IEEE TRANSACTIONS ON MICROWAVE THEORY AND TEC!HNIQIJIJ3, VOL. MTT-17, NO.

5, MAY 1969 259

of a two-slot transmission line on non-planar surfaces,” IEEE [lOb] F. G. Tricorni, Integral Equations. New ‘York: Interscience
Trans. Microwave Theory and Techniques, vol. MTT-13, pp. Publishers, 1957.
303-306, May 1965. [1OC] N. P. Vekua, Systems of Singular Integral Equations. Gronin-
.f8dl. J. P. Shelton. Jr.. “Impedances of offset ~arallel-counled stri~ gen, Netherlands: Noordhoff, 1967.
transmission lines,” IE’EE Trans. Microw;ue Theory ;nd Tec~- [11] I. Pa16cz, “The integral equation approach to currents and fields
iziques, vol. MTT-14,pp. 7–15, January 1966. in plane parallel transmission lines,” J. Math. Mech., vol. 14,
[Se] S. Yamamoto, T. Azakami, and K. Itakura, “Coupled strip pp. 541-560, April 1966.
transmission line with three center conductors,” IEEE Traos. [12] K, Nickel, “Losung eines Integralgleichungssysterns aus der
Microwave Theory and Techniques, vol. MTT-14, pp. 446-461, Tragfliigeltheorie,” Math. Z., vol. 54, pp. 81-96, 1951.
October 1966. [13] L. Lcwin, “The solution of singular integral equations over a mul-
[8f] — “Slit-coupled strip transmission lines,” IEEE Trans. tiple interval and applications to multiple diaphragms in rec-
Mic;owave Theory and Techniques, vol. MTT-14, pp. 542-553, tangular waveguides,” SIAMJ. of Appl. Math., vol. 16, pp. 417-
November 1966. 438, March 1968.
[8g] H. C. Okean, “Properties of a TEMtransmission line used in a
[14a] F. G. Tricomi, “On the finite Hilbert transform,” Quart. J. of
microwave integrated circuit
application,” IEEE Trans. on
Math., vol. 2, pp. 199-211, 1951.
Microwane Theory and Techniques, (Correspondence), vol. MTT-
[14b] H. Sohngen, “Zur Theorie der Endlichen Hilbert Transforma-
15, pp. 327-328, May 1967.
tion,” Math. Z., vol. 60, pp. 31-51, 1954.
[8h] H. Guckel, “Characteristic impedances of generalized rectang-
[14C] F. G. Tricomi, “The airfoil equation for a double interval,” Z.
ular transmission lines:’ IEEE Trans. Microwave Theory and
Angew. Math. Phys., vol. 2, pp. 402-406,1951,
Techniques, vol. MTT-13, pp. 27&274, May 1965.
[8i] M. K. Krage and G. I. Haddad, “The characteristic impedance [15a] J. C. Tranter, Integral Transforms in Mathematical Physics.
and coupling coefficient of coupled rectangular strips in a wave- New York: Wiley, 1951.
guide,” IEEE Trans. Microwave Theory and Techniques, vol. [15b] R. V. Churchill, “The operational calculus of Legendre trans-
MTT-16, pp. 302-307, May 1968. forms,” J. Math. and Phys., vol. 33, pp. 165-177, 1955.
[9] G. Wendt, “Statische Felder und Stattionare Strome,” Encyclo- [15C] G. Cinelli, “An extension of tbe finite Hankel transform and
pedia of Physics, S. Flugge, Ed., vol. 16. Berlin: Springer, applications,” Znternat. J. Engineering Science, vol. 3, pp. 539-
1958 (see especially page 37). 559, October 1965.
[lOa] N. I. Muskhelisvili, Singular Integral Equations. Groningen, [15d] D. Naylor, “On a finite Lebedev transform,’” J. Math. Mech.,
Netherlands: Noordhoff, 1967. vol. 12, pp. 375–383, 1963.

