit,
for which the total series capacitance, the total series
resistance and the Q of the resonant circuit with the
specimen in the holder, CT, Rsi, and Qt, are given as
CT=CMt+C'

(14)

R=RMi(^j>

(15)

\/Q,= uRCT

(16)

and
where 0Mt and RMt are defined by eqs (1) and (2).
Likewise, with the specimen out and with the
holder at spacing t2, the resonant circuit can be
represented by figure 2b in parallel with C,. The
total capacitance, the total series resistance and the
Q of the resonant circuit with the spacimen out of
the holder, CT, Rn, and Q, are given as
Ct=Cm.+C'

(17)

RM=RMo(j^j

(18)

l/Qo=o>R,0CT,

(19)

and
where (\f0 and R are given by eqs (3) and (4).
Equating the expression for CT in eqs (14) and (17)
will yield eq (11) of the previous section for the
measured value of e'.
An expression for t" is obtained from the difference,
QT 1 - ft" 1 , "sing eqs ( 1 6) and (19). Thus,
<^.-'---'--rf&{^J(?#v

'"

tcAT./t. A {q) (cv-r;) t1

'

-%r+k]

In the above equation CMo has been replaced bv

CTC, in accordance with eq (17). The capaci
tances, CT and
are obtained from the calibrated
Q-meter capacitor. The capacitance CHo is obtained
from the audio frequency calibration. Omitted
from the above equations are the high-frequency
corrections which must be applied to the values of
Q and C, as read from the 190A (7-meter. The
manner in which these corrections are carried out is
explicitly stated in the 190A (^-meter manual.
A more detailed description of ^-meter dielectric
measurements may be obtained from a paper by
Hazen [4] in which he shows how the (7-meter circuit
may be adapted for the measurement of very low
losses.
4.2. Two-Terminal Measurements at Audio
Frequencies
The General Radio (GR) 716C Schering biidge
was used for audio frequencies from lO2 Hz to 105
Hz, and the GR 1615A transformer ratio arm bridge
was used from 5X101 Hz to 2X104 Hz.
For audio frequencies the effects of the lead in
ductance can be ignored, and the circuits of figure 2
without the inductance are considered here. With
these bridges one measures the dissipation factor
and the effective series capacitance of the unknown.
However, when e"/'<0.1 the series capacitance
and the parallel capacitance of the unknown are
equal (within 1 percent) and this case is treated
here. For t"/t'^>0.\ the GR bridge manuals should
be consulted.
If the holder and specimen are connected as a
two-terminal device in an arm of the bridge and if
the procedure of measurement which was used with
the ft-meter is also used with the bridge, then the
result for t is the same as eq (11) and

(20)
t" = ({Dt-D0)(^^}>

Rearranging, dividing through by (\, and using

Chi=Ch one obtains,

In the above equation, the term in brackets is the

dominant term and it alone would be present if there
were no airgap or lead inductance. The term
{Cho/Cmo)* results from the effect of lead inductance,
which becomes significant above 30 MHz (see eq
(3)). The third term, [<,CZ+ C,) jCt]2, is the airgap
correction.
Equation (21) can be reduced to following working
equation :

(23)

where I), and D are the dissipation factors with the

specimen in and out of the holder, and CT is the
total capacitance in the arm of the bridge which
contains the holder. Just as with the V-ineter
measurement, (\ must be the same for the D, and
D0 measurements.
A variation on the above experimental procedure
which utilizes the standard capacitor in the bride*"
is preferable when using a bridge circuit and null
detector for a two-terminal measurement. First,
a bridge balance is obtained with the specimen and
holder in the bridge circuit. Then t,, D(, and the
standard bridge capacitor reading (\t are recorded

434 168

The sample is taken out of the holder. Keeping

the spacing fixed at t,, the capacitance of the
standard capacitor, C is varied to rebalance the
bridge. The specimen-out values, C,t and D0 are
recorded.
If a bridge measurement is made by changing
the standard capacitor while holding the electrode
spacing at t,, then a distinction must be made
between using the substitution method of measure
ment and using the bridge as a direct reading
instrument. In the substitution method, the un
known is connected in parallel with the standard
capacitor, whereas for the direct method the un
known is connected in the bridge arm electrically
opposite the standard capacitor. The GR 1615A
is most conveniently used to measure C and D
directly. The substitution method is used with the
GR 716C bridge by connecting a precision air
ballast capacitor to the "unknown direct" terminals.
When the electrode spacing is fi for both "in" and
"out" measurements and a direct measurement is
made, the total capacitance in the unknown bridge
arm will be different for the "in" and "out" balances.
Let CTt and CVo be the total capacitance in the
unknown bridge arm with the sample in and out of
the holder respectively. For a direct measurement
we have,
CBt=CMl+C.x+C'+^,

(24)

CM.=CMl+Ctl+C,

(25)

rom which ' can be calculated in terms of measurable

juantities. Using CHlCHo= C,t Cn we have for a
lirect measurement
'=ds(r'<-r--+,-f-''')

x[,_^_^.+g-'

(26)

nd
-*AJt\Ut

CTtU-)

and

(29)
where the bracketed term in eqs (26) to (29) is the
airgap correction term. To obtain the total capac
itance, CT, in the unknown arm of the GR 716C
Schering bridge a separate bridge measurement is
necessary. During the "in" and "out" measurement
the ballast capacitance, CB, remains fixed at CT.
After the specimen holder has been disconnected,
C'B is measured by switching the bridge to "direct"
and balancing the bridge by changing the standard
capacitor. The dial reading of the standard capac
itor is calibrated by measuring a known CB. The
calibration should also include the 1 pF capacitance
of the bridge terminals.
4.3. Measurements at Ultra-Low Frequencies
The Scheiber ultra-low-frequency bridge was used
in the frequency range 10"2Hz to 2X102Hz [1].
Although the bridge was designed for use with a
three-terminal dielectric cell, we have found that it
may be easily adapted for measurements with a
two-terminal cell. Ordinarily the leads from the
bridge to the electrodes are shielded at ground
potential and the shield encompasses the entire cell.
For the two-terminal holder the electrode leads are
shielded but the shield is abandoned at the base of
the holder where the electrodes connect. A shield
around the entire cell was improvised by lowering an
aluminum box over the holder and connecting the
the box to ground.
The substitution method of measurement with the
electrode spacing fixed at <, for the "in" and "out"
bridge balance is used. Equations (22) and (2'.i) of
Scheiber's paper [1] in conjunction with the airgap
correction term of this paper were used to calculate
the capacitance and conductance of the unknown.
For measurements below 10 Hz the surface con
ductivity of the sample can become a problem. It
can be eliminated by wiping and cleaning the edges
of the sample and thereafter handling the sample
with tweezers and by introducing dry nitrogen
through one of the ports of the cell.
4.4. Self-Resonant Measurements

For a direct measurement with the GR 1615A

unsformer bridge it is conveniently true that
T C, and CTi=C,t.
On the other hand, for the substitution measureent, Ct=Ct, = Ct and C<C. For t' and t"
e have,

^-nr-^+ST'

(28)

The two-terminal holder can be used up to 600

MHz by shorting the terminals with an appropriate
length conductor and exciting the resulting circuit
at its self-resonant frequency. The Boonton model
280A UHF (^-meter was used as a combined oscil
lator and detector. The oscillator was coupled to
the holder through a variable attenuator by means
of a single-loop inductive probe. The signal was
picked up with a capactive probe and rectified to
give a d-c voltage which was applied to a voltemeter
in the model 2X0A. The voltmeter is calibrated in

435 169

full and half-power levels which can be conveniently

checked by switching 3 (IB into the inductive probe.
The oscillator frequency was measured with a
Hewlett Packard model 524C frequency meter and
540A transfer oscillator.
Measurements were made by placing the specimen
in the holder and tuning the circuit by adjusting the
oscillator frequency. The frequency corresponding
to the half-power points is then determined and
recorded along with the electrode spacing. The
specimen is then removed and the circuit retuned
by adjusting the holder electrode spacing. The
frequencies corresponding to the half-power points
are again measured and recorded. One or more
additional measurements of electrode spacing and
corresponding resonance frequency must be made in
order to determine the change in capacitance with
resonance frequency.
Calculations are based on eqs (1) through (6),
where the terminals in figures 2a and 2b are shorted.
Thus we treat the problem as a simple series
LTCTRT circuit for which the dissipation factor DT
and total capacitance CT are

Finally
dt dCc A/,-A/
~djM ' dt ' 2toA.it.
u-t,
X 1

rv,-C, , r/-i
-.37;

5. Sample Data

The ratio of the frequency separation of the

half-power points to the resonance frequency is
measured to obtain the dissipation factor. The
total capacitance which cannot be measured directly
can be obtained by differentiating eq (31) to obtain

In order to illustrate the results which can be

obtained with the dielectric specimen holder de
scribed in this paper, room temperature measure
ments were made on two specimens machined from
the same 2.54-cm-diam poly(methy] methacrylate)
rod into disks of average thicknesses 0.3150 cm and
0.1577 cm. The disk faces were made flat and
parallel to better than 10 y. and no contact electrodes
were applied. Botli specimens were kept at about
23 C at a relative humidity of 50 percent for
several months before being measured. The measur
ing techniques used are those described in detail in
the previous sections, and the measurements on the
two specimens over t he whole frequency range took
about 10 hr. Calculations of t' and t" were done
by hand. The results of the measurements are
shown in figure 4 which is a plot of the values of
e' and t" versus log frequency for the 0.3150-cmthick specimen of poly (methyl methacrylate).
The data for the two specimens measured were
found to agree to within 1 percent for t and 5
percent for e" at all frequencies, and with just a few
exceptions, the agreement was within 0.3 percent
for e' and 2 percent for t" over the frequency range
covered. In addition, the highest frequency values,
which are the most crucial as far as the magnitude of
the circuit correction is concerned, agree with each
other and with measurements made using a re
entrant cavity [3] within 0.3 percent in t and 2
percent in t". Additional measurements mide on
specimens of polvchlorotrirluoroethvlene and fused
silica gave results which were also accurate to
within the limits claimed in this paper except for the
loss in fused silica which could not be detected at
high frequency.

dCT
2
= -fC
dft
Jm

6. Discussion of Measurement Errors

Dt = uCtRt>

(30)

1
(31)
One may obtain e' from eq (11) and t" from eq (13)
remembering that here CHi=Cr and
ADM=DMt-D0=uCT (RMt-RUt>).

(32)

Substituting (32) into eq (13) gives

(34)

Hence we can write

,, = _dCr
AA-A/o/C + g,V
dCT A/,-A//C;+C,V
dju
2C. \ C, J

(35)

Since the only part of CT which changes with spacing,

t, is the calibration capacitance Cc of the holder one
can write
dt_ dCe Aft-Afo/Cx+C,
m (gy 'j
dju' dt ' 2C,

(36)

The two-terminal method described in this paper

was intended to provide values of e' and t" over the
entire range of applicable frequencies without im
posing rigid requirements on specimen preparation
and cell construction. In addition to uncertainties
in the quantities explicitly appearing in the working
equations of the previous section, attention must be
given to the assumptions and approximations upon
which these equations are based, the errors due to
which will appear in an error capacitance term C.
The error capacitance (in fig. 2 and eq (5)) in
cludes corrections to the ideal assumed capacitance

436-170

1 0 24

LOG FREQUENCY
Figure 4. The dielectric constant (e') and lass index (") of a 0.3 1 50-cm-thick disk of com
mercial poly (methyl methacrylate) as a function of frequency as measured with the twoterminal dielectric specimen holder using various measuring equipment.
resulting from distortions of the electric field be
tween the measuring electrodes because of (a) tilted,
misalined or nonflat electrode surfaces, (b) the
presence of an airgap, (c) departure of the specimen
geometry from an ideal right circular cylinder,
(d) electrode edge effects, and (e) the presence of
air between the electrodes.
(a) The errors due to nonideal electrodes are all
second order effects with the actual capacitance
given by

where A and t are the electrode area and separation

and A represents the variation in t in the case of
tilted or uneven electrodes or the distance between
ixes of the electrodes in the case of misalined elec
trodes. It is not difficult to keep mechanical errors
to within 25 p. (25m=0.001 in.). With typical sepaations, t, of 1000 p. these errors amount to roughly
<0.1 percent. In addition, the tilt error increases
he capacitance whereas the misalinement error
iecreases it. Also since each determination of C
n volves a difference between two measurements, the
sfFects tend to cancel. Thus the mechanical conitruction need not require special attention.
(b) The airgap causes some distortion of the field
it the edge of the specimen (which increases with
;ap width and dielectric constant) and makes the
.ctual capacitance higher than its uncorrected value,
"his in turn leads to a high value for the specimen
apacitance Cr. An upper limit to this error (e' = )
or a typical measurement was calculated to be 0.2
ercent [6], and one can reasonably conclude that in
lost cases the error from this source is <0.1 percent.
(c) Geometric imperfections in the specimen also
sad to distortions. Rounded corners would cause
ii error similar to the airgap error. Uneven or nonarallel specimen surfaces lead to an actual capaciince greater than that assumed. The airgap capaciince is particularly affected by nonideal specimen
lrfaces. The specimen surface may well be tilted

with respect to the adjacent electrode surface so that

the airgap is a wedge of average thickness t=15 p.
and variation A=10 p. In this case the assumed
airgap capacitance may be in error by 20 percent.
(This calculation is based on the worst casethe
capacitance of two concentric but mutually tilted
circular disks in the limit of minimum separation.)
Fortunately, the airgap capacitance enters the calcu
lations in a very insensitive way so that an error of
20 percent in Ct would result in an error of perhaps
0. 5 percent in Cx, and the error would be such as to
increase the value of Cx above its true value. The
geometry of the specimen does not appear to be
particularly critical and special molding and machin
ing techniques are not required.
(d) Electrode edge effects have been studied
extensively [5]. With the specimen in place the
edge capacitance is higher than one assumes and
hence the calculated value of Cx is too high. In
general, this error increases with t' and also with the
proximity of the specimen to the edge of the electrode.
If the distance between the specimen and the elec
trode edges is twice the separation of the electrodes
then this error is <0.1 percent [5]. Thus it is ad
visable to design the holder accordingly keeping in
mind that measurement sensitivity decreases as the
ratio of electrode area to specimen area increases,
1.e., as Cz becomes a smaller part of the total
capacitance.
(e) Since e' for air is about 0.06 percent higher
than assumed, then the actual capacitance with
specimen removed is higher than calculated, thus
lowering e' by <0.1 percent. The effect on the
airgap is negligible.
A final approximation of the accuracy of the meas
urements can be obtained from an estimate of the
accuracy of the measured quantities in the working
equations. Generally, the calibration capacitances
can be measured to the nearest 0.1 percent, the
linear dimensions to the nearest micron and the area
of the sample using a traveling microscope to 0.02
percent. Considering eq (11) for e' the factors and

437-171

their percent uncertainty are C 0.3 percent; [(fe,

C^ + teAJt, C], 0.3 percent; the airgap correc
tion term, 0.5 percent. In eqs (22), (23), (27), and
(29) for e", the factors and their percent uncertainty
are CT, 0.2 percent; toA,/t 0.3 percent; A
3 percent; C/(CTC,), 0.3 percent; (D,-D),
(0.5% +0.00005) for GR 1615A; (D,-Dc), (2%
+0.0005) for GR 716C; and the airgap correct
ion 1 percent. For eq (37) the percent uncer
tainty of the terms are dt/dfu, 0.5 percent; dCJdl,
1 percent; and A/, and A/0, (0.2%+ 10kHz).
The total percent uncertainty for t' and t" is
obtained from the square root of the sum of the
squares of the uncertainties in the working equations.
For t the percent uncertainty is 0.7 percent over
the entire frequency range 10"2 to 6X108 H/>. For
t" we have (L5% + 0.0005) for the ultra low
frequency bridge, (1.8% + 0.0005) for the GR
716C bridge, (0.6% + 0.0001) for the GR 1615A
bridge, (4% + 0.0005) for the 260A and 190A
^-meters, and (1.6% + 0.0005) for the self-resonant
measurements.

7. References
[1] D. J. Scheiber, J. Res. NB8 6SC (Eng. and Instr.), No. 1.
23(1961).
]2] A. J. Bur, 1964 Annual Report of the Conference" on
Electrical Insulation, National Academy of SciencesNational Research Council Publication 1238, p. 70.
[3] A. H. Scott, D. J. Scheiber, A. J. Curtis, J. I. Lauritzen,
Jr., and J. D. Hoffman, J. Res. NBS 66A (Phvs. and
Chem.), No. 4, 269(1962).
[4) T. Hazen, Presented to the 13th Annual Wire and Cable
Symposium, Atlantic City, New Jersey, December
2-4, 1964.
[5] A. H. Scott and H. L. Curtis, J. Res NBS 22, 747(1939)
RP 1217.
[6] J. I. Lauritzen, Jr., unpublished results.

438-172

(Paper 69C3-195)

UNITED STATES DEPARTMENT OF COMMERCE

Luther H. Hodges, Secretary

NATIONAL BUREAU OF STANDARDS A. V. Astin, Director

Basic Magnetic Quantities and

the Measurement of the Magnetic
Properties of Materials

Raymond L. Sanford and Irvin L. Cooter

National Bureau of Standards Monograph 47

Issued May 21, 1962

315-411 O - 60 - 29

439-1

Preface
Since 1909 when the first NBS Circular on Magnetic Testing was issued,
successive revisions have been prepared in order to keep up to date with the
development of new magnetic materials and methods of testing. Since the
last circular, C 456, Magnetic Testing, was issued in 1946 the importance of
magnetic materials and testing methods has greatly increased. In view of the
many requests for information regarding magnetic quantities, materials, and
testing methods, the present revision and extension of theprevious circular has
been prepared. It supersedes Circular C 456, Magnetic Testing.

Preface

Page
n

1. Introduction

2. Magnetic quantities and units

2.1. Systems of unitsdimensions
2.2. Basic quantities
a. Magnetic induction, B
b. Magnetizing force, H
2.3. Magnetic constant, r
2.4. Derived quantities
a. Magnetic flux, <t>
b. Magnetomotive force, PL....
c. Magnetic reluctance, St.

i
1
2
2
2
2
3
3
3
4

3. Magnetic characteristics of materials

3.1. Magnetic permeability
-3.2. Magnetic susceptibility
3.3. Classification of materials
-3.4. Magnetic hysteresis and normal in
duction
3.5. Core loss

4
4
5
5

4. Measurement of static magnetic fields

4.1. General principles
4.2. Pivoted coil...
4.3. Pivoted magnet
4.4. Ballistic methods
a. Ballistic galvanometer
b. Intercomparison of mutual inductors
and calibration of test coils
4.5. Rotating test coil
4.6. Bismuth spiral
4.7. Hall effect
4.8. Saturable core
4.9. Nuclear magnetic resonance

7
7
~
8
8
8

6
7

10
11
11
12
12
I -'

5. Tests with direct current

5.1. Testing of materials
5.2. Ring method
5.3. Straight bar and solenoid
-5.4. Permeameters
a. MH permeameter
b. High-H permeameter
c. Fahy Simplex permeameter
5.5. Tests of low permeability materials
5.6. Magnetic susceptibility
5.7. Magnetic standards
5.8. Limits of accuracy
5.9. Requirements of standard specimens for
d-c permeameters
5.10. General precautions
6. Tests with alternating currents
6.1. Core loss and a-c permeability at power
frequencies
a. Voltmeter-wattmeter methods
b. Bridge methods
6.2. A-C Measurements at higher than power
frequencies
7. Typical magnetic properties of materials
7.1. Solid core materials
7.2. Electrical sheet and strip
7.3. Special-purpose materials.7.4. Ferrites
--7.5. Permanent-magnet materials
a. Quench-hardened alloys
b. Precipitation-hardening alloys
c. Work-hardened alloys
d. Ceramic magnet materials
e. Powder magnets
7.6. Feebly magnetic materials
8. References-

440-11

Pare
12
12
14
15
17
17
18
19
21
21
23
23
24
24
24
24
25
27
29
30
30
30
31
32
33
33
33
34
35
35
35
35

Basic Magnetic Quantities and the Measurement of the Magnetic

Properties of Materials
R. L. Sanford and I. L. Cooter
This paper gives general information regarding the two basic quantities, magnetic
induction, B, and magnetizing force, H, and also the magnetic constant r (often designated
by the symbols p, and nn). Information is also given regarding the magnetic properties
of various materials and methods and apparatus commonly used in the Magnetic Measure
ments Section for measuring these properties by means of reversed direct current or alter
nating currents of low frequency. Magnetic measurements peculiar to high frequencies are
not discussed. In view of the gradual adoption of the rationalized mksa system of units,
this system is included as well as the classical cgs electromagnetic system.

1. Introduction
The work of the Magnetic Measurements Sec
tion of the National Bureau of Standards includes
(1) testing of specially prepared test specimens in
tended for use as standards for checking magnetic
testing apparatus (2) investigation and develop
ment of magnetic testing apparatus (3) calibration
of mutual inductors, test coils, and instruments for
measuring magnetic fields, and (4) investigations
in the field of magnetics such as studies of phenom

ena associated with nuclear magnetic resonance or

the selection and testing of materials suitable to be
used as standards of magnetic susceptibility.
This monograph gives general information
regarding magnetic quantities and units, the mag
netic characteristics of various materials, and
methods and apparatus commonly used in the
Magnetic Measurements Section for magnetic
testing.

2. Magnetic Quantities and Units

2.1. Systems of UnitsDimensions
For many years magnetic quantities have been
expressed in the cgs electromagnetic 1 system of
units. In this system the centimeter, gram, and
second are taken to be the basic units in which the
concepts length, mass, and time respectively are
expressed. A three-dimensional system lacks the
"resolving power" necessary to distinguish be
tween the magnetic quantities denoted by the
s3Tmbols B and H. The ratio between these two
quantities in empty space,
or ju0, is arbitrarily
assigned the value unity in this system and there
fore n, or m> is usually omitted from the equations
without affecting the numerical values.
Another system based on the meter, kilogram,
and second as the units of length, mass, and time
respectively has been adopted by the International
Electrotechnical Commission (IEC) and is rapidly
gaining favor. It seems likely that this system
eventually will replace the cgs system. In this
system the ratio of B to H in empty space is not
unity and consequently it cannot be ignored. A
fourth dimension in addition to length, jnass, and
time is required to characterize magnetic and eleei Tliere are two other cgs systems, namely the electrostatic system and the
rausslan system, but In this monograph cgs means cgs electromagnetic
in less otherwise specified.

trical quantities in this system. It must be elec

tric or magnetic in nature. The IEC has chosen
electric current, /, as the fourth dimension.
When the four dimensions LMT1 are applied to
magnetic quantities the distinction between B and
H is clearly evident.
At the same time that the IEC adopted the
mksa (Giorgi) system of units a further step was
taken, namely, "rationalization." The object of
rationalization is to transfer the factor 4r from
linear equations to those having circular sym
metry. In the rationalized system, the ratio of
B to H in empty space,2 Tm is 4t X 10~7 h/m.
By applying four dimensions to quantities in the
cgs electromagnetic system (which is not inherently
three-dimensional, as is ordinarily assumed) addi
tional "resolving power" can be gained, the dis
tinction between B and // then becomes clearly
evident and mutual consistency between the two
systems of units is brought about. The symbols
L, M, T, and / denoting length, mass, time, and
current respectively may be applied to either
system.3
' Since this is a true definitional constant, whereas the symbol w denotes
relationships usually not constant, the symbol r Is used instead of f, or
p. to avoid confusion.
' Many writers consider Q (quantity of electricity) to be a more simple
concept than current and use It as the fourth dimension. The dimensions
then are LMTQ,

441-1

2.2. Basic Quantities

Symbol
There are two basic magnetic quantities and
a constant from which all other magnetic quantities are derived. They are magnetic induction,
B (often called flux density), magnetizing force
or magnetizing field, H (also called magnetic
field strength or magnetic intensity), and the
magnetic constant rm.

B
K
t
e.
a

i (Olorgi)
gauss
dyne
abampere
centimeter
volt
square centimeter,
second

tesla
newton
ampere
meter
volt
square meter
second

a. Magnetic Induction, B
The dimension 4 of magnetic induction is
A current-carrying conductor in a magnetic
field experiences a mechanical force, the magnitude
of which depends upon the magnitude of the cur
rent and the field, the length of the conductor
and its orientation in the field. This dependence
on direction and magnitude identifies the field as
a vector field. Its direction is considered to be
the direction of the conductor for which the me
chanical force is zero. The quantity measured by
the mechanical force experienced by a currentcarrying conductor in a magnetic field is called
magnetic induction, B.
If a linear conductor is perpendicular to the
direction of the field and the field is uniform along
its length
b=f/ii
where

B= magnetic induction
F= mechanical force
7 = current

and

[B]=MT-2I-K
b. Magnetizing Force, H
In view of the general acceptance of the concept
that the magnetic behavior of materials is due to
the presence within them of electrons in motion
(spins or rotations) it is reasonable to conclude
that magnetism is simply one of the manifestations
of an electric current. Magnetic induction at a
given point must therefore be due to the influence
of electric currents. The magnetizing influence of
an electric current is proportional to its magnitude
and depends upon its geometrical configuration.
The measure of the ability of an electric current to
produce magnetic induction at a given point is
called the magnetizing force (magnetic field
strength or magnetic intensity), H.
At the middle of a very long uniformly wound
solenoid
H=KNI/l

1= length of the conductor.

where
There is another phenomenon by which the pres
ence of a magnetic field can be detected and its
magnitude determined. If a conducting loop or
coil is placed in a magnetic field and the strength
of the field is varied, the orientation of the coil
in the field is changed or the coil is removed from
the field, an electromotive force will be induced
in the coil during the change. At any instant
while a change is taking place

H= magnetizing force
K=& constant depending on the system
of units
7V= number of turns
1= current
/= axial length of the solenoid.

The units for the two systems are as follows:

Symbol

mksa (Giorgi)

cgs

e=Na dB/dt
where

H
K
N.
I
I

e = instantaneous induced emf

2V=number of turns in the coil
a= average area of the turns
B= magnetic induction
<=time.

oersted . .
4ir
turn
abampere
centimeter

Ampere-turn per meter

1
turn
ampere
meter

The dimension of magnetizing force is

Provided that B is uniform over the area a, the
total change in induction, AB, in a given time is

[fl]=L->7.
2.3. Magnetic Constant, rm

f'

The units for the cgs and mksa (Giorgi) systems

are as follows:

The magnetic constant for any system of units

is the proportionality factor in the expression re
lating the mechanical force between two currents

442-2

* The square brackets in a dimensional equation denote "the dimension oL"

to their intensities and geometrical configurations.

In differential form this may be written

face, the product of the induction, B, by the area,

A, is the magnetic flux.
4>=BA

dF=Tm I, /jefl, X (eflj X O/nH

where
where

<f>= magnetic flux

B=magnetic induction uniformly distributed and
normal to the plane of the surface of which

rm

is the magnetic constant

dF

is the element of force of a current element

A= area.

Iidli

on another current element

IidU

at a distance r.

If the induction is not uniformly distributed over

the area, the surface integral of the normal com
ponent of B over the area is the magnetic flux

ri

is a unit vector in the direction from lx to Z3

is a dimensionless factor which is unity in

unrationalized systems and 4jt in rational
ized systems.

BdA
-JJ:
where
<#>= magnetic flux
B=Magnetic induction
and

The magnetic constant is also equal to the ratio of

the magnetic induction, B, to the corresponding
magnetizing force, H, in a vacuum. For this
reason it is often called the permeability of space
(or vacuum) and denoted by the symbol n, or ^
Since the symbol n, with or without subscripts is
used to denote several different relationships be
tween B and H which in general are not constant,
the symbol r is used to denote the magnetic con
stant. The value of Tm determines the system of
units employed. In the classical cgs system the
value of r is unity and it is a numeric. In the
rationalized mksa (Giorgi) system its value is
47rX10-7 and the dimension is LMT~2I~2 (henries
per meter).

<L4=an element of area.

It is important to note that although both B and
A are vectors, 0 is a scalar and cannot have
direction.
The units in the cgs and rationalized mksa
systems are
cgs

Symbol
maxwell
gauss
cm2

0
B
A

mksa
weber
- - tesla

The dimension of magnetic flux is

2.4. Derived Quantities
The principal derived quantities are associated
with the idea of a magnetic circuit, somewhat, but
not exactly, analogous to an electric circuit. It is
a magnetic structure which may contain one or
more airgaps or other "nonmagnetic" materials
and designed to contain certain continuous lines
>f magnetic induction. Ideally, none of the
induction would escape from the circuit. How
ever, since there is no insulator for magnetism,
:his ideal condition is rarely if ever realized. The
nagnetic flux which escapes from a magnetic cir
cuit is called leakageflux or simply magnetic leakage.

b. Magnetomotive Force,
In a magnetic circuit, the line integral of the
magnetizing force around the circuit is called the
magnetomotive force. It is proportional to the
total ampere-turns linked with the circuit.
F=KNI
where
J^=magnetomotive force
N number of turns

a. Magnetic Flux, <t>

/= current

Given a plane surface or cross section within the

toundaries of which there is a uniformly disributed field of magnetic induction, B everywhere
aving a direction normal to the plane of the sur

and
K=& constant depending on the system of units.
The units for the two systems are

443-3

Table 1.

Magnetic quantities and units

cgs Unit

Quantity and
dimension Symbol
(mksa)
Name
Induction

gauss

The dimension of magnetomotive force is [^] = I.

c. Magnetic Reluctance, Pf
The magnetic flux resulting from a given
magnetomotive force acting on a magnetic circuit
is determined by the magnetic reluctance Pi, of
the circuit. Thus

0 = flux
SF= magnetomotive force.
and
M= magnetic reluctance.
The units are

.T9t-

--

mksa
maxwellgilbert. .
()

Magnetic
Constant
LMI^'I'

weber
ampere-turn

oersted

r.

Flux
UMT-'l-i
Magnetomotive
Force
/
Reluctance PA
L-WT>1>
Normal per M
meability
LMT-'I'
Relative Per
meability
(dimen
sionless)
Susceptibility t
(dimensionless)

where

Symbol

Magnetizing H
foroo
-'/

To conten
Name of cgsemnlc
rationalized to rationi;mksa unit izedmks
Equation
muliiplj-'bv

teste
10-*
(weber/m1)
(/in
abamperes)
ampereturn per
meter
(/in
amperes)
" //
(in vacuum)

-12.57XIO-'

maxwell -JL*4 weber

io-<

arilpere
turn

gilbert

lC/4r
-7SJ8XIC

&~
B
"-H
M

4iXHH
-IISTX'.O-'
C)

In this table, "cgs unit" refers to the unrationalized cgs

system.

The dimension of reluctance is [R] = L~2M~'T2P.

3. Magnetic Characteristics of Materials

3.1. Magnetic Permeability
Magnetic permeability, denoted by a symbol
u, with or without certain subscripts, is a term
used to express various relationships between
magnetic induction, B, and magnetizing force, H,
in a material under specified conditions. The
simplest relationship is the ratio of induction to
the corresponding magnetizing force.
u=B/H.
This is usually called absolute permeability or
simply the permeability. The term has a specific
significance, however, only under certain definite
conditions. For certain types of material the
ratio is constant, not depending upon the degree
of magnetization. For other materials the ratio
depends upon the induction. If the corresponding
values of B and H are determined by a standard
ized normal procedure (to be described later) the
1 Previous to 1930 the cgs unit of reluctance was called the oersted, but
the International Electrotechnical Commission, In 1930 adopted the name
oersted for the cgs unit of magnetizing force, leaving the unit of reluctance
without a name.

ratio of B to H is called the normal permeability.

This is generally denoted by the symbol u, with
out subscripts. It relates only to points on the
normal induction curve. The symbol u with
various subscripts denotes either the normal
permeability at specified values of B or H or
the ratio of certain changes in B to the corre
sponding changes in H. In this discussion, the
term permeability (symbol u) denotes normal
permeability.
Permeability has the dimension
[U]=LMT-2I~2
in the rationalized mksa system.
The symbol uT denotes the quantity called
relative permeability. It is the ratio of the perme
ability of a material to that of space (or vacuum1.
Or it may be defined as the ratio of the pennrability of a material to the magnetic constat;;.
Mr=M/T
Since this is the ratio of two quantities harx:
the same dimensions, ur is seen to be a dimenscr

444-4

less ratio and its value is independent of the

system of units. The absolute permeability is
the relative permeability multiplied by the value
of r pertinent to the system of units employed.
In the cgs system, the value of rm is unity so that
the numerical value of the absolute permeability
in this system is the same as that of the relative
permeability. It is for this reason that in the
cgs system the normal permeability is considered
also to be a dimensionless ratio which leads to
the conclusion that B and H are quantities of
the same kind.
3.2. Magnetic Susceptibility
The magnetic induction, B, in a material due
to a magnetizing force, H, is made up of two
components, that induced in the space alone,
TmH, and that due to the magnetization e of the
specimen, Bt. That is
B=TJS+B
The ratio of these two components is called the
magnetic susceptibility, k
k=B,/rnH
since
B,=B-TmH
k=B(/TmH= {BjTJO) - 1

-1.

This differs from susceptibility in the classical

cgsem system by the factor 4a- because in the
cgsem system in which rm=l, the equations are
customarily written
B=H+4*J
J=(fi-fl)/4
k= JjH= (B-H)/4ttH= (m- 1)/4t.
J is called intensity of magnetization and has
the dimensions of magnetic induction. The
ratio of J to H, therefore, has the dimensions of
permeability although susceptibility is actually a
iimensionless constant. However, in the cgsem
system only three dimensions are employed, B
ind H are considered to be quantities of the same
cind, and n is a dimensionless ratio. Inthemksa
ystem as has been previously noted rm is not
mity and therefore in vacuum B is not equal to
Jy their ratio is not unity and so neither the
alues nor the dimension of rm can be ignored as
hey are in the classical cgsem system.
Susceptibility is used to characterize materials
-hose permeability differs from unity by only a
ery small amount. The susceptibility of a
laterial divided by its density, p, is called its
lass susceptibility, X.
The component due to the magnetization of the specimen Is called tho
trinsie Induction, Pi

X=k/p
The mass susceptibility multiplied by the atomic
weight A is called atomic susceptibility, XA.
XA=xA
3.3. Classification of Materials [32]
The magnetic properties of materials result from
the motions of electric charges within them. The
motions are either rotation in orbits or spins about
the axis of the charges. These motions constitute
equivalent electric currents and therefore produce
magnetic fields. The field-producing effect of an
orbit or spin is called its magnetic moment. It is
a vector whose direction is that of the axis of spin
or of the normal to the plane of the orbit and whose
magnitude is the product of the equivalent current
and the area enclosed by its path. The magnetic
moment of a magnetized body is the vector sum
of its internal moments. Dimensionally, \m\=Ul
where m is the magnetic moment.
In the past, materials have been classified as
diamagnetic, paramagnetic, or ferromagnetic.
However, two other classes are now recognized,
namely, antiferromagnetic and ferrimagnetic. The
magnetic properties of a single atom depend upon
the relative locations and directions of spins of its
electrons. The spin of the nucleus and orbital
motion of the electrons constitute such small parts
of the magnetic moment that they can generally
be neglected. The moments of two electrons in
the same orbit or shell whose spins are in opposite
directions mutually cancel. An atom has a per
manent magnetic moment if there is an excess of
spins in one direction.
The magnetic properties of materials composed
of associated atoms depend upon the arrangement
of the atoms with respect to each other. This
determines the kind of interactions between their
magnetic moments. The forces of interaction
(also called exchange forces) are functions of the
ratio of the distance Detween atoms to the diameter
of the orbit or shell in which the uncompensated
spins are located.
Diamagnetic materials have no permanent mag
netic moment. However, moments are induced
by the influence of a magnetic field. The induced
moments have a direction opposite to that of the
inducing field. Consequently the relative per
meability is less than unity. The diamagnetic
susceptibility is negative. The diamagnetic sus
ceptibility is very small, not more than a few parts
in a million, and usually independent of temper
ature and applied magnetic field. It is probably
present in all materials but usually masked by
other larger effects.
Paramagnetic materials have a permanent mag
netic moment but the interatomic spacing is so
great that there is negligible atomic interaction.
The relative permeability is greater than unity so
the susceptibility is positive. Although the effect
is greater than the diamagnetic effect, it is still

445-5

very small, not more than a few parts in a hundred

thousand. There are two kinds of paramagnetism
called strong and weak. Except for high fields of
the order of 10* gausses (1 tesla) or nigher the
strong paramagnetic susceptibility is constant
with field at a given temperature but varies in
versely with temperature at a given field. Weak
paramagnetism is due to the effect of conduction
electrons in conducting metals and is practically
independent of temperature.
Ferromagnetic materials have a relative per
meability greater than unity and generally very
high. The permeability is not constant but de
pends upon the degree of magnetization. Fer
romagnetic materials exhibit hysteresis, that is,
the induction corresponding to a given magnet
izing force depends upon previous magnetic history.
Furthermore, the intrinsic induction approaches a
limiting or saturation value as the magnetizing
force is increased indefinitely. Ferromagnetism
has sometimes been called a special case of para
magnetism but in view of the differences in the
processes involved, it should not be so classified.
Ferromagnetic materials are temperaturesensitive and when the material is heated, the
temperature at which it is transformed from the
ferromagnetic condition to the paramagnetic
condition is called the Curie temperature.
In ferromagnetic materials, the interatomic
distances compared to the diameter of the orbit
or shell in which the uncompensated spins are
located are such that neighboring atoms have
their magnetic moments alined parallel and in
the same direction. This is called positive
interaction. Not all of the moments in a body are
oriented in the same direction. Instead, they are
lined up in groups called domains which act as
magnetic entities. Each domain is spontaneously
magnetized to saturation but because they are
oriented in various directions the resultant
magnetization in any direction may be anything
from zero to the saturation value at which all
domains would be oriented in the direction of the
applied field.
Antiferromagnetic materials are those whose
interatomic distances are less than the critical
value so that the magnetic moments of neighboring
atoms line up parallel to each other but in opposite
directions, that is, antiparallel. The suscepti
bility of these materials is so low that they might
easily be mistaken for paramagnetic materials.
The experimentally distinguishable characteristic
is that the susceptibility increases instead of
decreasing as the temperature is raised until the
thermal agitation destroys the interaction.
Above this temperature the material becomes
paramagnetic. This is analogous to the Curie
Soint in ferromagnetic materials but is called the
!eel point for this class of materials.
Ferrimagnetic materials are those in which
unequal magnetic moments are lined up antiparallel to each other leaving a net permanent
moment. Permeabilities are of the same order of

magnitude as those of ferromagnetic materials

but are lower than they would be if all the atomic
moments were parallel and in the same direction.7
Under ordinary conditions the magnetic charac
teristics of ferrimagnetic materials are quite
similar to those of ferromagnetic materials.
3.4. Magnetic Hysteresis and Normal
Induction
One of the important characteristics of
ferromagnetic materials is the phenomenon of
magnetic hysteresis. This phenomenon is illus
trated in figure 1. If a demagnetized specimen
is subjected to the influence of a magnetizing force,
H, which is increased from zero to higher and
higher values, the magnetic induction, B, also
increases but not linearly with H. This is shown
by a curve oabcd. This nonlinearity is another of
the characteristics of ferromagnetic material.
B
m
Brs,

1
/

1
/

IHn I Hc

c
.2
Indue

n //
1 1

V 1
0
/

H
/ Magnetizing Force

Figure 1. Normal induction curve and hysteresis loops.

If the increase in H is stopped at a point such as
6 and then decreased, the induction does not
retrace the original curve in reverse order but lags
behind it as indicated by the curve b, Bn He, etc.
This lag is called magnetic hysteresis. The point
where the magnetizing force is zero is called the
residual induction, BT. The negative magnetizing
force at which the induction becomes zero is called
the coercive jorce, Hc. The closed curve starting
from b through BT, He, etc. back again to b is called
a hysteresis loop. The loop does not always close
at the first reversal of the magnetizing force but
will close after enough reversals have been made.
' There Is a technically Important class of material called ferrites. The
atomic Interactions In these materials Is mainly antiferromagnetic but there
is a net magnetic moment which gives them ferromagnetic characteristics
The ferrites are ceramic bodies having extremely high electrical resistivity
which makes them particularly valuable for use at high frequencies.

446-6

If the limits of H in each direction are equal, the

limits of B in the two directions will also be equal
and the material is said to be in a symmetrically
cyclic condition. The induction at the tip of such
a loop is called the normal induction. The ratio
of B to H at this point is called the normal
permeability n=B/H.
It is easy to see that the size of a hysteresis loop
for a given material depends upon the value of
induction at its tip. The normal induction curve,
oabcd, is the locus of the tips of a family of cyclically
symmetrical hysteresis loops.
The normal induction curve usually consists of
three distinct stages. In stage I the rate of
increase of B as H is increased is comparatively
small. The steep part of the curve represents
stage II of the magnetization. In this stage, a
small increase in H produces a relatively Targe
increase in B. In stage III the rate of increase of
B is again small. In this stage the intrinsic induc
tion, B(, asymptotically approaches a limiting
value which is called the saturation induction, B,.
For this reason, stage III is sometimes called the
saturation range. Magnetic saturation is another
of the distinguishing characteristics of ferro
magnetic materials.
The magnetizing force and induction at the tip
of a hysteresis loop are denoted by the symbols
Hm and Bm, respectively. As Bm is carried
higher and bigher in the saturation range there
comes a time when further increase does not pro
duce any further increase in B, and He. The
maximum values of these quantities are called
retentimty, BT and coercivity, Hc respectively.

For permanent-magnet materials, the important

part of the hysteresis loop is that portion in the
second quadrant between the residual induction
point, B and the coercive force point, Hc. This
is called the demagnetization curve. Points on
this curve are designated by the coordinates B and
Hd. The product of Ba and H for any point on
the demagnetization curve represents the energy
external to the magnet which could be maintained
under ideal conditions. A curve obtained by
plotting the products of the corresponding coordi
nates Bd and Hd as abscissas against the induction
Bd as ordinates is called the energy-product curve.
The maximum value of the energy-product,
{BdHd)m is a good criterion of the relative quality
of permanent magnet materials.
3.5. Core Loss
When materials are subjected to alternating
magnetic fields, as in the cores of transformers, a
certain amount of energy is expended which can
not be recovered but is dissipated in the form of
heat. This loss of energy is called core loss.
Core loss is made up of two major components,
hysteresis and eddy currents.8 The hysteresis
loss depends upon the area of the hysteresis loop
and the frequency of alternations. The eddy
currents are induced in the core by the alternating
magnetic flux and depend not only upon the fre
quency and maximum induction but also upon the
electrical resistivity of the material, the thickness
of the laminations, and the insulation between
them.

4. Measurement of Static Magnetic Fields

4.1. General Principles
There are two aspects of a basic phenomenon by
which magnetic fields can be measured. They
ire (1) the mechanical force experienced by a cur-ent-carrying conductor in a magnetic field or (2)
he electromotive force induced in a conducting
oop (test coil) when the magnetic induction enircled by or linked with the loop is changing,
ither because the induction is varying or because
he loop is moving with respect to the induction,
t is important to note that the quantity measured
y either method is by definition magnetic inducion, B. Magnetizing force, H, is calculated in
rms of either electric current and its geometry
r a measured value of B divided by the magnetic
instant, T. In the cgs system, rm is unity so
lere is no numerical difference between B and
f.
Consequently, it is general practice in certain
ises to measure B and call it II. In the mksa
astern, rra is not unity so it must be taken into
;cotint in determining the value of H in a magtic field.
Tfaere are several additional methods by which
agnetic fields can be measured. These are (1)
iange in electrical resistance due to a magnetic

field, (2) the Hall effect, (3) behavior of saturable

magnetic cores in a magnetic field, (4) optical
effects, and (5) nuclear magnetic resonance.9 The
measurement of a magnetic field in terms of the
mechanical force experienced by a currentcarrying conductor is, of course, basic. However,
accurate measurements by this method, although
simple in principle, require somewhat elaborate
apparatus and very careful experimental pro
cedure. Such apparatus is not commonly avail
able. A pivoted-coil arrangement is similar in
principle and often used on account of its con
venience if high accuracy is not required.
4.2. Pivoted Coil
A pivoted (or suspended) coil is essentially an
inversion of the well-known d'Arsonval type of
instrument. In the d'Arsonval instrument the
coil is located in the constant field of a permanent
magnet and its deflection against the restoring
force of a spring or suspension is a measure of
the current in the coil. The inversion consists
' A third component may be observed at high frequencies, but at lower
frequencies It is usually negligible compared with the total loss.
The methods commonly used in the field of terrestrial magnetism are
not described In detail but only such methods as find application primarily
in connection with the testing of materials are considered.