From Approximations to Exact Relations for

Characteristic Impedances

WOLFGANG HILBERG

Absfract—Approximations for the characteristic impedance of a eristic impedances could be given exactly orIly by means of
special two conductor stripline and for the general function K/K’ are
:lliptic integrals, but that in [1] the theory of elliptic inte-
derived by a transformation method recently described in the literature;
the first, second, and third approximations having a greatest relative error y-als was not needed at all. The calculation was made by
of the order of 10–s, 10–B, aad 10-12, respectively. They cao be introduced ~arious conformal mappings. In the following paper [2] a
into an algorithm which is based on elementary conforrnal mappings, and
:onnection was found between the theory in [1] and the
thus further approximations with rapidly vanishing errors can be derived.
The reanlts agree with those for the function K/K’ obtained by elliptic heory of transformations of elliptic iategrals, in particular
integral theory. Obviously no such theory is needed to calculate charac- he Landen transform. From this connection a general and
teristic impedances or the function K/K’ with arbitrary accuracy. Tbe
rery advantageous algorithm was derived to get the func-
advantages of the new method are illustrated for shielded coupled-strip
transmission lines, for which an extended diagram with extreme parameter ion K/K’. In the first part of this paper it will be proved that
values bas been worked out. t is possible to get the useful and general formulas of [2]
without going back to elliptic integral theory. Elementary
:onformal mappings as in [1] will suffice.
I. INTRODUCTION
In the second part, two coupled strip transmission lines
METHC)D has been developed [1] to achieve good u-e treated, as an interesting technical and somewhat more

A approximations
some
improve these approximations
for the characteristic impedances of
parallel and conical transmission lines,
step by step, so that
and to
at last
complicated example. Thus it will be illustrated
new method of calculation
(me, that it is generally
that the
is simpler than the conventional
applicable, and that the results agree
arbitrarily good approximations, i.e., exact values, can be with those obtained by the conventional theory. Further, it is
obtained. The interesting fact is that up to now these charac- ;hown that even in cases with extreme design parameters,
leeded, for example, for new developments, the new method
~ives the results faster and more exactly by elementary for-
Manuscript received May 8, 1968; revised December 26, 1968.
nulas than they can be obtained by tabulated elliptic inte-
The author is with the AEG-Telefunken Research Institute, Ulm,
Germany. yal functions.
260 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, MAY 1969

k = Cos a!, k’ = sin a. (6)

–––---H--L---
(a)
Later on two coupled striplines will be treated (Fig. 2). For
these arrangements it is usual to define even and odd com-
ponents of characteristic impedances. With these quantities
the calculation of arbitrary couplings can be carried out.
Cohn [3] calculated Z w,= and Z.dd fOr the lines in Fig. 2:

307r K(M)
z —— — (7)
“en – tiz K(kti)

where

‘e=‘anh(-a”’anh(:?)
Fig. 1. (a) Coplanar parallel transmission line. (b) Conical transmis-
sion line of a strip perpendicular to a conducting ground.

(8)
k: = <1 – k.’;
II. THE CONVENTTONALEXACT REPRESENTATION
OF CHARACTERISTIC IMPEDANCES and
It is well known that most characteristic impedances of 307r K(ko’)
transmission lines cannot be given exactly by elementary Zodd = — — (9)
<z K(kO)
functions. This is true without exception for strip transmis-
sion lines, which are in general use nowadays and are built in where
various ways. For these striplines the coplanar parallel
transmission line of Fig.
line of a strip perpendicular
l(b) are of central importance
l(a) and the conical
to a conducting plane in Fig.
[1]. This will be illustrated by
transmission
“=tanh(:a
“Coth(;%) ,10)
——
an example in Section V. The parallel transmission line of kd = 41 – ko’.
Fig. l(a) has the characteristic impedance,
The coupling factor K of an arrangement, specially designed
q K(k) as a directional coupler is well known:
z ~.— . (1)
4 K(W) K= z even – Z.dd
(11)
Here z ,VCII+ Zc,dd