447-7

in substituting the field to be measured for the

field of the permanent magnet. The coil, prefer
ably without the usual iron cylinder within it, is
mounted in a fixture so that it can be placed in
proper position in the field to be measured.
There are three different methods of making the
measurements. The most convenient is employed
in a commercially available instrument. By
this method, the current necessary to produce a
certain standard deflection is observed by means
of a milliammeter connected in series with the
coil and the adjusting resistors. Since the current
is inversely proportional to the field, the milliam
meter is calibrated in terms of the field strength
in gausses. It would also be possible to observe
the deflection of the coil due to a standard current
or to measure the torque required to bring the
deflection back to zero. The first mentioned
method is most convenient. The presence of an
iron core inside the moving coil is objectionable
because it distorts the field being measured.
One limitation of this type of instrument is
that for accurate results the field must be uniform
throughout the volume occupied by the coil so
that fields of small extent cannot be accurately
measured.

fields and therefore the calibration would be

affected.
Alternating fields and fields greater than 5,000
gausses can be measured by a modified form of the
instrument in which the Silmanal magnet is re
placed by a cobalt-plated surface on soft iron.
By this system, the general direction of the field
can be determined, but not its polarity.
The gaussmeter is regularly furnished with
probes either 1% in. or 5 in. long and diameter!3 of
0.052 in. or 0.090 in.
4.4. Ballistic Methods
Ballistic methods are so called because they
employ a ballistic galvanometer. In making a
measurement, a test coil connected to a ballistic
galvanometer is placed in the field with its axis in
the direction of the field and the deflection of the
galvanometer is noted when the test coil is either
suddenly removed from the field or rotated 180
about its diameter.10 This produces an impulsive
current in the galvanometer circuit which is
proportional to the total change in linkage between
the field and the test coil.
a. Ballistic Galvanometer

4.3. Pivoted Magnet

The pivoted coil has two disadvantages. It is
too large to go into small spaces such as the gap
in the magnetic circuit of a d'Arsonval instru
ment. Also it requires connections to a source
of current.
The development of permanent
magnet materials having extremely high coercivity has made it possible to produce an instru
ment by which magnetic fields can be measured
in terms of the torque on a very small magnet.
This instrument, called a gaussmeter, consists
of a light-weight shaft pivoted at each end and
supported between jewels. A small cylindrical
Siunanal magnet is mounted on the shaft near
its lower end. At the other end is a hair spring
for measuring the torque and a pointer which
moves over a scale, calibrated in gausses. The
magnet is about an eighth of an inch long in the
direction of the shaft. It is magnetized in a
direction perpendicular to the axis of the shaft.
The moving system is enclosed in a thin-walled
protective tube which extends perpendicularly
from the back of a case which houses the scale
and pointer. The pointer is mounted on the
shaft in a direction parallel to the magneticaxis of the magnet so that when the pointer reads
zero on the scale it also indicates the direction
of the field in which the magnet is located. To
measure the strength of the field, the instrument
is rotated about the axis of the probe until the
scale reading is a maximum. The axis of the
magnet is then at right angles to the direction
of the field and the scale reading shows its value
in gausses. It is important to avoid exposing
the magnet to a field greater than 5,000 gausses
as the magnets cannot be stabilized for higher

The ballistic galvanometer commonly used for

magnetic measurements is a moving-coil instru
ment of the d'Arsonval type.11 It is an instrument
for integrating electrical impulses of short dura
tion. The coil is usually suspended but some
pivoted-coil instruments are also in use. The
moment of inertia of the coil is high and the
restoring force due to the suspension or spring is
low so that the natural period on open circuit is
long, usually of the order of 20 sec or more. The
behavior of the galvanometer is controlled mainly
by the amount of resistance in its circuit. This
determines the amount of electromagnetic damp
ing due to the current induced in the coil as it
moves in the field of the permanent magnet. The
damping due to air friction is usually negligible
compared to the electromagnetic damping.
If the external resistance has a certain critical
value, the return of the coil to its zero position after
a deflection occurs in a minimum of time without
oscillation. This is called critical damping. The
external resistance for critical damping is an
important characteristic of a ballistic gal variometer.
If the external resistance is greater than the
critical value, the return to zero after an impulse
is by a series of oscillations of continually decreas
ing amplitude. The galvanometer is then said to
be underdamped.
If the external resistance is less than the
critical value, the galvanometer is overdamped.
The maximum deflection for a sudden impulse
comes sooner than when critically damped but
10 If the field arises from an electromagnet the field may be reversed by
reversing the current in the magnetizing coil.
11 For an excellent detailed discussion f the theory and operation of the
ballistic galvanometer see Harris, Electrical Measurements, pp. 301-341,
Wiley, 1952.

448-8

the return to zero is slow. For magnetic measure

ments, the ballistic galvanometer is usually used
in the overdamped condition. Some operators,
however, like to have the galvanometer critically
damped or very slightly underdamped. Under
this condition it is necessary that the impulse be
complete before the coil has moved appreciably
from its zero position. Otherwise, if the impulse
is prolonged, the integration will not be complete
and the reading will be low.
For measurements of magnetic fields it is usu
ally possible to produce sudden brief impulses but
in the testing of ferromagnetic materials, the im
pulses are likely to be of such long duration that
it is necessary to use a very heavily overdamped
galvanometer if good accuracy is to be obtained.
As a rule, the intrinsic sensitivity of a ballistic
galvanometer is sufficiently high to permit an in
crease in damping by shunting the galvanometer
as shown in figure 2. The figure shows alternative
methods of connecting a ballistic galvanometer to
a test coil. The choice of methods is a matter of
taste. In both diagrams TC is the test coil, RS
is a series resistor for adjusting the sensitivity, and
RP is a resistor connected in parallel with the
galvanometer when the key K is closed on the lower
contact. In diagram A the galvanometer can be
short-circuited by closing K on the upper contact.
This will stop the motion of the coil almost in
stantly if it is swinging. In diagram B the re
sistor RC is permanently connected to the gal
vanometer and so adjusted that when key K is
open the galvanometer is critically damped.
This allows the galvanometer coil to return to its
zero position after a deflection in minimum time
without oscillation.12 Some operators omit RP
when this connection is used.
If the restoring force could be made zero, the
only control of the galvanometer deflection would
be the electromagnetic damping.13 The motion
of the galvanometer coil would then follow the
change in flux linkage in the test coil exactly and
the final deflection would not be influenced by the
time taken by the change. An instrument whose
torsional restraint is so small that this condition
is closely approximated is called a fluxmeter. As
a rule, however, it is more convenient to have a
suspension or spring stiff enough to give a fairly
stable zero. A galvanometer which is sufficiently
overdamped to approach the performance of a
:rue fluxmeter is said to have fluxmeter charactersties.
If the circuit of a galvanometer with a weak
suspension or spring is not free of temperature
gradients or contacts between dissimilar metals,
hermal electromotive forces may be very troubleorne by causing an unsteady zero and spurious
leflections. Also, it is desirable to have a method
or setting the galvanometer exactly on zero
[uickly and easily. Figure 3 is a diagram of a
" In practice the resistance is often adjusted so that the galvanometer is
i^htly underdamped and will pass beyond zero once before coming to rest,
bis makes it possible to be sure that the coil is swinging freely between tho
o'e pieces without friction.
U Except for a slight amount of nlr damping which usually is negligible.

RS
-AVAWAV

(bg)
TC

K
T

(B)
Figure 2. Connections for ballistic galvanometer.
-wv-

>

RS
,.

zs

Figure 3. Zero adjuster and thermal emf compensator.

device which has been found to be very useful.
Two slide-wire resistors are connected in parallel
to a small dry cell through a high resistance. The
two sliders are connected to opposite ends of a
resistor of low value which is inserted in the gal
vanometer circuit. Contact ZS has a detent so
that it can always be set at the same position.
To compensate for thermal emfs, the slider TC is
set so that the galvanometer does not drift when
contact ZS is on the detent. To set the zero,
contact ZS is moved one way or the other till the
galvanometer reads zero. It is then quickly re
turned to its neutral position. The galvanometer
is calibrated by means of a mutual inductor of
known value whose secondary winding (in fig. 4)
is connected permanently in series with the test
coil.M It is often convenient to adjust the sensi'< If necessary, to avoid the possible effects of stray fields, a noninductivo
resistance equal to that of the secondary coil may be substituted for the coil
after calibration.

449-9

RTC
L
UMMfl

Figure 4. Calibration circuit for ballistic galvanometer.

tivity so that the scale is direct reading in terms
of induction thus avoiding the necessity of multi
plying the reading by an odd-valued scale constant
to obtain the value of the field in specified units.
The calibrating current which must be switched
in the primary of the mutual inductor is calculated
by the formula
KBaN
where
7e=current
K=& constant dependent upon the system of units
5=induction
aN= area-turns of the test coil
Z=mutual inductance.
In the cgs system / is in amperes, B is in gausses,
aN is in cm2-turns, Lm is in henries, and K is
10-8. In the mksa system
is in amperes, B is
in teslas, aN is in m2-turns, Lm is in henries, and
K is unity.
If the field is to be measured by withdrawing
the test coil, the resistance in the galvanometer
circuit is adjusted so that when Ic is suddenly
reduced to zero the deflection is 10 cm. If the
field is to be measured by flipping the coil through
180 or by reversing the field, the galvanometer
is adjusted to read 10 cm when Ie is reversed.
If the value of H is desired, it is only necessary
to divide the observed value of B by Tm. In the
cgs system Tm is unity so that H is numerically
the same as B. In the mksa system Tm is 4irX 10~7
so that it is not possible as it is in the cgs system
simply to measure B and call it H.
b. Intercomparison of Mutual Inductors and
Calibration of Test Coils
At the National Bureau of Standards the stand
ard of mutual inductance is one built with con
siderable care for use in determining the unit of
resistance. The value of mutual inductance has
been computed in terms of the dimensions of
the inductor and is known to a few parts in a mil
lion. Since ballistic magnetic measurements are
carried out with direct currents, the working stand
ards of mutual inductance are calibrated by com
parison with the primary standard using a direct-

Figure 5. Diagram of connections.

current method somewhat similar to one proposed
by Maxwell [l].18 A diagram of connections is
shown in figure 5. Lmi and J.mi are the mutual
inductors to be compared. Lni must have the
greater value. The primary windings are con
nected in series to the battery, B, through the
reversing switch, S, the rheostat, C, and the am
meter, A. The secondary windings are connected
so as to oppose each other. Ri and lit are adjust
able precision resistors. It is important that the
ballistic galvanometer, , be connected as indi
cated because if it is connected between points
/?and G leakage currents might give trouble unless
the insulation between coils in the inductors is
practically perfect. A measurement is carried out
as follows. With switch St, closed Rt is set at
some convenient value and the primary current
is set at a value not exceeding the current-carrying
capacity of the primary windings of the inductors.
R2 is then adjusted so that upon reversal of the
primary current by switch S there is no residual
deflection of the galvanometer. If the self-induct
ances of the secondary windings of the two in
ductors are not equal there may be a small double
kick but this is generally not troublesome. When
balance is obtained, the value of
is
T

-T

Rl.

"LR is the total resistance of the secondary circuit

of Lm\ including Ru R2, the secondary winding of
Lml, and all the leads. The excess over
R2 is
determined by setting Ri and R2 to zero, opening
switch St, and measuring the resistance of the rest
of the circuit by means of a Wheatstone bridge
connected across SV Correction for the resistance
of the 1 eads to the bridge can be made by measuring
their resistance with switch S closed. The usual
precautions should be taken against the effect of
stray fields. Leadsshouldbe twisted and the two in
ductors should be located at some distance from each
other and so oriented that no interaction between
the primary of one inductor and the secondary of
' Figures In brackets Indicate the literature references on page 33.

450-10

the other can be observed. Some errors may

result from capacitance between the primary and
secondary circuits. This can be minimized by
repeating the observations after reversing the
connections of the primary of one inductor and
the secondary of the other and averaging the
results.
It is usually not feasible to calculate the value
of area-turns of a test coil with a satisfactory
degree of accuracy, especially if it has more than
one layer. It is much better to determine the
value experimentally. If a test coil is placed at
the middle of a long slender solenoid whose pitch
of winding is uniform and accurately known, its
area-turns can be calculated from the mutual
inductance between the solenoid and test coil
and the induction per unit current in the solenoid.
The mutual inductance can be determined by the
method described above. It is important that
the axis of the test coil be alined with that of
the solenoid. The error due to lack of alinement
is proportional to 1 cos 8 where 0 is the angle
between the two axes. If the angle is less than
2.5, the error from this source will not exceed
0.1 percent.
The value of area-turns is calculated from the
equation
{aN)=KLJC
where
aN= area-turns
K= a constant depending on the system of units
Lm=the mutual inductance
C=the value of induction per ampere in the
solenoid.
In the cgs system, K is 10s, Lm is in henries, and
C is in gausses per ampere. Area is then in square
centimeters. In the mksa system K is unity,
Lm is in henries, C is in teslas per ampere, and area
is in square meters.
If a coil whose area-turns value is to be deter
mined is too large to insert in an available solenoid
or has a handle which cannot be detached, the
calibration field may be obtained by means of a
Helmholtz arrangement. This consists of two
identical very short coaxial coils, the axial distance
between them being equal to their radius. This
produces a very uniform field along about the
middle third of the axis between the coils. If a
is the radius in centimeters, N is the number of
turns in each coil, and i is the current in amperes,
the magnetizing force at the midpoint between
the two coils is
H= /^n2\m oersteds.

This reduces to
0.899lA/i
, ,
H=
oersteds.
a

To convert to ampere turns per meter (mksa

units) multiply by 103/4ir=79.58. If coils having
very small values of area-turns are to be tested
this can be done in the field between the poles of
an electromagnet. The field must be uniform
over the area of the test coil and measured by a
standard method. Since it is generally not
feasible to reverse the current in the electromagnet,
the reading must be made either upon suddenly
removing the coil from the field or preferably by
rotating it through 180 about an axis which is
perpendicular to the direction of the field.
4.5. Rotating Test Coil
A continuously rotating test coil can be used
to measure magnetic fields. The coil is rotated
at constant speed about a diameter. It is con
nected through a commutator or cam-operated
reversing switch to a d-c instrument which indi
cates the average voltage induced in the coil.
If the reversal comes at the time when the a-c
voltage induced in the coil is zero, the reading
on the instrument will be a maximum and will
be proportional to the magnetic field. Commercial
instruments operating on this principle have the
rotating coil at the end of a long shaft which has
at its other end a synchronous motor. The shaft
has to be long enough to avoid errors due to stray
fields from the motor. The accuracy of the
measurement depends directly upon the accuracy
with which the frequency of the power source is
controlled. The apparatus is generally calibrated
by observations in known fields.
4.6. Bismuth Spiral [2]
One of the so-called secondary effects which is
utilized to measure magnetic fields is the change
in electrical resistance of a conductor when it is
subjected to the influence of a transverse magnetic
field. Most metals show this effect, but only in
bismuth is it sufficiently large to be of practical
value for the measurement of magnetic fields.
The bismuth spiral is made of small insulated
bismuth wire wound bifilarly in a flat spiral, all
the turns being in one plane. The ends are usually
soldered to heavy copper leads. The spiral is
held in place and protected from mechanical
damage by thin discs of mica which are fastened
to the copper leads. The diameters of the spirals
range from 0.5 to 2.0 cm and they are ordinarily
about a millimeter thick overall. It is essential
that the bismuth be of very high purity. Other
wise, anomalous and uncertain results will be
obtained. The major difficulty encountered in
the use of a bismuth spiral is that it has two
temperature coefficients; not only does bismuth
have the ordinary temperature coefficient in the
absence of a magnetic field but also the change in
resistance due to a magnetic field depends upon
the temperature. If proportional change in re
sistance is plotted against the value of magnetic
field, the curve is not linear below about 5,000
gausses. Above this value the curve is practically

451-H

linear but the projection of the linear part does not

pass through zero.
Bismuth spirals are not much used at present
but the principle might be useful for some
applications.
4.7. Hall Effect
The Hall effect furnishes a very convenient
means for the measurement of magnetic fields.
Instruments are available which utilize this prin
ciple. It can be very simply stated as follows.
If a thin strip or film of metal, usually bismuth,
indium antimonide, or indium arsenide, has a
current in it, two points can be found at opposite
ends of a line approximately at right angles to the
current which will be at the same electrical
potential if no magnetic field is acting. If a
magnetic field is applied at right angles to the
plane of the strip a difference of potential appears
between the two points. This difference of poten
tial is proportional to the intensity of the magnetic
field. In practical instruments a-c is used so that
the emf produced by the action of a magnetic
field can be amplified and thus provide increased
sensitivity. Calibration is carried out in magnetic
fields of known value.
4.8. Saturable Core
A saturable core 16 is one which reaches practical
saturation under the influence of a relatively low
magnetic field. Thus a very small change in the
effective magnetizing force will produce a large
change in the inductance of a coil containing a
saturable core. This principle is used for the
measurement of magnetic fields. This is done by
energizing a coil containing a saturable core with
alternating current. The frequency is usually a
few kilocycles per second. If a steady magnetiz
ing field such as that of the earth is tdso present,
there will be induced in the circuit a second
harmonic component. This second harmonic
voltage is proportional to the intensity of the
superposed steady field. The steady field is de
termined in either of two ways. The amplified
second harmonic voltage may be filtered out and
measured or the steady field may be neutralized

by means of direct current in a coil surrounding the

core. When this is done, the second harmonic
voltage disappears. The value of the field is then
calculated in terms of the direct current, and the
turns and dimensions of the coil. This is the more
precise method because variations in the amplifier
only change the sensitivity with which the balance
can be made. If the second harmonic voltage is
measured directly, the apparatus must be cali
brated in fields of known values.
4.9. Nuclear Magnetic Resonance
Nuclear magnetic resonance is another phenom
enon by which it is possible to measure homo
geneous magnetic fields with very high accuracy.
The measurement is based on the equation
2*v=yvB
or transposing,

In this equation B is the field in gausses, v is the

resonance frequency in cycles per second and 7- is
a quantity called the gyromagnetic ratio. The
gyromagnetic ratio yp is the ratio of the magnetic
moment of a nucleus to its angular momentum.
The measurement is made by immersing a
sample of material containing nuclei, whose gyro
magnetic ratio is known, in the steady field to be
measured. The sample is then subjected by means
of a surrounding coil to an alternating magnetic
field at right angles to the steady field. The
frequency of the alternating field is adjusted to be
equal to the resonance frequency of trie nuclei in
the sample. At this frequency which is directly
proportional to the intensity of the steady field
there is an absorption of energy from the exciting
circuit. This is generally observed by means of a
cathode ray oscilloscope. B is then calculated,
using the equation
r> gausses 1717
K=
where v is the resonance frequency and yp is the
gyromagnetic ratio.

5. Tests With Direct Currents

5.1. Testing of Materials
General Principles. The principal object of
testing ferromagnetic materials by d-c methods is
to obtain data from which normal induction curves
and hysteresis loops can be plotted. This is done
either by using suitably shaped specimens such as
rings which constitute the entire metallic part of
the magnetic circuit or by means of permeameters.
Several devices using this principle arc described in the book, Saturating
Core Devices, by L. R. Crow, The Scientific Publishing Company, Yincenncs. Indiana.

A permeameter is a magnetic circuit provided

with magnetizing and test windings in which a
test specimen can be inserted so as to compose a
part of the circuit. Since there is no material
which acts as an insulator with respect to mag
netism, the magnetic circuit must be arranged so
as to produce the greatest possible uniformity in
the distribution of magnetic induction across the
section and along the length of that part of the
specimen directly involved in the test. Several
permeameters have been developed which differ
" To convert from gausses (ces) to tes'.as (mksa) multiply by 1 X 10*.

452-12

from each other in form of magnetic circuit and

method of determining corresponding values of
H and B. In the Koepsel [3] and Esterline [4]
permeameters. the ends of the specimen are
clamped between iron pole pieces resembling those
of a d'Arsonval instrument but much larger. The
magnetizing winding surrounds the specimen.
Compensating coils in series with the main mag
netizing winding are mounted on the pole pieces.
Their function is to compensate for the magnetic
reluctance of the pole pieces. In the Koepsel
apparatus there is a regular d'Arsonval moving
coil whose pointer gives an indication of the
value and direction of B in the specimen when
there is a current of fixed value in the coil. Values
of H are estimated in terms of the current in the
magnetizing winding. In the Esterline apparatus,
the d'Arsonval coil is replaced by a rotating arma
ture. When this armature is driven at a constant
speed the induced voltage is taken to be a meas
ure of B in the specimen. Neither of these
permeameters is capable of very high accuracy
and neither is at present in common use.
Other permeameters have been developed by
DuBois [5], Hughes [6], Picou [7], Iliovici [8],
Niwa [9], and Burrows [10]. The Burrows per
meameter is based on the principle of so distrib
uting the magnetomotive force around the
magnetic circuit that there is negligible magnetic
leakage, at least from the specimen under test. It
was for many years accepted as the standard
apparatus for d-c magnetic testing. However,
the operation of this apparatus is tedious and
time-consuming and is unduly sensitive to nonuniformity in magnetic properties along the
length of the specimen. Also, it requires two
similar specimens. The MH permeameter [11]
was developed with the object of eliminating or
minimizing these drawbacks and has been adopted
by the Magnetic Measurements Section as the
standard method for d-c magnetic testing in the
range of magnetizing force up to 300 oersteds
(approximately 24 X 103 amp-turns/m) . The Fahy
Simplex permeameter [12] has been in general use
for many years for tests in the same range of
magnetizing force and for this reason is used for
testing when so requested. The advent of very
hard magnetic materials requiring the use of very
high magnetizing forces for testing led to the
development and adoption of the High-H permeam
eter [13] for magnetizing forces up to 5,000
oersteds (approximately 40X104 amp-turns/m).
These three methods, the MH, High-H, and
Fahy Simplex permeameters, are called ballistic
methods because a ballistic galvanometer is
employed in the measurements.
In ballistic tests, values of B are obtained in
terms of the deflection of a ballistic galvanometer
connected to a test coil (B-coil) surrounding the
specimen. The deflection is proportional to the
change in the induction linked with the test coil.
If the change is simply a reversal in direction,
then the induction is one half of the observed
change. There are two ways in which values of

H may be determined. One way is to calculate

it in terms of the current-turns per unit length
at a given section of the magnetic circuit. This
is the method employed in the testing of ring
specimens and in the Burrows permeameter.
The other way is to determine B in the air at a
suitable location and calculate the value of H
from the equation
H=B/Tm.
Since rm = l in the cgsem system, it has become
general practice to determine B 18 and call it H.
This does not lead to a numerical error in the
cgsem system. It has also become general practice
to call the coils by which this determination is
made //-coils. If H is calculated in terms of
current the magnetic circuit and magnetizing
windings must be so arranged that there is no
magnetic leakage from the part of the specimen
over which the J?-coil extends and the magnetizing
winding must be uniformly distributed over the
same portion of the specimen. This is the
principle involved in the Burrows permeameter
and approximated in the ring test.
If H is determined ballistically, it is not neces
sary that the magnetizing winding surround the
section of the specimen surrounded by the Bcoil but the magnetic circuit must be so arranged
that the field of induction in the region adjacent
to the specimen occupied by the H-coil is uniform
or if not it must be possible to extrapolate to the
surface of the specimen. This is the underlying
principle of the MH and High-H permeameters."
In ballistic tests, as described earlier, the gal
vanometer is calibrated by means of a mutual
inductor whose secondary winding is part of the
galvanometer circuit. Excepting during cali
bration, the secondary winding may be replaced
by a noninductive resistor having the same
resistance. This is to avoid errors which might
be caused by pickup from other parts of the
circuit.
It is customary to adjust the sensitivity of the
galvanometer so that the scale is direct-reading
in terms of induction or magnetizing force.
Thus, in the cgs system, the magnetic induction
on reversal may be 1,000 gausses/cm of deflection
or the magnetizing force may be 1, 10, or 100
oersteds/cm. By this procedure the necessity
of multiplying the scale reading by an odd-valued
factor is avoided. If H is measured in terms of
current in the magnetizing winding, only the
calibration for B needs to be made.
The calibrating current for a selected value of
B is given by the equation
e~

KBAN
T

" Any quantity determined in terms of an induced emf in a test coil is

by 'definition
not H.
The Fahy11 and
Simplex
permeameter operates on a somewhat diflerent prin
ciple and is described later.

453-13

where
Ie= calibrating current in amperes
K=a constant depending upon the system of
units.
B the selected value of induction
/l=the cross sectional area of the specimen
AT=number of turns in the test coil
Lm=mutual inductance in henries.
In the cgs system K is 10"8, B is in gausses, and
A is in square centimeters.
In the mksa system, K is unity, B is in teslas,
and A is in square meters.
In calibrating for a selected value of // the
equation becomes
T

KTmHaN

where
Ie= calibrating current in amperes
=a constant depending upon the system of
units.
rm=the magnetic constant
H=thc selected value of magnetizing force
aN= area-turns of the test coil
Z>m = mutual inductance in henries.
In the cgs system K is If)-8, rm is unity, H is
the selected value of magnetizing force in oersteds,
and aN is the area-turns of the test coil in cm2turns.
In the mksa system, K is unity, rm is 47rX10~7,
H is the selected value of magnetizing force in
anipere-tums per meter, and aN is area-turns in
m2-turns.
Since the calibration is usually made by reversal
of the calibrating current, care must be taken
when measuring changes in induction or magnet
izing force as in the determination of points on
the hysteresis loop to multiply the readings by 2.
Because the area of the B-coil is larger than that
of the specimen, a correction must be made to the
observed value of B to account for the "air"
flux existing in the space between the coil and the
specimen. Thus the induction, B, is
B=Boba-kTmH
where
Bob,=observed value of B (for specimen of area A)
, a A
k~ A
a = area of the test coil
.4= area of the specimen
rm=thc magnetic constant
fl==the magnetizing force.
In the cgs system, B is in gausses, Tm is unity
and II is in oersteds. In the mksa system, B is
in teslas, Tm is 47rX10~7 and H is in ampere-turns
per meter.
A: is a numeric and has the same value in either
system.

In making observations of normal induction,

the specimen should first be demagnetized to
eliminate latent polarization due to previous
magnetization. Tins is done by subjecting it to
the influence of reversals of magnetizing force of
gradually decreasing magnitude starting from a
peak value well into the third stage of magnetiza
tion. The frequency of reversal should be low
enough so that induced eddy currents will not
interfere with the demagnetizing process. About
lc/s has been found satisfactory for the average
case. If the specimen does not compose the
entire magnetic circuit and has a small cross
section compared with the rest, it may be neces
sary to demagnetize the rest of the magnetic
circuit using a larger specimen and then to repeat
the process with the smaller specimen in place.
After demagnetization, the magnetizing current
is set at the value corresponding to the lowestpoint desired and reversed several times until
successive readings of the induction are the same.
The specimen is then in a symmetrically cyclic
condition and the observed induction is the normal
induction. The corresponding magnetizing force
is then determined either by taking a ballistic
deflection or by observing the magnetizing current
according to the type of apparatus being used.
Additional points on the normal induction curve
are determined in the same way. Demagnetiza
tion need not be repeated if each point so deter
mined is higher than any preceding one. It is
the practice of some observers to start with the
highest point and then to demagnetize from each
point to the next lower one. This is sometimes
preferable, especially if the specimen is likely to
be heated unduly.
If points on a hysteresis loop are to be deter
mined, cyclic condition is first obtained by several
reversals of the magnetizing current corresponding
to the tip of the loop and the resulting values of
B and H observed. The current is then suddeidy
reduced in value by inserting additional resistance
in the magnetizing circuit. For points on the
negative side of the i7-axis, the current is simul
taneously reversed and reduced. The observed
changes in B and 27 are subtracted from the
values at the tip and the results thus obtained are
taken to be the coordinates of the required point
on the hysteresis loop.
The process is repeated for other points, care
being taken to reestablish cyclic condition at the
tip before each observation.
In setting up apparatus for ballistic testing, it
is important to twist the conductors in both
primary and secondary circuits and so to locate
the mutual inductor with reference to the rest of
the apparatus as to prevent errors due to stray
fields.
5.2. Ring Method
Although a ring would appear to be the idea
form of specimen for magnetic testing, it has
certain disadvantages which should always tx

454-14

taken into account. The principal advantage is

the absence of airgaps and end effects.
In the cgs system, H is calculated using the
equation
0.4 N 1
H=
D
where
H= magnetizing force in oersteds
N= number of turns in winding
7= current in amperes
19= mean diameter of the ring in cm.
To convert to rationalized mksa units (ampereturns per meter) multiply by 1000/4ir= 79.58.
This calculation is sufficiently accurate for most
purposes provided that the mean diameter of the
ring is large compared to its radial width. Since
the outer circumference of the ring is greater than
the inner circumference, the turns per unit length
are less on the outside than on the inside and
consequently the magnetizing force is not uniform
across the section. Table 2 shows the ratio of the
average value of H to that at the mean radius in
rings of either circular or rectangular cross section.
For difference no greater than 1 percent, a ratio
of at least 10 of mean diameter to radial width is
usually recommended.
Table 2. Ratio of average value of H to value at mean
radius in rings of rectangular and circular sections [13]
II.IH,
Ratio of diameter to radial thickness
Rectangular
1.0980
1.0397
1.0216
1.0137
1.0094
1.0069
1.0052
1.0033
1.0009

jk.
.. ...

Circular
1.0718
1.0294
1.0163
1.0102
1.0O70
1.0052
1.0040
1.0025
1.0007

It is not safe to assume, however, that simply

laking the ratio 10 or more assures accurate
3sults. Errors may also result from irregularity
l winding or from nonuniform magnetic properties
long the circumference of the ring. Irregularity
i winding can be minimized by the exercise of
ctreme care in winding or by the use of a winding
i

machine. However, errors due to nonuniformity

in the specimen which caD neither be calculated nor
conveniently determined experimentally may be
large and therefore, the use of ring specimens as
reference standards for checking the accuracy of
other methods is not recommended. The method
is also limited to low magnetizing forces which can
be applied without producing excessive heating.
In spite of the disadvantages of the ring speci
men for measurements of high accuracy or for
standardizing purposes, this form of specimen is
widely used m production testing for quality con
trol or the development of new materials, notably
the ferrites. Its use has been greatly facilitated
in recent years by the development of a machine
by which extremely uniform windings can be
rapidly applied. The secondary winding is
applied first so as to minimize the effect of air-flux
between coil and specimen.
A diagram of connections for the ring test is
shown in figure 6. Current is derived from the
battery, B, controlled by the resistors R and R',
and measured by the ammeter A or some more
precise method for measuring current. The cur
rent can be reversed by switch C or reduced by
opening switch C", or reversed and reduced by
operating both switches simultaneously. If
switch D is closed upward the current is in the
magnetizing coil M. If D is closed downwards,
the current will be in the primary winding, P, of
the calibrating inductor.
The secondary or test winding, T, is connected
in series with the secondary winding of the induc
tor and the series resistor, RS, and the parallel
resistor, RP, as shown. The ballistic galvanom
eter is connected through the key k across the
resistor RP.
The procedure for making a test is described on
p. 13 et seq.
5.3. Straight Bar and Solenoid
It is possible to make ballistic tests with a
straight specimen of uniform cross section mag
netized in a solenoid. In this case, the deter
mination of the magnetic induction at the mid
section of the bar offers no particular difficulty.
The principal problem lies in the determination of
the magnetizing force. This arises from the fact
that the magnetic circuit comprises not only the
bar but also the air-space in which the lines of

hi'-

*
C

Figure 6. Connections for the ring lest.

31S-411 O - 68 - 30

455-15

Table 3.

Demagnetizing coefficient and factor

II-IT DaB,
r

D = ratio of length to
diameter

Equation
Ellipsoid
N

10
15
20.
30
40
50
60.
70 .
80
100
150
200

0.2549
. 1350
.0848
.0432
.0266
.0181
.0132
.0101
.0080
.0054
.0026
.0016

Du Bois [15]

Mann [14]
N

Da

0.2550 0.0203
.1400 .0111
.0808 .00714
.0460 .00366
.0274 .00218
.0183 .00146
.0131 .00104
.0099- .00080
.0078 .00062
.0052 .00041
.0025 .00020
.0015 .00012

Shuddemagen [16]

A'

Da

Da

0.2160
.1206
.0775
.0393
.0238
.0162
.0118
.0089
.0069
.0045
.0020
.0011

0. 0172
.0096
.0062
.00313
.00189
.00129
.00094
.00071
.00055
.00036
.00016
.00009

0.2000
. 1040
.0655
.0335
.0206
.0139
.0101
.0077
.0061
.0041
.0020
.0012

0.0159
.0083
.0052
.00267
.00164
.00110
.00080
. 00061
.00049
.00033
.00016
.00009

0.2512
. 1235
.0747
.0367
.0222
.0150
.0109
.0083
.0066
.0045
.0022
.0013

Da
0.0200
.0098
.0059
.00292
.00176
.00119
.00087
.00066
.00052
.00035
.00017
.00011

logu2V-1.15-1.7SlogiiD
logiiD-0.05-1.75 log,i)
Note: The above values of Nana Da are calculated for cgsem system of units.
induction extend from one end of the bar to the
other. Thus the total magnetomotive force is not
confined to the space within the solenoid and con
sequently the magnetizing force available for the
bar is less than would be calculated in terms of
the ampere-turns per unit length in the solenoid.
Classical theory attributes this diminution to the
effect of fictitious poles near the ends of the speci
men and therefore the effect has been called a "de
magnetizing" effect. The effect is expressed by
the equation

H=the effective magnetizing force

H' = tbe apparent magnetizing force
Z?B=the demagnetizing coefficient
Bt= intrinsic induction
Tm=the magnetic constant.
In the cgs system, rm is 1. In the rationalized
mksa system, Ym is 47rX10~7.
Except for ellipsoids of revolution, DB cannot be
calculated but must be determined experimentally.
It has been found that the demagnetizing coefficient
is practically constant up to about the top of the
second stage of magnetization. Above this the
value decreases with increased induction. Within
the linear range, the value for bars of circular sec
tion is a function of the ratio of the length to
diameter of the specimen 20 but published tables
by different experimenters are not in perfect
agreement.
According to Bozorth [32], the demagnetizing
factor depends not only upon the dimensional
ratio but also upon the permeability. He also
10 For sections other than circular, the diameter is taken to be the diameter
of the circle which has the same area as that of the bar.

gives graphs by which apparent permeability can

be converted to true permeability or vice versa.
Most of the tables available at the present writing
are given in terms of "intensity of magnetization"
rather than in terms of Bt. The equation com
monly given is
H=H'-NI
where H and H' are the true and applied magnet
izing force respectively, N is the demagnetizing
factor, and / is the "intensity of magnetization."
The intensity of magnetization as understood here
is Bt/4ir so that
2V"=4rZ?fl.
Table 3 shows some published values of the
demagnetizing factors together with equations
which practically average the values given by
Mann [14], DuBois [15], and Shuddemagen [16].
These equations provide a means for estimating
the demagnetizing factor for intermediate values
of the dimensional ratio.
The diagram of connections shown in figure 6
for the ring test is substantially right for the sole
noid test. The only changes required are to take
M to be the solenoid winding and T to represent
the test coil at the middle of the test bar. In cal
culating the applied magnetizing force H' , if the
solenoid is very long compared to its mean diam
eter and if the cgs system of units is used, the
equation is
#' = 0.4ir 7i /
where
H' = applied magnetizing force in oersteds
n = number of turns per cm
1= current in amperes.
If the solenoid is relatively short, a more nearly
accurate value can be obtained by using the

456-16

equation
u,_0AmIl
where
i=length of the solenoid in cm
and
</=mean diameter of the winding in cm.
A more accurate value for multilayer coils can be
obtained by calculating each layer separately and
adding the results. To convert from oersteds to
rationalized mksa units, multiply bv 1000/4*=
79.58.
Unless the test coil is wound directly upon the
specimen its area should be determined and the
standard air-flux correction should be applied to
the observed value of B. (See p. 14.)
On account of the difficulty of obtaining accu
rate values of the demagnetizing factor this
method is not capable of high accuracy but for
some purposes it may be the most convenient for
straight specimens if approximate results are
good enough.

Figure 7. MH permeameter.

Figure 8. Test coil assembly for MH permeameter.

5.4. Permeameters
a. MH Permeameter
The MH 21 permeameter [11] was designed with
the purpose of producing an instrument which
would not have the disadvantages inherent in
the Burrows permeameter but would have com
parable accuracy. It should also be an absolute
instrument in the sense that its constants can be
determined from its own dimensions and therefore
do not have to be calibrated by reference to any
other permeameter. The MH permeameter fulfils
these essential requirements. In addition, it has
the advantages that it requires only a single
specimen and is more simple and rapid to operate
than the Burrows permeameter. Furthermore, it
is much less sensitive to the effect of nonuniform
magnetic properties along the length of the
specimen. It is also used for ordinary magnetic
testing within its proper range; that is, up to
300 oersteds or 24X103 amp-turns per meter.
The permeameter is shown in figure 7. The
magnetic circuit is made of laminated electrical
sheet of armature grade. The laminated struc
ture reduces the effect of eddy currents induced
in the circuit by the changes in flux incident to
the testing procedure. The specimen and the
symmetrical U-shaped yokes which span it rest
on two end pieces supported by Bakelite blocks.
The yokes are recessed so as to make good contact
with the end pieces and the ends of the specimen
in such a way that no clamping is required. This
eliminates mechanical strain in the specimen.
The symmetrical yokes promote a uniform distri
bution of magnetic induction across the section
of the specimen and along its length. The dis" MH is a designation for "medium //."

tribution is also improved by auxiliary windings

around the ends of the yokes and the end pieces.
These are connected in series with the main
magnetizing coil which surrounds the specimen.
The proper number of turns in these auxiliary
coils depends upon the magnetic properties of the
yokes and end pieces and is determined experi
mentally for each instrument. The preferred
length of test specimen is 28 cm but specimens as
short as 24 cm can be used if necessary.
The test-coil assembly is shown in figure 8.
This assembly is held in place within the magnet
izing coil by two Bakelite rings. A platform
extending between these rings serves to hold the
B-coil in place and to support flexible specimens
such as strips of electrical sheet so as to prevent
errors due to mechanical strain.
A 100-turn .B-coil extends over the middle 3 cm
of the specimen. Just above it, but not surround
ing the specimen, is a compensating coil adjusted
to have the same value of area-turns as the 2?-coil.
This coil is connected in series opposing the Bcoil so that the reading of the galvanometer upon
reversal of the magnetizing current is propor
tional to the intrinsic induction, B,. Thus the
usual air-flux correction is not required. If
desired, the total induction, B, can be obtained
simply by adding to the observed value of B, the
value of TmII. In the cgs system r= 1. In the
mksa system, rm=4rX 10~* The 5-coil is able
to accommodate specimens as large as 1X3 cm in
cross section.
The //-coil system is also shown in the figure 8.
It consists of two rectangular coils with nearly
identical values of area-turns mounted on a turn
table below the specimen. The upper coil extends
into a recess in the supporting platform so as to

457-17

Figure 9. Connections for the MH permeameter.

come close to the surface of the specimen. The
turntable can be rotated through 180 about a
vertical axis by means of a rack and pinion arrange
ment operated by an iron-clad solenoid and
plunger located far enough away from the mag
netic circuit to avoid interference. The lower
coil has the larger value of area-turns and is
adjusted to equality of effective value of areaturns by means of a noninductive shunting
resistor. If the two coils are connected in series
opposing and rotated the reading of the galvanom
eter will be proportional to the gradient of the
magnetic field. A reading taken by means of
the upper coil alone together with the gradient
permits extrapolation to the surface of the speci
men and thus the determination of the magnet
izing force acting on the specimen.
The use of rotating coils (flip coils) has some
advantages over a fixed-coil system, especially
when determining points on a hysteresis loop.
With fixed //-coils, it is necessary to calculate
values of Ha for a point on the loop in terms of
the difference between two large quantities.
With the flip coils the value of Hd is determined
directly with suitable sensitivity and observations
can be repeated without the necessity of reestab
lishing a symmetrically magnetized condition.
The Burrows criterion for perfect demagnetiza
tion is that the induction for a given low mag
netizing force shall be a maximum. This is not
feasible to apply experimentally. With the flip
coils, it is only necessary to observe whether or
not the magnetizing force corresponding to a
given low magnetizing current lias the same
absolute value regardless of the direction of the
current. If there is a difference, the presence of
residual magnetization is indicated. The demag
netizing procedure should then bo repeated.
Taking into consideration all the possible
sources of error inherent in the apparatus, it is
estimated that by the exercise of great care in

taking the readings and averaging several reading

for each point, values of induction are accurate
to within about 0.5 percent. Values of magnet
izing force will come within the same accuracy
or 0.05 oersted, whichever is larger. For routine
tests, values of either induction or magnetizing
force accurate to within 1 percent should be
obtained without difficulty.
Figure 9 is a diagram of connections for t he
MH permeameter.
Battery E furnishes the magnetizing current
which is controlled by resistors R and /?' and
measured by ammeter A or by a standard shunt
and potentiometer. With switch D closed up
ward the current is in the primary winding P of
the calibrating inductor Lm. If it is closed down
ward, the magnetizing winding is energized.
With selector switch F closed on b, the B-eo\\
B is connected to the galvanometer circuit
through the calibrating resistor RB,. If F
closed on h, the H-coils are in the galvanometer
circuit through the calibrating resistor RH,. The
selector switch, G, determines whether the upper
coil, u, alone is in the circuit or the coils u and
/ are connected in series opposing. Resistor *A
is for adjusting the effective area turns of coil
I to be equal to the area-turns of coil u.
noninductive resistor, is to compensate for the
resistance of coil I so that the resistance of the
galvanometer circuit will be the same whether
one or two coils are connected.
It should be noted that for a specimen which is
magnetically nonuniform along its length, the
values of B and H obtained by the MH per
meameter relate only to the part of the specimen
linked with the B-coil.
b. High-H Permeameter
The High-H permeameter [13] shown in figur--'
10 is used for testing "hard" magnetic materials
at magnetizing forces up to 5,000 oersteds

458-18

Figure 11. Pole pieces of the high-H permeameler,

Figure 10. High-H permeameler.

(40X104 amp. turns/meter). Since only a short
length of the specimen is included in the test,
it is possible to apply the high magnetizing forces
without appreciable heating. The magnetic cir
cuit is made of laminated electrical sheet of
armature grade. The specimen, of rectangular
cross section, is held in pole pieces extending
between two U-shaped yokes. The construction
of the pole pieces is indicated in the sketch of
figure 11. They have longitudinal channels into
which filler pieces are fitted as shown. The
specimen is located in the space between the
bottom of the channel and the filler. The dis
tance between pole pieces, or gap length, is
adjustable; and scales are provided to indicate
the gap length up to 10 cm, which is the maximum
generally used. The filler pieces are fastened at
their outer ends to heavy brass plates which are
clamped to the ends of the pole pieces so as to
keep the fillers parallel with the bottom of the
channels. The main magnetizing coils surround
the pole pieces and auxiliary coils are wound on
the yokes.
Two similar //-coils are mounted one above the
other on a vertical shaft below the specimen.
They can be rotated 180 by means of a small
motor through beveled gears and a long horizontal
shaft. The lower coil lias a slightly greater value
of area-turns than the upper coil and is adjusted
to have the same effective value by a noninductive
shunt. If the two coils are connected to the
galvanometer in series, opposing, and rotated
through 180 from a position such that their axes
are parallel to the direction of the field, the
deflection is proportional to the radial gradient
of the field. If the axis of the upper coil is set to
be the same distance below the surface of the

specimen as the distance between the axes of the

coils this gradient can be used directly to extrap
olate the value obtained by the upper coil alone
to the surface of the specimen.
In order to keep the air correction small,
individual B-coils are wound on thin brass forms
to fit specimens of various sizes. The coils are
5 mm long and usually have 25 turns.
Relatively short specimens can be used but for
specimens 5 cm or less in length, it is better to
butt the ends against the pole pieces rather than
to attempt to insert them in the regular way.
It is estimated that for magnetizing forces from
100 to 5,000 oersteds (8X103 to 40X10* ampturns/meter) it is possible under favorable condi
tions to obtain values of either induction or
magnetizing force which will be accurate within
0.5 percent. Under ordinary conditions of routine
testing, the errors probably do not exceed 1
percent.
The wiring diagram shown in figure 9 applies
also to the High-H permeameter if it is under
stood that the magnetizing coil, M, is on the pole
pieces instead of over the specimen.
c. Fahy Simplex Permeameter
For several years, the Fahy Simplex permeam
eter [12] has been in use in many laboratories for
routine magnetic testing. Since this permeameter
tends to give average values of the properties
of a nonuniform specimen, agreement with another
permeameter which includes only a relatively
short section in the test should be expected only
for a uniform specimen. Unless a specimen has
been checked for uniformity, it is safer to assume
that it is nonuniform, and lack of agreement
should not be interpreted as an error of either
permeameter.
For checking other Fahy Simplex perineameters,
the Magnetic Measurements Section has calibrated
one by reference to a standard permeameter
using uniform bars as standards. This instru
ment is used as a working standard and consti^iies
a practical means of making the check.

459-19

Figure 13. Modified Fahy Simplex permeameter.