III. APPROXIMATIONS

The coplanar parallel transmission line of Fig. l(a) is con-

verted into the conical line of Fig. l(b) by simple conformal
in particular
mappings [1], where the relation exists
—

Y’
/Jo
— = 1207r ohms = 377 ohms. (2b) a 7r 4 ‘Y

co arc tan — . (12)

T=T– ti ?J
K(k) is the complete elliptic integral of the first kind,
Therefore, only one transmission line needs be regarded in
the following general considerations. Because of the shorter
(3)
expressions for the moduli the conical line of Fig. l(b) is
chosen. For this line a first approximation can be derived in
and k and k’ are the so-called moduli with the mutual an elementary way [1], containing two expressions for dif-
dependence ferent ranges of validity, having a common boundary and
~, + ~1, = ~. (4) which together cover the whole range of definition:

For Fig. l(a) with the characteristic impedance (1) the

moduli are
z= ~ln
[12 cot ~ ,

‘=sin(2arctant9* (5)

“= ‘C’s(’arc’anla W2’O’(:-;I 4 ’13)

z=

The conical transmission line of Fig. l(b) has the same char-
acteristic impedance (l), but the moduli are given by
HILBERG: CHARACTERISTIC IMPEDANCE APPROXIMATIONS AND EXACT RELATIONS 261

The greatest relative error arises exactly at the point The corresponding approximation for the function K/K’ is
a = Tr/4, Z= ~/4 and is smaller than 2.36X 1(F3. By compari- then
son of the formulas for the characteristic impedance in (6)
and (13) the relation can be obtained:
%-=+
’”[2:%1’
Cotl ~ — 1
a l+k 2 forl &
SK,_ ‘m and L<k <l.
cot—. — k= . (14) d2 –
2 d l–k’ ‘r
Cotz ; + 1 (17)
.4

Herewith a corresponding approximation for the elliptic

integral function can be set up: 1
1 and O~k~=.
42

The numerical treatment showed these approximations to

have a relative error lower than 3.10-’.
Continuing the mappings, a further approximation for the
conical line can be obtained:
(15)
<l+cosa+#4cosa
z =~ln
[
2
v’l+cosa-q4cosa 1 ‘

for~~Z~~ and O:SCYs~.

(18)
v’1 + sin a + &4 sin a-
The greatest relative
In regarding
the two expressions
error is the same as in (13).
these approximations
in (13) and(15)
it is conspicuous
are built very similarly.
that
z = V; /In
[
2
41+sina–~4 sin a.’ 1
The second expression is essentially the reciprocal of the
first one, only a and k are replaced by the complementary
parameters 7r/2– a and k’. In just the same way the second
The corresponding approximation for the function K/K’ is
expression was derived from the first in [1]. It can easily be
.—
shown that this procedure can be generalized. From an K dl+k+qG
approximation for one of the ranges, e.g. for
0~as~/4withv/45Z<~, or OSk2S~with
the range
l< K/K’<w,
F“2r
‘In 2 —
[ dl+?c-qz 1 ‘

an approximation with the same accuracy is immediately

obtained for the other range. This becomes especially clear
when looking at the expression K(k)/K(k’). If k runs from O (19)
to l/@, k’ runs from 1 to l/v’Z because of (4). Thus, K @ + k’ + +4)/
knowing the function
gets those in the remaining
values in the range OS k< l/ti2,
range 1//2 S k <1 by taking
one
the
—
K’
= 27r/in
[
2
@
—
+ k’ – +’41G’ -1 ‘

reciprocal of the function and exchanging k’ and k.

for O<~51 and Osks-$.
From the relations (38) and (39) in [1], obtained by con-
formal mapping, a further approximation results by suitably
choosing half the relations and introducing the same bound- The approximations (18), (19) have a greatest relative
aries as in (13) and (15): error smaller than 4X 1&12, thus supplying far better values
than contained in all known tables. Furthermore it should

z .~lnz 1 + 4COS
_ a–> be noted that in using the approximations
ficulties of interpolation,
tlhere are no dif-
unavoidable with tabulated func-
[ 1 – 4COS a 1
tions. At last it should be remarked that elementary func-
tion are often much better suited for general calculations
than elliptic integrals, especially, for example, for integration
(16) and similar operations.