Figure 12. Magnetic circuit of Fahy Simplex permeameter.
The magnetic circuit of the Fahy Simplex
permeameter is indicated in the sketch of figure 12.
The yoke, Y, is made of laminated silicon sheet.
Two soft iron blocks, H, make contact with the
ends of the specimen. Clamps hold these blocks
and the specimen in place against the pole pieces
of the yoke as shown. The magnetizing winding
is wound on the yoke. The 7?-coil for measuring
the magnetic induction has 100 turns uniformly
wound on a brass form which surrounds the
specimen and extends over its whole length
between the pole pieces. For determining the
magnetizing force, a uniformly wound //-coil of
many turns on a straight nonmagnetic form
extends horizontally between the upper ends of
the //-blocks.
In order to gain the advantages of "flip coils"
for the determination of magnetizing force,
especially for points on the h\Tsteresis loop, the
NBS instrument was modified by providing a
means of rotating the //-coils through 180 about
a vertical axis. The modified instrument is
shown in figure 13. The //-blocks had to be
lengthened slightly so that the //-coil could clear
the magnetizing coil but experiment showed that
this did not affect the accuracy of the results.
As previously stated, the flip coil arrangement
makes it possible to determine values of Ht on the
hysteresis loop with adequate sensitivity and to
repeat the observation without the necessity of
repeating the process of producing a symmetrically
cyclic condition for each repetition of the observa
tion. It also provides a convenient and positive
indication of proper demagnetization. Demag
netization has been complete if for a very small
magnetizing force, the values of H are equal for
reversals with the magnetizing current in either
direction.
The Fahy Simplex permeameter is not an
absolute instrument. It requires calibration by
reference to an absolute instrument which is
taken to be a standard; if it were not for the
inevitable magnetic leakage, the theory of the
instrument would be comparatively simple. The

//-coil is assumed to indicate the difference of

magnetic potential between the ends of the
//-blocks. These blocks then in effect transfer
the ends of the coil to the ends of the specimen.
On account of leakage, however, the actual
situation is not quite so simple. Experiment
has shown that if the actual value of area-turns
in the //-coil is used in calculating H, the value
thus obtained is not correct.
The single-yoke arrangement leads to a non
uniform distribution of induction both along the
length and across the section of the specimen.
Also, since the magnetizing coil is on the yoke
rather than on the specimen, some of the leakage
flux is linked with the //-coil. The reluctance of
the //-blocks should also be taken into consider
ation. In view of all of the potential sources of
error inherent in the design, it was somewhat
surprising to find that a simple correction pro
portional to the observed magnetizing force
would bring the results practically in line with
the results obtained with an absolute instrument
if uniform specimens are used in the intereomparison. This correction is most conveniently
applied by assigning to the //-coil a value of
what might be called the "effective" area-turns
which is somewhat different from the measured
value. It has been found that if the specimen
is laminated, the maximum number of strips for
which reasonable accuracy can be obtained is
ten. The error is apparently a function of the
number of air gaps between the pole pieces and
the //-blocks.
With so many factors involved, it is not feasible
to give a precise estimate of the accuracy ob
tained but in general, under normal conditions,
values of either B or H for normal induction may
reasonably be expected to be accurate within 2
percent. For the determination of coercive force,
another phenomenon must be taken into con
sideration. It has been found that when the
average induction as indicated by the /?-coil is
zero, the B in the part of the specimen next to
the pole pieces has already reached a negative
value while the outer part next to the //-blocks is

460-20

still positive. Since the i/rblocks are in contact

with this outer part, the value of Hc will be low.
The error depends upon the slope of the hys
teresis loop at the Hc point. If it were vertical,
there would be no error. As a rule, the slope is
less and consequently the error is greater the
higher the value of Hc.
The Fahy Simplex permeameter has found
favor primarily on account of its easy manipula
tion and the requirement of only a single specimen.
Its accuracy is probably within the range of
uniformity of most magnetic materials.
The wiring diagram is like that of figure 9
except that coil I and resistor sh are omitted and
the magnetizing coil M is on the yoke instead of
over the specimen.

5.6. Magnetic Susceptibility

The measurement of magnetic susceptibility re
quires the use of much more sensitive apparatus
than is needed for the measurement of ferromag
netic properties. Just as in the measurement of
magnetic fields, there are only two basic methods
for the determination of magnetic susceptibility.
One is the measurement of the minute mechanical
forces experienced by a body in a nonuniform
magnetic field. The other is by means of induc
tion methods. For precise measurements, the
force methods are more commonly used.
The principal methods in use by the Magnetic
Measurements Section are known as the Gouy
method [18], the Faraday method [19], and the
Thorpe and Senftle method [20].

5.5. Tests of Low-Permeability Materials

"AT
Materials having a permeability only slightly
in excess of unity are generally tested in a straight
solenoid. Since the intrinsic induction for such
materials is very low, the correction for selfdemagnetization is small and can be neglected
without appreciable error in the value of mag
netizing force calculated in terms of the current.
In order to obtain a satisfactory degree of pre
cision in the determination of the intrinsic induc
tion it is necessary to balance out the direct
effect of the magnetizing field. This can be done
by connecting the secondary of a mutual inductor
of variable value in series-opposition with the
test coil which usually has several hundred turns
The primary of the inductor is connected in
series with the solenoid. The mutual inductor is
adjusted so that, with no specimen in the testcoil, there is no deflection of the ballistic galvanomter upon reversal of the magnetizing current
The galvanometer can then be used at its maxi
mum sensitivity. When a specimen is inserted
within the test coil and the magnetizing current
is reversed, the galvanometer deflection is propor
tional to the intrinsic induction. The galvanom
eter is calibrated by means of a standard
mutual inductor in the usual way.
In the Fahy Low-Mu permeameter [17] a some
what different arrangement is employed. Two
similar test coils are used. The specimen is in
serted in one coil and the other constitutes a com
pensating coil to balance out the effect of the
magnetizing force on the main test coil. The
compensating coil has a value of area-turns some
what greater than that of the other coil. In order
to balance the system with no specimen present,
the compensating coil is shunted with a resistor
whose value is adjusted so that, with no specimen
in the test coil, no deflection of the galvanometer
results from a reversal of the magnetizing current.
Then when a specimen is inserted in the test coil
and the magnetizing current is reversed, the re
sulting deflection is proportional to the intrinsic
induction. The galvanometer is calibrated in the
usual way by means of a standard mutual inductor.

i
Figure 14. Gouy apparatus.
The Gouy apparatus is illustrated by the dia
gram of figure 14. A cylindrical specimen, s, of
uniform cross section, A, is hung from a sensitive
balance with its axis vertical. The lower end is
located in the uniform part of the magnetic field
between the poles, P, of a powerful electromagnet,
and usually above the center line of the pole
pieces. The specimen should be long enough to
bring its upper end into a much weaker field. The
force due to the application of the magnetic field
is measured by the balance. The magnetic sus
ceptibility is then calculated by using the equation
1=kA(Bl2-Bi2)l2T

22

Transposing,
2iy
AW-BJ)
" Koto: In the clnssical cescm system of units, the equation is usually elven
as /-t.4(Hi*Hi')!2. This equation is dtmensionally Inhomoiteneous but
Cives the same numerical results as the other because In this system rm is
unity and there is no numerical distinction between R and H.

461-21

cgsem

mksa
(rationalized)

dynes

newtons

rm=magnetic constant

A= area of specimen

cm2

4irX10"7
ma

1= field at lower end

gausses

teslas

B2= field at upper end

gausses

teslas.

standard samples of known susceptibility. The

standard must be of the same size and shape as
the specimen to be tested and located in the same
position in the nonuniform field.

&=volume susceptibility
/= force

The use of rationalized units yields values of

magnetic susceptibility which are 4jt times the
values obtained by the use of unrationalized
units. (See pp. 4-5). The Magnetic Measure
ments Section generally uses the nuclear magnetic
resonance method (see pp. 12-13) for measuring Bi
and a pivoted magnet instrument for measuring
B2. (In the classical cgsem system, it is customary
to measure B and call it H.) The Gouy apparatus
at NBS is equipped for measurements at tempera
tures as low as that of liquid hydrogen.
The Faraday method is adapted to measure
ments of magnetic susceptibility of specimens too
small for the Gouy method. A small sample of
magnetically isotropic material experiences no
mechanical force when placed in a uniform mag
netic field. However, if there is a gradient in the
field, a force will be experienced. In this case
0'

mxBdB/dy :
yTransposing,
mBdB/dy

Figure 15. Vertical beam Faraday apparatus.

where
cgsem

mksa
(rationalized)
(see pp. 5-6)

x=mass susceptibility
rm=magnetic constant

4*X10-7

dynes

newtons

rn = mass

grams

kilograms

B= magnetic field

gausses

teslas

cm

m.

/= mechanical force

y= distance

Here as with the Gouy method, the values in a

rationalized system are 4tt times those in an
unrationalized system.
Measurements with this method on an absolute
basis are attended with some difficulty. The
apparatus is generally calibrated by the use of
,J The force Is given as f=mxHdHldy when the classical cgsem system of
units Is used. This equation is dimensionally inhomogeneous hut gives rise
to no numerical error because r Is unity and no distinction is made between
B and H in the classical cgsem system.

The apparatus is shown diagrammatically in

figure 15. The specimen, S, is mounted at the
lower end of a quartz beam, and a permanent mag
net, M, is mounted near the upper end. The
beam is supported by a horizontal berylliumcopper wire near the midpoint; the specimen is
carefully positioned between the pole pieces in
the region where the product of the field by the
field gradient is constant over the volume of the
specimen. Measurements are made by applying
a field and then restoring the beam to its original
position by means of a torque exerted on the
permanent magnet by a current in the solenoid, u.
The position of the beam is indicated by observing
the tip, T, with a telescope and scale.
The Thorpe and Senftle method, described by
Thorpe and Senftle [20], is generally used for
measuring small samples of material. The method
is illustrated in the diagram of figure 16. A
quartz spring supports the specimen initially in a
position between the poles of the magnet and
slightly below the center of the pole faces. After
the deflection of the spring, due to the application

462-22

and
g = acceleration due to gravity,
_2gATm

The induction methods most generally used are

a-c bridge or balance methods 24 and the vibration
magnetometer.24 These methods involve the in
ducing of voltage in a coil as a result of changing
flux linkages due to the applied field or a position
change of the coil or sample. These methods
have the advantages that they use uniform mag
netic fields and cover a larger range of suscepti
bilities than the force methods; however they
usually require a standard material for calibration
purposes.
5.7. Magnetic Standards
Figure 16. Quartz Helix magnetic susceptibility balance.
of the field, has been observed for this position,
the magnet is lowered to a new position and
another observation of the deflection is made.
This procedure is repeated for several selected
positions of the magnet until there is no observ
able change in deflection when the magnet is
moved to a lower position. From the proper
integration of the curve of displacement of the
sample versus the distance of the magnet from
the sample, the maximum value of the magnetic
field between the poles of the magnet, the mass of
the sample, and the elastic constant of the quartz
spring, it is possible to calculate the magnetic mass
susceptibility of the sample by the equation
2AKYm

where
X = mass susceptibility
A = area under the curve
K= elastic- constant of the quartz spring
r= magnetic constant
to = mass of the sample
B = the maximum field of the magnet.

In view of the nature of the magnetic units, it is

obviously not feasible to realize them in concrete
form. The calibration of magnetic testing appara
tus is carried out in terms of the electrical units by
the use of standard shunts, potentiometers (or
ammeters), and mutual inductors. The basic
standards are (1) a mutual inductor so designed
that its value can be calculated from its dimensions
(2) a standard resistor, and (3) a standard cell.
These basic standards are used to calibrate work
ing standards and these working standards in turn
are used for the calibration of the measuring cir
cuits. For the intercomparison and standardiza
tion of permeameters it is necessary to make use
of carefully selected and prepared test specimens
whose magnetic properties are accurately deter
mined by some standard method. It may perhaps
be proper to refer to such test specimens as mag
netic standards.
The preparation and maintenance of standards
of magnetic susceptibility present some difficult
problems. The material must be of high purity
(ferromagnetic impurities are particularly trouble
some) and means must be provided to prevent
contamination during handling or storage. Ques
tions of stability and temperature effects are also
important. Standards of susceptibility are required
for the calibration of apparatus with which it is
difficult or impossible to obtain absolute results.
The Magnetic Measurements Section has a con
tinuing program for the investigation of standard
materials such as water, benzene, nickel chloride,
and hydrated ferrous ammonium sulfate (Mohr's
salt) and the improvement of methods of prepara
tions and measurement of standards specimens.

Since
5.8. Limits of Accuracy
m=Kh/g
where
h = measured static deflection of the quartz
spring

The problem of magnetic testing consists in

determining simultaneous values of magnetic
s* MagnctochemistryP. W. Selwood Intcrsdenoe Publishers.
a A Vibrating Sample Magnetometer by N. V. FrederickIRE Trans
action on Instrumentation 1-9, No. 2 (Sept. 1960).

463-23

induction and magnetizing force. It is relatively

easy to obtain fairly accurate values of induction,
but the accurate determination of the correspond
ing magnetizing force is more difficult. It is
only by the exercise of great care in the selection of
test specimens and manipulation of the testing
apparatus that an accuracy of 1 percent can be
attained. The influence of the quality and con
dition of the test specimen is of great importance,
especially in the standardization of permeameters
because inaccuracies really arising from the con
dition of the specimen itself should not be charged
to the testing apparatus.
5.9. Requirements of Standard Specimens
for D-C Permeameters

It is well known that specimens of steel freshly

heat-treated are not metallurgically stable; that
is, changes in internal structure or condition may
go on for some time. These changes are accom
panied by corresponding changes in magnetic
properties. It is necessary, therefore, to make
sure that specimens to be used as magnetic
standards arc metallurgically stable. This can
be accomplished by either natural or artificial
aging.
It is quite obvious that irregularity in cross
section along the length of a specimen would
have an effect similar to that of a variation in
magnetic permeability. For this reason, it is
important that care should be used in preparing
the specimen to maintain a uniform cross section.
5.10. General Precautions

Specimens to be used as standards for the

calibration or intercomparison of permeameters
should be chosen and prepared with the following
points in view: (1) magnetic uniformity along
the length, (2) metallurgical stability, and (3)
uniformity of section.
If the specimen varies in permeability along
its length, errors are introduced in the measure
ments which cannot be calculated or eliminated
by compensation, and which may be of consider
able magnitude. It is possible to have errors
due to this cause alone which amount to 25
percent or more. Moreover, various methods arc
sensitive to this influence in varying degrees.
It is obvious, therefore, that such specimens
should not be used for the intercomparison or
standardization of testing apparatus. Various
methods for the determination of the degree of
uniformity of magnetic-test specimens have been
proposed [21], but probably the most satisfactory
one is to prepare a specimen much longer than
is required for the final form and to make measure
ments at suitable intervals along its length. If
the results of these measurements arc in agree
ment, then the specimen is, from this point of
view, satisfactory to use as a standard.

6.

Tests With

In the calibration and use of magnetic standards,

it is necessary to avoid (1) mechanical strain, (2)
variations in temperature, and (3) mechanical
vibration.
Mechanical strain influences the magnetic
properties of materials to a marked degree. It is
important, therefore, in the calibration and use
of magnetic standards that they be clamped
without bending. The effect of bending is par
ticularly noticeable in materials of high permea
bility and in the steep part of the magnetization
curve.
The effect of variations in temperature is not
negligible [22], and care should be taken that
standards be not heated during the course of a
test. The temperature coefficient is not constant
and varies for different materials or even for the
same material with different heat treatments.
Mechanical vibration should be avoided in
magnetic testing. It has a tendency to increase
the apparent permeabilitv and to decrease the
hysteresis. This is generally not a serious factor,
but for work of high accuracy, it should be con
sidered and the specimens protected from ex
cessive vibration.

Alternating Currents

6.1. Core Loss and A-C Permeability at Power

Frequencies
Magnetic tests with alternating currents arc
extensively employed by producers and users of
magnetic core materials in connection with quality
control and for obtaining design data. These
tests are generally made by methods specified by
the American Society for Testing Materials.
Since magnetic materials are continually being
improved and new and better methods of testing
are being developed, the specifications require
frequent revision in which the Magnetic Measure
ments Section cooperates. For this reason, it is
neither necessary nor desirable to give detailed
descriptions of the current standard methods.

The general principles are presented here but for

specific details, the reader is referred to the latest
Standards on Magnetic Materials issued hy the
ASTM [23].
The principal objects of testing with alternating
currents are to determine core loss, a-c perme
ability, and incremental permeability of laminated
magnetic core materials. The test specimens
usually consist of strips of flat-rolled material
assembled in the form of a square magnetic
circuit inserted in four solenoids which are per
manently mounted on a suitable base. Each of
the solenoids has two uniformly distributed
windings. The corresponding windings of the
four coils are connected in series to form primary
and second windings having an equal number ol

464-24

turns. The primary winding carries the magnet

izing current. The secondary winding is used to
energize voltmeters and the voltage coil of a
wattmeter (if used). A third winding may be
required in certain types of tests. The set of
coils mounted on a base is generally called an
Epstein frame.
Since flat-rolled core material has directional
properties, half of the specimen strips are generally
cut with their long dimension in the direction of
rolling and half in a direction at right angles to
the direction of rolling. Certain materials with
oriented grain structure may have the strips all
cut in the direction of rolling. When the strips
are cut "half and half" they are so assembled that
strips on opposite sides of the square are cut in
the same direction.
There are two principal types of testing methods
in which the Epstein frame is used, voltmeterwattmeter methods, and bridge methods.
a. Voltmeter- Wattmeter Methods
The basic diagram of connections for these
methods is shown in figure 17. The a-c source
should deliver as nearly a sinusoidal voltage as
possible with a minimum amount of distortion
when fully loaded. The current is controlled
preferably by means of a suitable autotransformer
rather than by series reactance. Excessive re
sistance or inductance in the circuit causes an
undesirable amount of distortion in the wave
form. Connected in series with the primary
winding, P, of the test frame are: an ammeter, A,
of low internal impedance, the primary winding of
a calibrated mutual inductor, Lm, the current coils
of a wattmeter, W, designed specially for measure
ments at very low power factors and the primary
winding of a compensating mutual inductor, Lme
'if used).
The secondary winding of the compensating
inductor, Lmc (when used) is connected in series
opposition to the secondary winding, S, of the test
frame. When the selector switch Sw is closed
downward, the rms-reading voltmeter, rms Vm,
the average-reading voltmeter,26 Av Vm, and the
potential coil of the wattmeter, W, are connected
in parallel to the secondary winding S of the test
frame through the secondary winding of the com
pensating inductor, Lme. The 50-cm Epstein test
requires a specimen consisting of 10 kg (22 lb) of
material cut into strips 50 cm (19% in) long and
3 cm (Hit in.) wide. This relatively large sample
previously was required in order to obtain a fair
average result for material which was not very uni
form even throughout a single sheet. The strips
are assembled in the test frame with butt joints at
the corners, which are clamped to maintain a low
magnetic reluctance at the joints. Improved
manufacturing techniques now permit the use of
much smaller samples and the 25-cm Epstein test 27
requiring only 2 kg (4.4 lb) is now standard.
* This is an instrument whose moving system responds to the average
value
of toearerectified
alternating
current.
r Tests
often made
on smaller
specimens but the 2 kg sample is stand
ard at present.

rth

rth

rrti

rrv
Test Frame

Sw
RMS
Vm

Figure 17. Diagram of connections for Epstein tests by

the voltmeter-wattmeter methods at power frequencies.
In carrying out tests using the 25-cm Epstein
frame, a better joint at the corners than the butt
joint used in the 50-cm apparatus is required on
account of the shorter magnetic circuit. A doublelap joint has been adopted. This requires strips at
least 28 cm (llKs in.) long. Strips up to 50 cm
(19% in.) long can be used if necessary. The
width of the strips is the same as in the 50-cm test,
i.e., 3 cm (l%e in.).
The sample is first weighed and then inserted in
the test frame in four equal groups in such' a way
that at each corner the strips of adjacent groups
successively overlap and groups at opposite sides
of the square are cut in the same direction. No
insulation other than the natural oxide is used
between strips except in the case of oxide-free ma
terial. The corners may be clamped to reduce the
magnetic reluctance but it must be done carefully
so as not to introduce excessive mechanical strain
in the strips.
The apparatus is then connected to the a-c
source and the current is adjusted so that the
voltage of the secondary coil of the test frame as
indicated by the average-reading voltmeter cor
responds to the desired maximum induction in the
specimen. This voltage is given by the equation
4BmANf
10"
where
ve= average voltage
B =maximum induction in gausses
A = cross-sectional area of the specimen,
cm2
N = number of turns in secondary coil
/ = frequency in cycles per second
(If B is in teslas and a is in m2, the factor 108 is not
needed.)
The cross-sectional area of the specimen is deter
mined from the mass, density, and length of the
strips
. m
A~4U
where

465-25

A = area in cm2
m = the mass in grams
/ = length of a strip in cm
5 = density in g/cm3.

The density for most of the conventional mate

rials is assumed from their chemical composition.
For other materials, the density must be deter
mined experimentally. For silicon-iron alloys the
values range from 7.55 to 7.85 depending upon the
silicon content. (See table 4.) Densities of
nickel-iron alloys vary from 7.85 to 8.90. (See
table 5.)
Table 4. Silicon-iron alloys
Silicon content range

Assumed density

Percent
0 to 0.5
0.5to2.0
2. 0 to 3. 5
3. 5 to 5.0

glcm'
7.85
7. 75
7.65
7.55

in which
'0=average voltage due to air flux
rm=47rX10-?
i/=peak magnetizing force in amp-turns/in
and areas are in m2.
In the cgs system, the peak magnetizing force
is calculated by using the equation

"
where

Hp=peak value of magnetizing

force in oersteds
N= number of turns in pri
mary winding
7p=peak current, amperes
l= assumed length of the mag
netic circuit =94
cm.

Table 5. Iron-nickel alloys

Density determined from straight lines joining points given below.
Nickel

Density

Percent
0
30
50
80
100

glem1
7.85
8.00
8.26
8.64
8.90

After the frequency and voltage have been prop

erly adjusted, the wattmeter is read. The value
obtained includes not only the total loss in the
core but also the losses in the instruments con
nected to the secondary winding of the test frame.
The loss in each instrument is E'/R where E is the
rms voltage and R is the ohmic resistance of the
instrument.
If the wave-form distortion as indicated by the
form factor, iff, differs from 1.111 (form factor of a
sine wave) by more than 1 percent, it is necessary
to make a correction to the observed value of the
core loss to account for the fact that whereas the
hysteresis component of the total loss is a function
of the average voltage, the eddy-current compo
nent is a function of the rms voltage [24].
If the cross-sectional area of the specimen is
much less than that of the secondary coil of the
test frame, a correction must be made for the extra
induced voltage due to the air flux. This correc
tion can be calculated from the equation
p ANjYMa-A)
a~
108
where
Ea= average voltage due to air flux
N= number of turns in secondary coil
/= frequency in cycles per second
rm=magnetic constant=l
a = area of secondary coil in cm2
A= area of specimen in cm2
ffp=peak magnetizing force in oersteds.
If the mksa system of units is employed, the equa
tion becomes
Ea=4NfTjff,(a-A)

OAtNI,
la

In the mksa system, H, is in ampere-turns/

meter and la is in meters (0.94 meter) and the
equation is
NT,
H,=l.
The peak value of the magnetizing force, H, is
determined in terms of the peak current and the
number of turns per unit length in the magnetizing
winding, P. The peak current is proportional to
the average voltage induced in the secondary
winding of the mutual inductor, Lm
I,=KEm.
The factor K is determined by observing the
average voltage induced by a sinusoidal current
of known value.
The a-c permeability is taken to be the ratio of
the maximum induction to the peak value of
magnetizing force.
It is convenient to compensate for the extra vol
tage due to air flux by means of the compensating
mutual inductor (Lmc in the diagram) . The mutual
inductance is adjusted to be equal to that between
the primary and secondary windings of the test
frame (no specimen inserted). If this is done, the
indicated induction is the intrinsic induction, Bt,
in the specimen.
The specific core loss is calculated bv dividing
the corrected value of total watts by the "active
weight" of the specimen. The active weight is
mli
47
where
m:=active weight in kg
m= weight in kg
t*i=assumed length of magnetic circuit=94 cm
1= length of strips.

466-26

b. Bridge Methods
Generally, core loss and a-c permeability are
determined by the voltmeter-wattmeter method.
In many cases, however, there are certain advan
tages, such as increased sensitivity and greater
frequency range, in making these measurements
by a bridge method. For measurements at low
inductions (not over 1,000 gausses or 0.10 tesla)
and at high frequencies, the bridge methods have
been used successfully for several years. More
recently, however, bridge circuits have been de
veloped by which measurements may be made at
higher inductions. The ASTM [23] has approved
as standard a modified Hay bridge for use in
measurements at inductions up to 10 or 12 kilogausses in nonoriented steels or 16 to 18 kilogausses
in oriented silicon steels. For measurement at
higher inductions, the distortion in the wave form
of the exciting currents is considerable and it has
been found necessary to use special techniques
[25] in order to obtain accurate results with bridge
methods.
Figure 18 is a diagram of a modified Hay bridge
currently approved as standard by the ASTM
[23]. The assumed equivalent impedance of the
coil containing the test specimen is shown in the
diagram at the right. The inductance of the coil
alone without the specimen, Lw, and its resistance,
Ra, are considered to be in series as shown. They
are balanced by the capacitance Ci and the resis
tance 2? respectively. Ordinarily the inductance
of the coil alone is so small that it is negligible and
C'a need not be used.
The apparent additional resistance, Ru due to
the core loss and the inductance, Lu due to the

permeability of the test material are considered

to be in parallel. They are balanced by the
resistance,
and the capacitance C\. Ra and Rc
are fixed resistors whose product is equal to the
product of the impedances of the other two arms
of the bridge. Since Rc carries the total exciting
current it must have such current-carrying capac
ity that it will not be heated excessively.
D is a detector, tuned to the fundamental fre
quency of the voltage supply. This detector
should have a high impedance and high sensitivity
to the fundamental frequency. Previously, vibra
tion galvanometers have been employed for this
purpose, but they have been almost entirely
replaced by electronic devices of various kinds.
The a-c source is connected to the bridge
through an isolating transformer. In the arrange
ment recommended by the ASTM, this trans
former has a tapped secondary for convenience in
selecting the proper range of voltage to be applied.
The primary is fed from an autotransformer con
nected to the generator. This provides the fine
control of the applied voltage. Other arrange
ments may be employed provided that there is
isolation between the source and the bridge and
the series resistance is not used to control the
current. Resistance tends to distort the wave
form of the induced voltage and may lead to
excessive errors. [25]
For incremental values, the d-c magnetizing
current may be supplied to the secondary winding,
N2. The average a-c voltage is then read across
the primary winding, Nt. The IR drop in this
winding causes a slight error but this is usually
negligible. The d-c source must furnish a steady
current. A storage battery is ordinarily the best

Equivalent Impedance
of Test Frame
Figure 18. Modified Hay Bridge.
467-27

source as rectified a-c requires a great deal of

filtering. The d-c circuit contains a series inductor
of at least 10 h when carrying maximum current.
This is to limit the a-c current in the d-c circuit
to a negligible value.
The maximum induction in the specimen is
indicated by a voltmeter connected to the second
ary winding N2 of the test frame. The voltmeter,
Fave, indicates the average voltage.28 The maxi
mum induction in terms of average voltage in cgs
units is

cases. Fortunately, it has been found that a

satisfactory degree of agreement can be obtained
if balance is reached about 3 min after demagneti
zation. If the test induction is higher than 1,000
gausses (0.100 teslas) the drift is not observed.
Rw and Cw (if used) are preset according to the
equations

and
n'
Cw~rjtc

(C in farads) .

"ave=4 BANJX10-*
where
^ve= average volts
B= maximum induction, gausses (if the air-flux
compensating inductor is used, B is the
intrinsic induction, Bt)
-A = area of specimen, cm2
N= number of turns in winding
/= frequency, cycles per second.
In mksa units,
B= induction in teslas
A = area in m2
and the factor 10~8 is omitted. The other factors
remain the same.
The rms value of the exciting current can be
determined by connecting the rms-reading volt
meter across the resistor Rc as indicated.
As in d-c testing, the first step, especially if
measurements are to be made at low values of
induction, is to demagnetize the specimen. This
is accomplished by first setting switch S2 so as to
take Rc out of the circuit and opening switch S3 and
then with switch S4 closed, setting the a-c source
to zero and connecting it to the bridge. The
voltage is then raised until the average voltmeter
indicates that a maximum induction of the order
of 14 kilogausses has been reached. The current
is then gradually and steadily decreased until the
indicated induction is somewhat lower than the
lowest at which a measurement is to be made. If
this point is below about 1,000 gausses (0.100
teslas) it will be necessary to take into account the
drift in values which takes place immediately
after demagnetization. This drift is rapid during
the first few seconds but the rate of change
decreases as time goes on. For this reason, if
reproducible results are to be obtained, a standard
procedure must be followed. For the best
reproducibility, it is necessary to wait several
hours after demagnetization before making the
test. This would be very inconvenient in most
11 According to ASTM specifications, the scale of this instrument is cali
brated in terms of average volts multiplied by 1.111, the form factor of a pure
sine wave. This may be confusing in some instances, especially when an
rms-reading voltmeter and average-reading voltmeter are used together to
determine the form factor.

These balance the resistance, Rw, and the induct

ance, Lw, of the test winding, N (specimen not
inserted), thus avoiding the necessity of calculating
the effect of Rw and Lw.
Since Re must carry the total exciting current
and the voltage drop across it must be less than
10 percent of that across the test winding to avoid
excessive distortion of the wave form, its proper
value will depend upon the range of inductions
at which measurements are to be made. Ordi
narily, 1,000 ohms is used for the lowest range
(10 to 50 gausses), 100 ohms is used for the
medium range (50 to 500 gausses) and 10 ohms
for the higher range (500 to 10,000 gausses). The
resistor Rtt may be set so as to make permeability
direct-reading in terms of Ch or core loss may be
direct-reading in terms of Rb. Directions for
setting may be found in the ASTM specifications.
Readings are made as follows. With switch ^
set to connect Rc into the circuit and switches
Si, S3, and St closed, the applied voltage is raised
to the point at which the desired induction as
indicated by the average-reading voltmeter is
reached. Balance is then obtained by successively
adjusting Rb and Cb until the detector shows a
minimum reading. The balance equations are
as follows
I, = Wt
and
7,

RaRc

where
L, = inductance due to the permeability of the
test material, henries;
and
/?i=increased apparent resistance due to core loss
in the test material, ohms.
In the cgs system
Ll~
and since

468-28

OAtNM^
ion, XMt

it follows that

where
7i-e=d-c current, amperes

W^iXio1
OAttNM

and

In these equations
HL = effective a-c permeability of the test
material
li = assumed effective length of the magnetic
circuit=94 cm
Ny = number of turns in the primary winding.
In the mksa system, the equation is

where
Zx = assumed effective length in meters=0.94
and
.A=area in square meters.
Areas are determined from the length, mass, and
density of the strips.
Core loss, Pe, is calculated from the observed
values of Rl and 2?,
c

RaRc

The specific core loss PB.; is

m,
where m.\ is the active mass as indicated above.
Reactive or quadrature power is
E?
the unit is called the var. The specific value is
Pt/mi where mi is the effective mass, determined
as shown above.
Incremental values of permeability and core
loss are determined in the order of increasing
values of biasing magnetizing force. Direct
current is supplied to the secondary winding and
adjusted to the required value and reversed several
times to establish a cyclic condition in the speci
men. The a-c is then applied and the bridge is
adjusted and readings taken as described above.
The d-c magnetizing force is calculated as
follows :
In the cgs system, the biasing or d-c magnetizing
force is

li is in meters
H is then ampere-turns per meter.
For incremental tests,
meter is connected to
For this test values of
are limited to not more
tesla).

the average-reading volt

the primary winding Ni.
incremental induction, B&
than 1,000 gausses (0.1000

6.2. A-C Measurements at Higher Than Power

Frequencies
Magnetic materials have been developed which
are particularly useful at high frequencies on
account of their extremely high electrical resist
ance. They are ceramic materials called ferrites.
Magnetic tests of these materials are usually made
on specimens in the form of small rings. Although
the Hay bridge can be adapted to make measure
ments at the higher frequencies, it has the dis
advantage that each ring must be wound
individually. In order to obviate this difficulty,
Haas [26] has developed a radiofrequency
permeameter similar in principle to an instrument
developed earlier for power frequencies by Kelsall
[27]. It is shown diagrammatically in figure 19.
The radiofrequency permeameter consists es
sentially of a transformer. The primary is wound
on a powdered ring core. The secondary consists
of a central conducting tube over which the ring
specimen can be placed. The one-turn secondary
is completed by a coaxial outside cylinder and end
plates which form a single short-circuited turn
which is linked with both the primary core and the
specimen as shown in the equivalent circuit. The
top plate is removable for insertion of the test
specimen. Measurements are made of the input
Cover Plate

where
Hd= biasing magnetizing force, oersteds
Id.c =d-c current, amperes
/! = assumed length of the magnetic circuit, cm.
In the mksa system

Specimen
Primary Core

Primary Winding

Primary Core
Primary Winding
Single Turn Secondary
Cover Plate
Specimen

N2Id.c
Figure 19. Radio frequency permeameter.
469-29

impedance of the primary winding with no

specimen and the secondary open-circuited, with
the secondary circuit closed but no specimen, and
with the secondary circuit closed and a specimen
in place. These measurements can be made with

a (?-meter or an impedance bridge. From the

values thus obtained, initial values of complex
permeability and losses can be computed. Details
of the measurements and computations can be
obtained from the original paper [26].

7. Typical Magnetic Properties of Materials

Magnetic materials are used under a great
variety of conditions both for cores and for per
manent magnets. In order to meet many of these
special requirements, it is frequently necessary to
"tailor make" the material. This may require
"special" alloying elements, heat-treating tem
peratures and procedure, and special mechanical
working. In general, however, the metallic mag
netic materials employed in practice are alloys
mainly composed 01 one or more of the ferromag
netic metals, iron, nickel, and cobalt. Other
elements are always present, either as undesired
impurities or as alloying elements added for the
purpose of producing certain desired character
istics. In figure 20 are shown typical normal
induction curves of annealed samples of iron,
nickel, and cobalt of comparatively high purity.
These curves are given only for the purpose of
general comparison and should not be considered
as representing critical values. Small variations
in the degree of purity or in the annealing pro
cedure lead to substantial differences in normal
induction.
Magnetic materials employed in commercial
practice may be considered under the following
six classifications: (1) solid core materials; (2)
electrical sheet or strip; (3) special alloys; (4)
ferrites; (5) permanent-magnet materials; and
(6) feebly magnetic materials. In view of the
rapid and continued progress in the development
of magnetic materials, it is only feasible to give
data characteristic of several of the materials
commercially available at present. These data
indicate merely the range of properties obtainable
20

10 15

CO

Nl

ID
5

in the various classes of material and should not

be considered as critical, since considerable varia
tion from these values will be found in practice.
7.1. Solid Core Materials
These materials are used for the cores of directcurrent electromagnets, relays, field frames of
d-c machines, etc. The principal requirement is
high permeability particularly at relatively high
inductions. For the majority of uses, it is desir
able that the coercive force and hysteresis be
low. The principal materials employed are soft
iron, low-carbon steel, cast iron, and an alloy of
approximately 49 percent of cobalt, 49 percent
of iron, and 2 percent of vanadium known as
Permendur or Supermendur. Permendur is char
acterized by very high permeability in the upper
part of the normal induction curve and a saturation
induction approximately 10 percent greater than
that of pure iron. Its cost is relatively high,
however, and its use is limited in general to pole
tips in which a very high induction is required.
Several varieties of soft iron are available, such as
Armco iron, Norway iron, and Swedish charcoal
iron. These irons are especially refined to reduce
impurities and to make as pure iron as is com
mercially feasible. A typical composition is 99.91
percent of iron, 0.02 percent of carbon, with small
percentages of manganese, phosphorous, and
sulfur. Low-carbon steel should not have more
than 0.0 to 0.2 percent carbon and should contain
only the usual small amounts of the ordinary
impurities. Cast iron has a relatively low per
meability and is used principally in field frames
when cost is of primary importance and extra
weight is not objectionable. Cast iron is high in
carbon (about 3%) and also contains about 3
percent of silicon and varying percentages of
phosphorous, manganese, and sulfur.
The best magnetic properties for these materials
are obtained by a suitable annealing treatment
after machining and fabrication processes have
been completed. The properties of cast iron can
be greatly improved by malleabilizing, a process
that converts a large part of the carbon to the
amorphous form.
Typical normal induction curves for solid core
materials are given in figure 21.
7.2. Electrical Sheet and Strip

0
0

50

100
150
200
H, OERSTEDS
Figure 20. Typical norml induction curves of annealed
samples of iron, nickel and cobalt.

The term electrical sheet (or strip) is com

mercially used to designate iron-silicon alloys
produced in sheet or strip form and used as core

470-30

25
F-(;o-v
20

SOFT 1 BOM______

in 15
c/7

'

CAST STEEL

MALLEI BLE
CAST 1 RON

<
" 'CAST 1 RON
10
20
OERSTEDS
Figure 22. Typical normal induction curves for electrical
sheet.
150
100
200
H, OERSTEDS
Figure 21. Typical normal induction curves for solid core
materials.
SO

materials in alternating-current apparatus such

as transformers, motors, electromagnets, or relays.
The principal requirements are high permeability,
low hysteresis, and high electrical resistivity. The
several grades differ mainly with respect to their
silicon content, which ranges from 0.5 percent to
approximately 5 percent. Alloys containing the
higher percentages of silicon are practically nonaging; that is, the permeability and losses do not
change appreciably with time. The required
magnetic properties are produced by annealing.
By a suitable combination of cold-rolling and
heat treatment, electrical sheet may be produced
in which the majority of the crystals are given a
favorable orientation. Such material has consid
erably better magnetic properties in the preferred
direction than ordinary grades, their maximum
permeability is approximately twice as high and
they have much lower core losses combined with
higher permeability at high induction than
ordinary grades. Figure 22 shows typical normal
induction curves for two grades of electrical sheet
and oriented-grain material. The improvement
in the oriented-grain material is particularly
conspicuous in the upper part of the normal
induction curve.
The different grades of electrical sheet and strip
are usually sold on the basis of guaranteed maxi
mum values of total core loss, as determined in
accordance with the specifications of the American
Society for Testing Materials [23]. The common
designations of the various grades are armature,
electrical, motor, dynamo, and transformer. The
transformer grades are further subdivided into
classes denoted by numerals corresponding to the
core loss under standard conditions. Armature,
electrical, and motor grades are used principally
in small motors, a-c magnets and starting trans315-411 O - 68 - 31

formers. The dynamo grade is used in highefficiency rotating machines and small trans
formers. The transformer grades are used in
power and radio transformers.
7.3. Special-Purpose Materials
For certain applications, special alloys and other
materials have been developed which, after proper
fabrication and heat treatment, have superior
properties in certain ranges of magnetization. For
instance, alloys of nickel and iron with possible
small percentages of molybdenum or chromium
have very high values of initial and maximum
permeability. Alloys of this class which may
have from 70 to 80 percent of nickel are called
Permalloys. Special attention to the purity of
the constituents, the manufacture, fabrication, and
heat treatment has resulted in an alloy of nickel,
iron, molybdenum, and manganese that has a
maximum permeability greater than one million.
This alloy when commercially prepared and rolled
into thin tapes (0.001 to 0.004 in. thick) has a
maximum d-c permeability between 300,000 and
900,000. An alloy of 50 percent nickel and 50
percent iron is called Hipernik. Another alloy
having a small percentage of copper in its compo
sition is called Mumetal. The characteristics of
these alloys differ in detail but in general they have
high initial and maximum permeability, low
hysteresis, and low saturation values. The alloy,
Supermendur, of 49 percent of iron, 49 percent of
cobalt and 2 percent of vanadium has high
permeability which persists at higher values of
induction than the nickel-iron alloys.
Typical permeability curves for several specialpurpose alloys are given in figure 23 and figure 24.
A certain alloy of nickel, cobalt, and iron after
suitable heat treatment has very nearly constant
permeability for inductions below 1,000 gausses
and is called Perminvar. The 50-50- nickel- iron

471-31

125000

\ PE RMALLOY
t 75000
2
* 50000
HIPERNIC
25000
\MUMt
0

2500

5000
7500
10000
B, GAUSSES
Figure 23. Typical permeability curves for special purpose
alloys I.
alloy can also be heat-treated so as to have similar
characteristics.
Many of the above described alloys are rolled
into thin tape (as thin as 0.0001 in.) and spirally
wound cores are prepared from the tape. This
permits the designer to make use of their direc
tional properties as well as to use them at much
higher frequencies by decreasing eddy-current
effects.
Another series of magnetic alloys of copper,
nickel, and iron is temperature-sensitive having an
approximately linear relation between permeabil
ity and temperature. These are called Thermalloys. The principal use is in the compensation of
watt-hour meters for temperature variations.
They are also used in certain types of thermal
relays.
Although chemical composition and impurities
are very important in the preparation of a mag
netic alloy, experience has shown that with the
same chemical composition a wide variety of
magnetic properties can be obtained by varying
the mechanical working and type of heat treat
ment. This is very striking in the nickel-iron and
iron-silicon alloys and has led to the development
of materials having a practically rectangular
hysteresis loop. These are of great importance in
the field of electronics.
7.4. Ferrites
Fcrrites are nonmetallic ceramic materials
whose extraordinarily high electrical resistivity is
especially useful for high-frequency applications.
They are finding increasing use in applications
such as electronic computers, antenna rods,
isolators, and magneto-strictive devices. They
are made of iron oxide combined with certain bi-

Figure 24. Typical permeability carves for special purpose

alloys II.
valent oxides, hydroxides, or carbonates of metals
such as manganese, cobalt, nickel, copper, zinc, or
magnesium. The process of manufacture is sim
ilar to that of other high-grade ceramic materials.
The powdered materials, which must be of high
purity and proper particle size, are mixed in
suitable proportions, pressed or extruded to the
desired shapes, and fired. This procedure produces
chemical compounds of the metallic oxides.
These materials are often called "soft" ferrites
to distinguish them from certain ceramic materials
used for the manufacture of permanent magnets.
In common with other ceramics, they are mechan
ically hard and brittle and require special tools for
cutting or grinding. Also, since they require no
critical materials, they are relatively low in cost.
Although the saturation induction of the soft
ferrites is not high, this is not important because
they are generally used at low inductions. Their
relative initial permeabilities (100 to 1,500) are
very good for the high frequencies involved. The
principal disadvantage of the ferrites is their low
Curie points which range from 100 to 500 C.
This is important if they are to operate at temper
atures much higher than room temperature.
Typical magnetic properties of some ferrites at
room temperature are shown in figure 25. In
figure 26 is shown the temperature dependence of
magnesium-manganese ferrite.

472-32

M , oerstedf
Figure 25. Typical normal induction curves for "soft"
ferrites.

10
15
20
H, oersteds
Figure 26. Temperature dependence of a magnesium-man
ganese ferrite.

7.5. Permanent-Magnet Materials

I, the Japanese metallurgist Honda announced the

development of a new permanent-magnet steel
having cobalt as the principal alloying element,
which he named KS magnet steel. This steel also
contained substantial percentages of tungsten,
chromium, and molybdenum. The coercive force
of this material is about four times as great as that
of the tungsten or chromium magnet steels. This
development stimulated further investigations
which led to the commercial production of several
types of cobalt magnet steel having various per
centages of cobalt. High-cobalt magnet steels
have from 35 to 41 percent of cobalt, and lowcobalt steels have from 8 to 9 percent. These
steels also have tungsten and chromium in addi
tion to the cobalt. An intermediate cobalt steel
has 17 percent of cobalt and about 9 percent of
chromium. Tungsten is sometimes substituted
for part of the chromium.
The quench-hardened alloys can be forged and
machined from the ingot.

Although it is possible to make permanent

magnets of almost any kind of steel that is capable
of being hardened by heat treatment, it is best to
use materials especially produced for the purpose.
Magnets made from other types of material are
likely to be inferior in quality or unnecessarily
expensive. Before the development of the special
magnet steels, magnets were generally made of
plain high-carbon tool steel. This type of steel
is relatively inexpensive, but its magnetic prop
erties are so greatly inferior to those of the special
steels that now it is practically never used for
making permanent magnets.
Permanent-magnet materials may be grouped
in five classes as follows: (a) Quench-hardened
alloys; (b) precipitation-hardened alloys; (c)
work-hardened materials; (d) ceramics; and (e)
iron-powder compacts. Figure 27 shows typical
demagnetization curves for several permanentmagnet materials.

b. Precipitation-Hardening Alloys
a. Quench-Hardened Alloys
Tungsten, chromium, and cobalt magnet steels
have been in use for many years. Prior to the
first world war, tungsten steel was the standard
high-grade permanent-magnet material. The
optimum tungsten content is between 5 and 6 per
cent, with about 0.6 percent of carbon. During
World War I, when tungsten was scarce, chromium
magnet steel was developed and extensively used
in this country. There are two principal grades,
one containing about 1 percent of chromium and
0.6 percent of carbon, and the other containing
about 3.5 percent of chromium and 0.9 percent of
carbon. The 3.5-percent chromium steel is used
in many applications. Short I3' after World War

The most important permanent-magnet ma

terials are of the precipitation-hardening variety.
This group includes the Alnico, Cunife, Cunico,
and Vicalloy. The process of hardening in these
alloys is related to the change from a single phase
at high temperature into two new phases when the
temperature is lowered beyond a certain value.
The fact that high values of coercive force can be
obtained with alloys containing no carbon was
anounced in 1932 by Seljesater and Rogers [28]
in the United States, Koster [29] in Germany, and
Mishima [30] in Japan. These precipitationhardening permanent-magnet alloys contained
aluminum, nickel, and iron. Although the resid
ual induction was relatively low, the coercive forec

473-33

Figure 27. a jYjpieai demagnetization curves for permanent

magnet materials.

-700

-600

was so high, about 475 oersteds, that new applica

tions of permanent magnets were made possible.
Further investigation showed that the addition of
cobalt to the aluminum, nickel, and iron could
increase both the residual induction and coercive
force. There are more than twenty different com
mercial varieties called Alnico having coercive
force values ranging from 400 to 1,100 oersteds
and maximum energy products from 1.2X10" to
approximately 7X108 gauss-oersteds. The Al-NiCo-Fe alloys are very hard and brittle, so that they
cannot be formed by the usual methods of forging
and machining but must be either cast or molded
in powder form and then sintered. Final finishing
is done by grinding.
Other alloys that belong to the precipitationhardening group, however, are unusual for high

coercive force permanent-magnet material in that

they are so ductile that they can be formed readily
by rolling or drawing and can be drawn to fine
wires before the final heat treatment [31]. Cunife,
Cunico, Silmanal, and Vicalloy are alloys of this
type.
The platinum alloys have very high values of
coercive force but their low residual induction and
high cost limit their usefulness.
c. Work-Hardened Alloys
Several ordinarily "nonmagnetic" alloys of iron
may become ferromagnetic after cold working due
to a phase change in the material. Stainless steel
(18% chromium, 8% nickel) is "nonmagnetic" at
room temperature after being rapidly cooled from
1,200 C in the usual process of manufacture.
However, if it is hardened by cold working such as

474-34

drawing through a reducing die it may develop

properties such that it makes an acceptable per
manent-magnet material at room temperature. If
this work-hardened alloy is then reheated to a
high temperature and cooled slowly it regains its
original nonmagnetic condition at room tempera
ture. Another alloy that shows this property
contains 45 percent iron, 15 percent Ni, and 40
percent Cu. Nesbitt [32] has measured a co
ercive force of 240 oersteds and a residual induction
of 4,400 gausses in wire of this composition which
after quenching from 1,000 C, was then cold
drawn from 0.026 to 0.006 in. Increasing the
percentage of iron to 60 percent, decreasing the
percentage of copper to 25 percent, with 15 per
cent nickel produced an alloy that after similar
treatment as above resulted in a coercive force of
170 oersteds and residual induction of 11,000
gausses.
d. Ceramic Magnet Materials
A commercial development in permanent-mag
net materials which is increasing in importance
each year is the barium ferrite or ceramic perma
nent-magnet material. This is a chemical com
pound and has mechanical characteristics similar
to other ceramics. They are hard, brittle, have
a lower density than metals, and extremely high
electrical resistivity. The basic ingredients are
barium carbonate and iron oxide. These ma
terials in powdered form are compressed in dies
under high pressure to the required shape. This
compacted material is then sintered at a high
temperature. This process produces a material
which has an Hc of approximately 2,000 oersteds,
a B, of approximately 2,000 gausses, and a de
magnetization curve which is practically a straight
line. Further improvements in ceramic materials
have resulted in a highly-oriented barium-iron
oxide whose magnetic properties, on a weight basis,
are almost equal to those of Alnico V. The co
ercive force is approximately 2,000 oersteds, the
residual induction 4,000 gausses, and the energy
product is 3.5 times that of the unoriented variety.
At right angles to the direction of grain orienta
tion, however, this material exhibits negligible
permanent-magnet properties and has a perme
ability of only approximately 1.0.
e. Powder Magnets
Although pure iron is usually regarded as a
high-permeability or "magnetically soft" material
yet theory has predicted and experiments have
proved that compacts of pure iron powders may
produce very good permanent magnets. Powder
magnets have been produced of iron and iron

alloys (such as 70% iron and 30% cobalt) (Permct)

with particle size of about 10_s cm diameter with
Hc up to 500 oersteds and energy product of
1.5X106 gauss-oersteds. The permanent-magnet
properties result from the discrete small particles
of a single phase instead of from the presence of
two or more phases as in most other metallic
permanent-magnet material. Further experimen
tal work with particle size and shape and processes
of manufacture have produced, in the laboratory,
magnets with energy products comparable to
those of Alnico V and theoretical considerations
predict even higher values.
Manganese-bismuth permanent magnets also
belong to this group. This material is an aniso
tropic aggregate of crystals of the intermetallic
compound manganese bismuthide (Mn Bi) and is
a product of powder metallurgy. Manganese
bismuthide is prepared from the chemical action
between molten bismuth and powdered manganese
when heated to approximately 700 C in an inert
atmosphere of argon or helium. Cooling is ac
complished in such a manner as to produce
crystallization of the compound. Laboratoryproduced material may have a residual induction
of approximately 4,800 gausses and a coercive
force of 3,600 oersteds with energy product values
as high as 5X106 gauss-oersteds.
Powder metallurgy has also produced sintered
Alnico magnets. These magnets have greater
mechanical strength and more uniform magnetic
properties than the cast variety at the expense of
a slight decrease in the magnetic properties.
Magnet materials prepared from metal oxides
such as cobalt ferrites and Vectolite have been
made and used for many years; however, they
have been practically superseded by the barium
ferrites.
7.6. Feebly Magnetic Materials
It is occasionally desirable to use metallic
materials for tools or parts of equipment which
require practically nonmagnetic materials. Oxy
gen-free, high-conductivity
copper,
copperberyllium alloy, and pure aluminum are often
used where mechanical strength is not required.
Aluminum-bronze, nickel-copper, stainless steels,
and manganese steels can be used if a perme
ability not exceeding 1.10 is permissible. In
using such a material, it should be kept in mind
that even when the bulk material is sufficiently
nonmagnetic, mechanical working, surface con
tamination or temperature effects may make it
unsatisfactory.

8. References
[1] Gray, Absolute measurements in electricity and mag
netism, 2d Ed. p. 535, MacMillan Co., London
(1921).
[2] George S. Smith, Use of bismuth bridge magnetic
fluxmeter for ac fields, Trans AIEE 58, p. 52 (1939).
George S. Smith, A new magnetic fluxmeter, Trans
AIEE 56, p. 441 (1937).

[3] A. Koepsel, Apparat zur bestimmung der magnetischen eigenschaften des eisens in absoluten mass
und directer ablesung, ETZ. 15, p. 214 (Apr. 12,
1894).
C. W. Burrows, An experimental study of the koepsel
permeameter, BS Sci. Pap. 11, 101 (1914) S228.

475-35

4] R. B. Treat and J. W. Esterline, A new magnetic

testing apparatus, Elec. World 30, 696 (1897).
[5] H. E. J. G. Du Bois, Magnetic circuits and their
measurements, Electrician 27, 634 (1891).
[6] E. Hughes, A magnetic bridge for testing straight
specimens, Proc. Phys. Soc. (London) 37, 233
(1925).
[7] R. V. Picou, Permeametre universal, Bui. Soc. Int.
des Elec. [2] 2, 828, (1902).
[8] A Iliovici, Sur un nouveau permeametre universal,
Bui. Soc. Int. des. Elec. [3] 3, 581 (1913).
[9] Y. Niwa, A null method for testing magnetic prop
erties of materials, Researches Electroteeh. Lab.
(Tokyo) No. 142 (1924).
[10] C. W. Burrows, The determination of the magnetic
induction in straight bars, Bui. BS 6, 31 (1909)
S117.
[11] R. L. Sanford and P. H. Winter, A permeameter for
magnetic testing at magnetizing forces up to 300
cersteds, J. Research NBS, 45, (July, 1950)
RP2109.
[12] F. P. Fahy, A permeameter for general magnetic
analysis, Chem. and Met. Eng., 19, 339 (1918).
R. L. Sanford, Performance of the Fahv simplex per
meameter, BS J. Research 4, 703 (1930) RP174.
[13] R. L. Sanford and E. G. Bennett, An apparatus for
magnetic testing at magnetizing forces up to
5,000 oersteds, J. Research NBS 23, 415 (1939)
RP1242.
14] C. R. Mann, Phys. Rev. 3, 359 (1896).
15] H. Du Bois, Wied Ann. 7, 942 (1902).
16] C. L. B. Shuddemagen, Proc. Am. Acad. Arts Sci. 43,
185, (1907).
[17] The Fahy low-mu permeameter for measuring low
permeability, Iron Age 130, 330 (1932).
[18] L. G. Gouy, Compt. rend. 109, 935 (1889).
[19] M. Faraday, Experimental researches Vol. Ill,
Taylor and Francis, London, pp. 27 and 497 (1855).
[20] Arthur Thorpe and Frank E. Senftle, Absolute method
of measuring magnetic susceptibility, Rev. Sci. Inst.
30, No. 11, 1006-1008, (Nov. 1959).

[21] R. L. Sanford, Determination of the degree of uni

formity of bars for magnetic standards, Bui. BS
14, 1 (1916) S295.
[22] R. L. Sanford, Temperature coefficient of magnetic
permeability within the working range, Bui. BS 12,
1 (1915) S245.
[23] See latest edition ASTM Standards MetalsTest
Methods.
[24] G. Camilli, A flux voltmeter for magnetic tests, Trans
actions AIEE XLV, pp. 721-728 (1926).
[25] I. L. Cooter and W. P. Harris, Investigation of an
alternating-current bridge for the measurement of
core losses in ferromagnetic materials at high flux
densities, J. Research NBS 57, No. 2, (Aug. 1956)
RP 2699.
W. P. Harris and I. L. Cooter, Improved bridge
method for the measurement of core losses in ferro
magnetic materials at high flux densities, J. Re
search NBS 00, No. 5, (May, 1958) RP2865.
[26] Peter H. Haas, a radio-frequency permeameter, J.
Research NBS 51, No. 5, (Nov. 1953) RP2454.
[27] G. A. Kelsall, Permeameter for alternating-current
measurements at small magnetizing forces:, J. Opt.
Soc. Am. and Rev. Sci. Instr. 8 (Feb., 1924).
[28] K. S. Seljesater and B. A. Rogers, Magnetic and me
chanical hardness of dispersion-hardened iron allovs,
Trans. Am. Soc. Steel Treating 19, 553 (1932).
[29] W. Koster, Dauermagnetwerkstoffe auf der grundlage
der auscheidungs hartung, Stahi U. Eisen 53, 849
(1933).
[30] Mishima, Magnetic properties of iron-nickel-alumi
num allovs, Ohm (July, 1932) abstracted in Iron
Age 130, "346 (1932).
[31] I. L. Cooter and R. E. Mundy, Cunife wire magnets
of small size, J. Research NBS 59, No. 6, (Dec.
1957) RP2808.
[32] Richard M. Bozorth, Ferromagnetism, D. Van
Nostrand Co., Inc. p. 401 (1951).

476-36

INSTRUMENT SOCIETY of AMERICA

PREPRINT^ 20th Annual ISA Conference and Exhibit October 4-7, IMS, tot Anfalai
to mmUitk wriu ISA. 530 Wmx. fmmt *mm. Ktaon*. Pa. UIII
THE CALIBRATION OF PERMANENT MAGHBT STANDARDS *
by Irvln L. Oooter
Chief, Magnetic Measuremerits Section
Electricity Division
National Bureau of Standards
Washington, D.C. 20234
ABSTRACT
of a set of Helaholti colls. This accurately
knovn flux density is compared vlth the flux
density in the gap of the permanent magnet
standard through the use of a suitable transfer
Instrument.

The increased demand for the calibration of

permanent magnets used as reference standards
has required a nev magnet calibration facility
at the National Bureau of Standards, Washington,
D.C. The method procedure and apparatus used in
the calibration of reference magnets are
described.

THE MEASUREMENT Of MAGNETIC FIELDS

INTRODUCTION
Magnetic fields are essential to the operation
of many devices used for defense, research and
industrial applications. Many of these
applications require that the flux density in
the gap of a magnetic structure be knovn to an
accuracy of 0.1 to 5 percent. Modern materials
and design have increased the efficiency of
these magnetic structures and have Improved the
stability of the magnetic field In the gap.
Improvements and Innovations In the measurement
techniques have permitted theBe magnetic field
measurements to be more easily and generally
made with increased accuracy. The measuring
system employed usually consists of tvo parts;
a probe head that is Inserted in the gap of the
magnetic structure and an indicating or
measuring instrument. Since the probe head and
measuring instrument system are usually cali
brated to read flux density, they are often
called gaussmeters or magnetometers. (Perhaps
later the name vlll be changed to teslameters) .
The more sensitive gaussmeters use electronic
amplification vlth adjustable gain for cali
bration purposes. The gaussmeters are usually
calibrated by inserting the probe head in a
reference field and adjusting the gain of the
amplifier so the measuring instrument indicates
the correct flux density. The reference field
in many cases is the flux density in the gap of
a permanent magnet. Therefore, it Is very
essential that the permanent magnets used as
standards be calibrated and maintained as
reference standards in the laboratory.

There are many methods of measuring magnetic

field. Several of these methods may be called
absolute methods. Included among these methods
are:
1. The test coil method. This requires
a knovn value of area turns of a test
coll for flipping, vibrating or rotating
purposes.
2. Nuclear Magnetic Resonance Magnetometer.
3. Electron Paramagnetic Resonance Magnetometer
k. Rubidium Magnetometer
5. Helium Magnetometer
Several other methods by vhlch magnetic fields are
commonly measured use devices dependent on other
effects such asi
a. The Hall Effect
b. Magne tore si stance
c. Saturable magnetic cores
d. Optical effects
e. Torque
f. Susceptibility - paramagnetic aiid dlamagnetlc
There are many other effects that can be but are
Infrequently used such as:
g. Deflection of electron beam
h. Magnetostriction
1. Ettlngshausen effect
J. Thermal effects

The Magnetic Measurements Section, Institute

ABSOLUTE METHODS FOR MEASURING MAGNETIC FIELDS
for Basic Standards, of the National Bureau of
Standards, Washington, D.C, has a service of
At the present time in the Magnetic Measurements
calibrating these reference standards for the
Section there are tvo methods used for the
many users. The calibration procedure is based
absolute measurements of magnetic fields.
on the accurate measurement of the flux density
In the gap of an electromagnet or at the center
Contribution of the National Bureau of Standards, not subject to copyright.
477-1

1. The Teet Coil Method

2. Nuclear Magnetic Re tonance Method
The Teat Coll Method la uaed vhen It la
desired to measure field less than 0.02 tesla
(200 gausses) In the gap of the electromagnet
and for the smaller flelda produced at the
center of the Helmholtz Coils. The Nuclear
Magnetic Resonance Method Is uaed for measuring
the magnetic field in the gap of the electro
magnet where the field is approximately 0.02
tesla or more.
1. The Test Coil Method
This method requires that the area turnB of a
test coll be determined by basic consideration.
Referring to figure 1 - S Is a long, single
layered solenoid; the magnetic field at the
central region of this solenoid may be computed
from the geometry of the solenoid and winding,
and the current in the ^winding; C is a search
coil placed concentric with the solenoid in the
central region. L is a mutual Inductor whose
mutual inductance has been calculated from Its
dimensions and this value is known to a few parts
in a million. A potential divider circuit is
connected across the secondary of the mutual
Inductor and R^ is connected in series with the
coll C through the ballistic galvanometer B0.
The primary of the mutual inductor is connected
In series with the winding of the long solenoid
and provision is made to reverse and to measure
the current in these windings. It is easily
shown
that at balance
R..L
iiu
1 m
= (Rx + V*
Where AN Is the area turns of the coll In o2 turn, L it the mutual Inductance of the Inductor
In henries, and It is the solenoid constant in
teslas/ampere. The values of k and L are
determined from mechanical measurement.
2. Nuclear Magnetic Resonance Method
The phenomenon of nuclear magnetic resonance
makes it possible to measure homogeneous fields
with very high accuracy. Under the best
conditions it is possible with thla method to
measure magnetic fields within a few ppm. The
measurement Is based on the equation
2xv = V B
'P
or by transposing

In this equation B is the flux density in teslas,

v la the resonance frequency In hertz (c/s) and
v Is a quantity called the gyromagnetic ratio.

The gyromagnetic ratio, y is the ratio of the

magnetic moment of a nucleus to its angular
momentum. Usually the proton resonance of a
water sample is observed. The accepted value at
present for yJS.* for water Is 4.25759 x 10? hertz/
tesla.
v
The measurement is made by Immersing a sample
of material containing nuclei whose gyromagnetic
ratio Is known, in the steady homogeneous magnetic
field to be measured. The field must be uniform
within 0.1 percent over the volume of the sample
in order to obtain a good resonance curve. The
water sample in the probe is subjected by means
of a surrounding coil to an alternating magnetic
field at right angle to the steady field. The
frequency of the alternating field is adjusted to
be equal to the resonance frequency of the nuclei
in the sample. At thib frequency which is directly
proportional to the Intensity of the steady field,
there is an absorption of energy from the exciting
current. This Is observed by means of a cathoderay oscilloscope and the resonance frequency is
read from a digital counter. B is then calculated
using the equation
B *

hw teBlaB

Where v is the resonance frequency In hertz.

TRANSFER DEVICES
A few of the permanent magnet standards are
designed to be directly measured with a NMR
gaussmeter; however, most of these standards axe
designed to be used with the Hall effect,
rotating coil, search coll or torque devices.
Measuring Instruments based on these principles
are thus used as transfer devices between the
permanent magnet standards and the reference
fields.
1. The Hall Kffect
This effect furnishes a very convenient means
for the measurement of magnetic fields where
probes having very small dimensions are needed. As
a result of extensive development work many
commercial instruments are available using this
basic principle. The principle can be stated very
simply using figure 2. If a thin strip or film of
metal, usually bismuth, indium, antimonlde or
indium arsenide has a current, I, in it, two
points (a and b) can be found at opposite ends of
a line approximately at right angles to the
current which will be at the same electrical
potential, if no magnetic field, B, is acting. If
a magnetic field is applied at right angles to the
plane of the strip, a difference of electrical
potential appears between these two points. This
difference of electrical potential is proportional
to the intensity of the magnetic field. The effect
is large enough that for a fairly large field, the

478-2

voltage may be measured directly without

amplification. For small fields and for Increased
sensitivity alternating current is used or
produced so that the emf generated by the action
of the magnetic field could be amplified. Instru
ments using this principle are generally non
linear and require calibration in the range in
which they are to be used.
2.

shaft near its lower end. At the other end is a

hair spring for measuring the torque and a pointer
which moves over the scale, for fields higher
than 0.5 tesla a modified form of the Instrument
Is used, In which the permanent magnet is re
placed by a material which becomes magnetized by
Induction.
PERMANENT MAgNET STANDARDS AND THEIR CALIBRATION

Search Coils

These colls are frequently used in conjunction

with integrating devices such as ballistic
galvanometers, flux-meters, and electronic
integrators. In making the measurement a test
coil connected to the Integrating device is
placed in the magnetic field with its axis in the
direction of the field and the deflection or
reading observed when the test coil is either
removed from the field or rotated 180* about its
diameter. The signal produced is proportional
to the total change In flux linkage between the
field and the test coil.
A continuously rotating test coll is also
often used to measure magnetic fields. The
coil is rotated at constant speed about a dia
meter. It is connected through a commutator or
cam operated reversing switch to a d-c instru
ment which indicates the average voltage induced
in the coll. If the reversal comes at the time
when the a-c voltage Induced in the coll is zero,
the reading on the Instrument will be a maximum
and will be proportional to the field.
Commercial instruments operating on this
principle have the rotating coll at the end of a
long shaft which has at its other end a
reversing commutator and synchronous motor. The
shaft has to be long enough to avoid errors due
to stray fields from the motor on the probe and
to prevent the effect of the field being
measured on the motor Itself. The accuracy of
the measurement depends directly upon the
accuracy with which the frequency of the power
source is controlled and variations in the
resistance and spurious voltages at the reversing
commutator are minimized. The voltage output of
this Instrument is linear with respect to the
magnetic flux density being measured except
at small fields where effects due to commutation
may become significant. This instrument is
usually calibrated through the use of two or
more fields.
3. Pivoted Magnets
These are usually torque devices also used
with permanent magnet standards. As a result
of the development of permanent magnet materials
having extremely high coercivity, it has been
possible to produce an Instrument by which
magnetic fields can be measured in terms of
torque on a small magnet. This Instrument
consists of a light weight shaft pivoted at
each end and supported between Jewels. A small
cylindrical, high Hc magnet is mounted on a

As may be deduced from the description of the

various types of probe heads and measuring devices
used for transfer Instruments in the above section,
there are large variations in permanent magnet
standards. Among these variations are field
strength, the physical size and shape, air gap and
access opening dimensions, guide and stop arrange
ments for the probe head, temperature control,
designed accuracy and other characteristics.
Therefore, calibration of these standards requires
a considerable range of reference magnetic fields
and several types of probes and measuring instru
ments. This may be seen from figure 3- The
Magnetic Measurements Section In the Electricity
Division of the Institute for Basic Standards,
National Bureau of Standards, Washington, D. C. ,
is equipped for and has calibrated permanent
magnet standards having fields in the range of
0.0001 to 1.6 teslaa. This range will be
extended to 2 tesla In the near future. The
facilities used for setting up the reference
fields are a Helmholtz coil, a 7" symmetrical
yoke electromagnet, and a 12" symmetrical yoke
electromagnet
The reference field to be used when calibrating
permanent magnet standards having gap fields of
0.02 tesla or less are produced using a set of
Helmholtz colls or the 7" electromagnet. These
reference fields are measured using search coils
whose area-turns have been measured using the
method 1 described In the section entitled
ABSOLUTE METHODS TOR MEASURING MAGNETIC FIELDS.
The gap field of the permanent magnet standard
is compared with the reference field through a
suitable transfer instrument.
>
Reference fields greater than 0.02 tesla are
produced using the 7" or 12" electromagnet. These
reference fields are usually measured by the
nuclear magnetic resonance technique. For these
fields, the probe head of the transfer Instrument
most generally used are the search coll, rotating
coil and Hall effect. Since the permanent magnet
standards requiring calibration are not uniform
with respect to gap dimensions or In other respects,
it is usually required that the Baae type or
preferably the same probe head be used in the
calibration procedure as will be used In the field
measurement when the magnet is used as the standard.
This Is particularly true when search colls are
used as probes since frequently they are custommade and it is unlikely that we would have a
search coll having the same geometry. Also In
many cases the flux-density In the magnet

479-3

submitted for teat Is not uniform over the volume

of the active portion of the probe head.
The new calibration setup using the 7" electro
In the calibration procedure, the probe head
magnet was made possible through the financial
and measuring Instruments are used as transfer
support of the Bureau of Naval Weapons. Robert E.
devices only betveen the reference field of an
Mundy had the major responsibility for the super
electromagnet and the field of the permanent
vision of the setup and Its use In the cali
magnet standard. The reference field la
bration program of the section.
supplied In the gap of the 7" or 12" electromagnet.
This field 1b constant and homogeneous over a
volume larger than the measurement probe head of
a nuclear magnetic resonance (NMR) magnetometer.
Figure k Is a photograph of the calibration set
up for permanent magnets using the 7" electro
magnet. Location and Identification of the
various components may be seen from figure 5The NMR probe head Is placed near the center
of this highly uniform magnetic field. The
transfer probe head in the case shown, is a Hall
effect device that is inserted In the gap of the
permanent magnet and a reading Is obtained on the
Hall effect measuring device. In our case we
UBe a sensitive digital voltmeter as a readout
device of the gaussmeter. The Hall effect
probe Is placed In the gap of the electromagnet
adjacent to the NMR probe and the current from
the power supply of the electromagnet is
adjusted until the same reading Is obtained on
the digital voltmeter as when the Hall probe
was In the gap of the permanent magnet standard.
The frequency of the radio frequency field
applied in NMR magnetometer is then varied until
the resonance curve appears on the cathode ray
oscillograph. This resonance frequency Is then
measured by means of the electronic counter. The
flux-density existing In the gap may then be
calculated from the equation
B v x 10"'
where B Is this flux density in teslas and v is
the resonance frequency In hertz. This
calculated flux density In the gap of the
reference Is equal to the flux density that
exists In the gap of the permanent magnet
standard. The transfer probe Is always positioned
as closely as possible to the NMR probe when in
the electromagnet and Is positioned for a
maximum reading near the center of the air gap of
the permanent magnet standard, if any degrees of
freedom are possible. However, it is highly
desirable and Is usually required that
appropriate guides and Btops be supplied to the
permanent magnet standard If accuracy of 1
percent or better Is desired. Usually the value
of the magnetic field in the permanent magnet is
reported to an accuracy of 0.1 to 5 percent
depending on the individual need. Permanent
magnet standards having flux density from 0.0001
to 1.6 teslas have been calibrated.

480-4

481-5

482-6

483-7

484-8

Encyclopedia of Chemical Technology

Volume 7, 1965

Electrical Testing
Arnold H. Scott
Methods of testing for resistivity, dc dielectric conductivity,
dielectric constant and ac loss, dielectric strength, arc resistance,
and tracking and erosion under moist conditions are described to
give a.person working in chemical technology an understanding of the
types of electrical tests that can be made and some idea of their
usefulness .

JOURNAL OF RESEARCH of the National Bureau of StandardsC. Engineering and Instrumentation

Vol. 65C, No. 1, January-March 1961
An Ultra Low Frequency Bridge for Dielectric
Measurements
Donald J. Scheiber
(October 11, 1960)
The bridge described is capable of measuring the parallel capacitance and resistance of
dielectric specimens in the frequency range of 0.008 to 200 cps. It employs no earthing
device and is directly connected to a three terminal oscillator. The substitution method is
employed. The capabilities of the bridge are experimentally tested by measurements upon
known capacitors and resistors and by comparison with results obtained using a Schering
bridge near 100 cps. The apparatus is capable of accuracies previously unattained at these
low frequencies. Capacitances between zero and 100 pf may be measured to an accuracy
of (0.05% + 0.002 pf) at frequencies above 5 cps. Below 5 cps the accuracy becomes
[0.05% + (0.002+ 2X105//A,D)pfl, where /is the frequency in cps and HD is the equivalent
resistance in ohms across the detector terminals. Conductances between 10 " and 10"15
mhos may be measured to an accuracy of about (1 % + 3/X 10~15 mhos) when />0.1 cps.
When /<0.1 cps the accuracy is about (1% + 2X10~18 mhos). The dielectric constant
(') of a specimen may be determined to an accuracy proportional to that of capacitance
measurements. The loss index (") may be determined to an accuracy of about (1%
+ 5X 10-*/Ct) when />0.1 cps and (1% + 3X 10"5//C,) when /<0.1 cps. Here C, is the
vacuum capacitance of the specimen expressed in picofarads. The effects of stray imped
ances shunting the bridge ratio arms, are investigated. Useful modifications of the bridge
are discussed.

485-1

Proceedings from the Electrical Insulation Conference

IEEE-NEMA, 1967

Low Frequency Dielectric Behavior

William P. Harris
Electrical engineers, as well as polymer physicists and
materials researchers, can gain useful information from a study
of the behavior of dielectrics, and of high-megohm resistors at
very low frequencies. This is illustrated by examples.
There are two main methods of obtaining ultra- low- frequency
data. One is to apply a dc voltage producing a time- dependent
current which corresponds to an inverse frequency plot. The
other is to use an alternating voltage, usually sinusoidal, and
apply this to a bridge circuit including the material under
study. Recent developments in the latter methods are presented,
along with some typical results.

Key Words -

Bridge, Dielectrics, Dielectric Measurements,

Dielectric Phenomena, Electrical Measurements,
Low Frequency, Measurements, Operational Am
plifiers, Ultra Low Frequency.

486-2

Subject Index1
(Reference Is to volume number, boldface, followed by page number of this volume]
A

Volume and Page

Absolute ampere determination
3-6, 3-10, 3-69A
Absolute electrical units
3-4,3-9
Absolute ohm determination
3-5, 3-9, 3-70A
Ac-dc comparators, peak.
3-310
Ac-dc comparators, rms
3-293
Ac-dc transfer standards, see also Thermal con
verters
Ac-dc transfer standards
3-278,
3-281, 3-293, 3-310, 3-318, 3-325A, 3-326A
Accuracy of NBS electrical calibrations
3-29, 3-44
Ampere, definition
3-8

Volume and Page

Four-terminal resistors, calibration.
3-174, 3-200
Pour-terminal resistors, series-parallel buildup
3-140

Gyromagnetic ratio, proton.

Hall effect
3-452, 3-478
High voltage measurements, see Voltage measure
ments.
I
Inductance measurements
3-17
Inductive voltage dividers, calibration
3-357,
3-362, 3-371, 3-380A
Inductive voltage dividers, design
3-380A
Inductive voltage dividers, international com
parison
3-37 1
Inductors, calibration.
3-259, 3-265
Inductors, computable
3-272A
Inductors, for high voltage
3-389
Inductors, mutual
3-265
International system of units (SI)
3-4, 3-9
International weights and measures, bureau
3-1

B
BIPM
Bridges, see specific type

3-1

Calibration, see specific electrical instrument or

standard
Capacitance measurements
3-12
Capacitors, air

3-244
Capacitors, calibration
3-254, 3-271 A, 3-272A
Capacitors, fused-silica dielectric
3-237
Capacitors, stability
3-240
Capacitors, standard
3-237
Capacitors, voltage dependence
3-244
Coaxial chokes
3-248, 3-266, 3-353
Current comparators
3-24, 3-329
Current comparators, calibration
3-330
Current measurements, ac
3-22
Current measurements, dc
-3-19
Current-ratio standards, international compari
son
3-329
Current transformers
3-329
Current transformers, calibration 3-23, 3-330, 3-379A
Current transformers, design
3-379A

K
Kelvin double-bridge
-- 3-168,3-200
Kelvin double-bridge, effect of link circuit
3-200
3-35

Laboratories, standards .
M
Magnetic constant
Magnetic materials
Magnetic measurements
Magnetic standards
Magnetic units
Magnetometers
Maxwell-Wien bridge
MKSA units
Mueller bridge
Mutual inductors.

D
Dielectric constant
3-12, 3-429
Dielectric measurements.. 3-427, 3-431, 3-485A, 3-486A
Dielectric specimen holder
3-431
Differential thermocouple voltmeter
3-318
Direct-reading ratio sets (DRRS)
3-170, 3-271A
Direct-reading ratio sets, calibration
3-187
Direct-reading URS
3-198
E
Electrical units
3-4, 3-9, 3-67A, 3-68A
Energy measurements, electrical
3-19
The letter "A" following the page number Indicates an abstract only.
315-411 O - 68 - 32

3-15

487

3-442
3-470
3-441
3-463
3-18, 3-441
3-477
3-259
3-4
3-166
3-265

N
Nuclear magnetic resonance.

3-452, 3-478

Oil baths

3-111,3-114

P
Peak a-c to d-c comparators, calibration
Permanent magnets, calibration
Permeameters.
Potentiometers, calibration
Potentiometers, URS type
Power measurements, electrical
Proton gyromagnetic ratio

Volume and Page

3-310
3-477
3-457
3-184, 3-233A
3-290
3-19
3-15

R
Resistance alloys
3-153
Resistance measurements
3-13
Resistors, calibration
3-161, 3-231A, 3-232A
Resistors, construction
3-136, 3-155
Resistors, effects of humidity
3-158
Resistors, for high voltage
3-383, 3-389, 3-416
Resistors, four-terminal
3-164, 3-231A
Resistors, load coefficient
3-160
Resistors, National Reference Group
3-135
Resistors, series-parallel buildup
3-140
Resistors, stability
3-137, 3-160
Resistors, standard
3-135, 3-159
Resistors, temperature coefficient
3-154
S
Shunts, see Resistors
Skin effect
3-404
Standard capacitors
3-237
Standard-cell comparator (Brooks)
3-84
Standard cells, calibration
3-83, 3-131 A
Standard cells, construction
3-91
Standard cells, contact clamps
3-117
Standard cells, effect of current
3-103
Standard cells, emf-temperature coefficient
3-98
Standard cells, enclosures
3-111
Standard cells, internal resistance
3-103
Standard cells, international comparisons
3-85
Standard cells, National Reference Group
3-81
Standard cells, oil baths
3-111, 3-114
Standard cells, stability
3-82
Standard resistors
3-135, 3-159
Standard volt boxes, see also Volt boxes
Standard volt boxes, calibration
3-204
Standards laboratories, practices
3-35
Surge measurements
3-25,3-389

T
Volume and Page
Thermal converters
3-20, 3-279, 3-281, 3-300
Thermal converters, calibration
3-282,
3-293, 3-300, 3-324A
Thermal converters, frequency errors
3-287
Thermal converters, thermoelectric errors
3-286
Thermoelements, see Thermal converters
Thermometer bridges, calibration
3-180
Thompson-Lampard theorem
3-11
Transfer standards, see Ac-dc transfer standards
Transformer bridges
3-244, 3-265, 3-271 A, 3-350
Transformer ratios, measurement
3-349
Transformers, see Current and voltage trans
formers
u
Universal ratio sets (URS)
3-177, 3-184
Universal ratio sets, calibration
3-233A, 3-296
Universal ratio sets, direct-reading
3-198
V
Volt boxes
3-204,3-292
Volt boxes, calibration
3-204, 3-217, 3-222, 3-297
Volt boxes, calibration console
3-210, 3-217
Volt boxes, effects of rod resistances
3-212, 3-223
Volt boxes, effects of self-heating
3-206, 3-224
Voltage dividers, see Inductive voltage dividers
and Volt boxes
Voltage measurements, ac
3-22
Voltage measurements, dc
3-15
Voltage measurements, high
3-17, 3-383, 3-424A
Voltage measurements, surge
3-25, 3-413
Voltage measurements, system
3-290
Voltage transformers, calibration
3-335, 3-343
Voltage transformers, cascade type
3-342
Voltage transformers, international comparison.. 3-342
w
Watthour meters, calibration
Wattmeter, standard
Weston (cadmium sulfate) standard cells
Wheatstone bridges
Wheatstone bridges, calibration
Z
Zener reference diodes
Zener reference diodes, calibration
Zener reference diodes, stability

488

3-325A
3-325A
3-90
3-164
3-179
3-118,3-125
3-119, 3-127
3-120

Broadhurst, M. G
Brooks, P. P. B
Bur, A. J
Cameron, J. M
Chinburg, C. H
Cones, H. N
Cooter, I. L
Cutkosky, R. D

Author
Volume and Page
3-431
3-187, 3-232A
3-431
3-131A
3-380A
3-413
3-9, 3-439, 3-477
3-69A,
3-70A, 3-237, 3-271A, 3-272A, 3-349

Defandorf, F. M
Driscoll, R. L
Dunfee, B. L
Dunn, A. F
Dziuba, R. F

3-424A
3-69A
3-9, 3-204, 3-329, 3-379A
3-371
3-222

Eicke, W. G., Jr

3-118, 3-125, 3-131A

Flach, D
Forman, N. L
Griffin, J. E...

Lee, L. H
Lewis, A. B
Lisle, R. V
Lowrie, P. H., Jr
Marzetta, L. A
McGregor, M. C
Moore, W. J. M

3-310
3-379A
3-290,3-318

3-184, 3-198, 3-200, 3-233A

Zapf, T. L

489

3-310
3-271A
3-329, 3-342

Ramaley, D

Weaver, F. D
Wenner, F
Williams, E. S
Wolf, H. K

1 The letter "A" following the page number Indicates an abstract only.

-.

3-1, 3-8, 3-67A, 3-140, 3-272A

3-325A, 3-379A, 3-383, 3-389, 3-413
--3-9
3-342

Thomas, J. L
Turgel, R. S

3-271A
3-342

Volume and Page

3-237
3-325A
3-357
3-114, 3-217

Page, C. H.
Park, J. H
Peterson, C
Petersons, O

Sanford, R. S
Scheiber, D. J
Scott, A. H
Shafer, J. F
Shields, J. Q
Silsbee, F. B
Smith, R. L
Souders, T. M
Spinks, A. W
Sze, W. C

Hamer, W. J
3-73
Harris, F. K
3-4, 3-9, 3-271A, 3-342
Harris, W. P
3-9, 3-486A
Hermach, F. L.__. 3-9, 3-29, 3-275, 3-281, 3-290, 3-318
Hersh, J. F
3-271A
Hess, E. E
-- 3-234A
Homan, D. N
3-265
Houghton, J. R
3-326A
Kotter, F. R
Kusters, N. L

Index 1

3-439
3-485A
3-232A, 3-427, 3-485A
3-198
3-244, 3-349
3-33, 3-67A, 3-68A, 3-231A, 3-379A
3-379A
3-222
3-325A
3-335, 3-342, 3-362, 3-371
3-135, 3-149
3-300
3-44, 3-326A
- 3-231A
3-281, 3-290, 3-324A
3-380A

3-254,
3-259, 3-272A, 3-325A, 3-357, 3-371, 3-380A

NATIONAL BUREAU OF STANDARDS

The National Bureau of Standards1 was established by an act of Congress March 3,
1901. Today, in addition to serving as the Nation's central measurement laboratory,
the Bureau is a principal focal point in the Federal Government for assuring maxi
mum application of the physical and engineering sciences to the advancement of tech
nology in industry and commerce. To this end the Bureau conducts research and
provides central national services in three broad program areas and provides cen
tral national services in a fourth. These are: (1) basic measurements and standards,
(2) materials measurements and standards, (3) technological measurements and
standards, and (4) transfer of technology.
The Bureau comprises the Institute for Basic Standards, the Institute for Materials
Research, the Institute for Applied Technology, and the Center for Radiation Research.
THE INSTITUTE FOR BASIC STANDARDS provides the central basis within the
United States of a complete and consistent system of physical measurement, coor
dinates that system with the measurement systems of other nations, and furnishes
essential services leading to accurate and uniform physical measurements throughout
the Nation's scientific community, industry, and commerce. The Institute consists
of an Office of Standard Reference Data and a group of divisions organized by the
following areas of science and engineering :
Applied MathematicsElectricityMetrologyMechanicsHeatAtomic Phys
icsCryogenics2Radio Physics2Radio
Engineering2Astrophysics2Time
and Frequency.2
THE INSTITUTE FOR MATERIALS RESEARCH conducts materials research lead
ing to methods, standards of measurement, and data needed by industry, commerce,
educational institutions, and government. The Institute also provides advisory and
research services to other government agencies. The Institute consists of an Office of
Standard Reference Materials and a group of divisions organized by the following
areas of materials research:
Analytical ChemistryPolymersMetallurgy Inorganic Materials Physical
Chemistry.
THE INSTITUTE FOR APPLIED TECHNOLOGY provides for the creation of appro
priate opportunities for the use and application of technology within the Federal Gov
ernment and within the civilian sector of American industry. The primary functions
of the Institute may be broadly classified as programs relating to technological meas
urements and standards and techniques for the transfer of technology. The Institute
consists of a Clearinghouse for Scientific and Technical Information,3 a Center for
Computer Sciences and Technology, and a group of technical divisions and offices
organized by the following fields of technology :
Building ResearchElectronic Instrumentation Technical Analysis Product
EvaluationInvention and Innovation Weights and Measures Engineering
StandardsVehicle Systems Research.
THE CENTER FOR RADIATION RESEARCH engages in research, measurement,
and application of radiation to the solution of Bureau mission problems and the
problems of other agencies and institutions. The Center for Radiation Research con
sists of the following divisions:
Reactor RadiationLinac RadiationApplied RadiationNuclear Radiation.
1 Headquarters and Laboratories at Gaithersburg, Maryland, unit
2 Located at Boulder, Colorado 80302.
Located at 5285 Port Royal Road. Spring-field. Virginia 22151.

otherwise noted ; mailing address Washington. D. C 20234.

NBS TECHNICAL PUBLICATIONS

PERIODICALS
JOURNAL OF RESEARCH reports National
Bureau of Standards research and development in
physics, mathematics, chemistry, and engineering.
Comprehensive scientific papers give complete details
of the work, including laboratory data, experimental
procedures, and theoretical and mathematical analy
ses. Illustrated with photographs, drawings, and
charts.
Published in three sections, available separately:

N0NPERI0DICALS
Applied Mathematics Series.
tables, manuals, and studies.

Mathematical

Building Science Series. Research results, test

methods, and performance criteria of building ma
terials, components, systems, and structures.
Handbooks. Recommended codes of engineering
and industrial practice (including safety codes) de
veloped in cooperation with interested industries,
professional organizations, and regulatory bodies.

Physics and Chemistry

Papers of interest primarily to scientists working in
these fields. This section covers a broad range of
physical and chemical research, with major emphasis
on standards of physical measurement, fundamental
constants, and properties of matter. Issued six times
a year. Annual subscription: Domestic, $6.00; for
eign, $7.25*.
Mathematical Sciences
Studies and compilations designed mainly for the
mathematician and theoretical physicist. Topics in
mathematical statistics, theory of experiment design,
numerical analysis, theoretical physics and chemis
try, logical design and programming of computers
and computer systems. Short numerical tables.
Issued quarterly. Annual subscription: Domestic,
$2.25; foreign, $2.75*.

Special Publications. Proceedings of NBS con

ferences, bibliographies, annual reports, wall charts,
pamphlets, etc.
Monographs. Major contributions to the techni
cal literature on various subjects related to the
Bureau's scientific and technical activities.
National Standard Reference Data Series.
NSRDS provides quantitative data on the physical
and chemical properties of materials, compiled from
the world's literature and critically evaluated.
Product Standards. Provide requirements for
sizes, types, quality and methods for testing various
industrial products. These standards are developed
cooperatively with interested Government and in
dustry groups and provide the basis for common
understanding of product characteristics for both
buyers and sellers. Their use is voluntary.

Engineering and Instrumentation

Reporting results of interest chiefly to the engineer
and the applied scientist. This section includes many
of the new developments in instrumentation resulting
from the Bureau's work in physical measurement,
data processing, and development of test methods.
It will also cover some of the work in acoustics,
applied mechanics, building research, and cryogenic
engineering. Issued quarterly. Annual subscription:
Domestic, $2.75; foreign, $3.50*.
TECHNICAL NEWS BULLETIN
The best single source of information concerning
the Bureau's research, developmental, cooperative
and publication activities, this monthly publication
is designed for the industry-oriented individual whose
daily work involves intimate contact with science
and technologyfor engineers, chemists, physicists,
research managers, product-development managers, and
company executives. Annual subscription: Domestic,
$3.00; foreign, $4.00*.
Difference in price is due to extra cost of foreign mailing.

Technical Notes. This series consists of com

munications and reports (covering both other agency
and NBS-sponsored work) of limited or transitory
interest.
CLEARINGHOUSE
The Clearinghouse for Federal Scientific and
Technical Information, operated by NBS, supplies
unclassified information related to Governmentgenerated science and technology in defense, space,
atomic energy, and other national programs. For
further information on Clearinghouse services, write:
Clearinghouse
U.S. Department of Commerce
Springfield, Virginia 22151

Order NBS publications from:

Superintendent of Documents
Government Printing Office
Washington, D.C. 20402
U.S. GOVERNMENT PRINTING OFFICE : [968 O - 315-411

Announcement of New Volumes in the

NBS Special Publication 300 Series
Precision Measurement and Calibration

Superintendent of Documents
Government Printing Office
Washington, D.C. 20402
Dear Sir:
Please add my name to the announcement list of new volumes to be issued
in the series : National Bureau of Standards Special Publication 300, Precision
Measurement and Calibration.

Name
Company

Address
City
(Notification keyN : 353)

State

Zip Code

Official SI Unit Names and Symbols

fFor a oomplete statement of NBS practice, see]
LNBS Tech. News Boll. Vol. 82, No. 6, June 1888.J
Name
meter
kilogram..
second
ampere
kelvin 1
candela
radian
steradian
hertz
lumen
lux

Symbol

m
kg
s
A
K
cd
rad
sr
Hs
lm
be

Name
newton
joule
watt
coulomb
volt
ohm
farad
weber
henry
tesla

Symbol
N
J
W
C
V
11
F
Wb
H
T

Additional Names and Symbols approved for NBS use

curie'
degree Celsius *
gram
-

Ci
?C
g

mho
mole
Siemens *

mho
mol
8

i The same name and symbol are osed for thermodynamic temperature and temperature Interval.
(Adopted by the ISth General Conference on Weights 4 Measures, 1947.)
' Accepted by the General Conference on Weights & Measures for use with the 81.
> For expressing "Celsius temperature"; may also be used for a temperature Interval.
' Adopted by IEC and ISO.
Table for converting U.S. Customary Units to those of the International
System (SI)5
To relate various units customarily used in the United States to those of the International
System, the National Bureau of Standards uses the conversion factors listed in the "ASTM Metric
Practice Guide", NBS Handbook 102. These are based on international agreements effective
July 1, 1959, between the national standards laboratories of Australia, Canada, New Zealand,
South Africa, the United Kingdom, and the United States.
To convert from :
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)
(14)
(15)
(16)
(17)
(18)
(19)
(20)
(21)

inches to meters, multiply by 0.0264 exactly.

feet to meters, multiply by 0.3048 exactly.
feet (U.S. survey) to meters, multiply by 1200/3937 exactly.
yards to meters, multiply by 0.9144 exactly.
miles (U.S. statute) to meters, multiply by 1609.344 exactly.
miles (international nautical) to meters, multiply by 1852 exactly.
grains (1/7000 lbm avoirdupois) to grams, multiply by 0.064 798 91 exactly.
troy or apothecary ounces mass to grams, multiply by 31.103 48 . . .
pounds-force (lbf avoirdupois) to newtons, multiply by 4.448 222 . . .
pounds-mass (lbm avoirdupois) to kilograms, multiply by 0.453 592 . . .
fluid ounces (U.S.) to cubic centimeters, multiply by 29.57 . . .
gallons (U.S. liquid) to cubic meters, multiply by 0.003 785 . . .
torr (mm Hg at 0 C) to newtons per square meter, multiply by 133.322 exactly.
millibars to newtons per square meter, multiply by 100 exactly.
psi to newtons per square meter, multiply by 6894.757 exactly.
poise to newton-seconds per square meter, multiply by 0.1 exactly.
stokes to square meters per second, multiply by 0.0001 exactly.
degrees Fahrenheit to Kelvins, use the relation tK= (tF + 459.67) /l. 8.
degrees Fahrenheit to degrees Celsius, use the relation lc= (< 32)/1.8.
curies to disintegrations per second, multiply by 3.7X 1010 exactly.
roentgens to coulombs per kilogram, multiply by 2.579 760X 10"' exactly.
Systeme International d' Unites (designated SI in all languages).