z “v~/in
[
21+ti!!H
1 – tisin “1 a
, IV. RECURSIONSFOR ARBITRARILY HIGH ACCURACY
The cycle of stenographic projection and other simple con-
formal mappings, leading to the last approximation, can be
run through systematically again and again.
262 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, MAY 1969

The approximations arising by this procedure have errors

vanishing exceedingly fast. If, unlike the recursion formula
in [1], constant boundaries are set and the algorithms (49),
(50) in [1] are joined, the analytic result of the transforma-
tion cycle can be described in the following algorithm:

N=0,1,2, . . ..W.

fo=:. [$l.=2N2’1n[2K
::1
To derive this general and very rapidly converging algo-
v=1,2, .o. ,N. rithm for the function K/K’ the theory of elliptic integrals is
not needed at all, in contradiction to the former paper [2].
f,-, = ~ – arc tan i’/~W, In principle it is even possible to derive in this way the well-
known Landen transform of K(k). These remarkable and in
V=O, —1, —2, . ..(N(N - 1). (20) particular theoretically interesting facts cannot be traced
here.
Finally, it shall be mentioned that by rearranging the algo-
rithm (21), correspondingly good approximations can be
obtained in the same simple way for the inverse function
(K/K’)-’. From the technical point of view this means the
computation of the geometrical parameters of transmission
lines for given values of the characteristic impedance. For
practical claims of accuracy it should be simplest to solve the
above approximations for their arguments.

V. THE SHIELDED COUPLED STRIP TRANSMISSIONLINE

In this section it shall be shown, as an example of the prac-

In the last four rows it can be seen that the recursion tically important coupled striplines, why the coplanar strip-
starts from the approximation (13). Then the first improve- line of Fig. l(a) has a central importance, and how other
ment for N= 1 yields the approximation (18). The greatest cross sections of lines can be transformed into this one. For
relative error reduces from 2X 10-3 to 4X l&12. The ap- the coplanar stripline of Fig. l(a) a first approximation can
proximation (16) cannot be obtained with this transforma- be obtained from (13) together with (12):
tion cycle. It is left out. However, as we are free to introduce
——
every suitable approximation in the algorithm, (16) ean be ~r’/r + 1
chosen as the beginning. The first improved approximation
for N= 1 then has a greatest relative error far below I&lz.
z= ‘In
2r
2 _
[ #r’/r – 1 1 ‘

Starting from (18) yields a still better first improvement, and

for~sZS@ and lSEs(l+ @4.
so on. r
Summarizing all the results of the algorithm (20) based on (22)
(13) and (16), it can be realized that after converting the
results into approximations for the function K/K’, they are
identical with those obtained in [2] with the aid of the Lan-
den transform. The condensed general algorithm, compris-
ing all the results above, can be written in the following way:

N=0,1,2, . . ..c0. The greatest relative error of 2.4X l&’ is good enough for
the following.
k, = k, ii; = k’. Let us consider the shielded coupled strip transmission
1 – H-1 l–k, line of Fig. 2. For the calculation of the coupling, the charac-
k, = k;–l=———— k:+k~=l. teristic impedances Z,.em of the even mode (see Fig. 3(a)) and
l+kj!-l’ l+ k,’
Zo~~ of the odd mode (see Fig. 4(a)) must be known. If we
k; forv=l,2, . . ..N. take the even mode fist and pick out a quarter of the line
cross section (see Fig. 3(b)), then this part will have the
k-N forv=O, –l, . ..(N(l) l).
characteristic impedance 2 Z.V,.. A mapping intothe~ axis
HILBERG : CHARACTERISTIC IMPEDANCE APPROXIMATIONS AND EXACT RELATIONS 263

I
t
I t ----E -----
Fig. 5. Shielded strip transmission line.

I with the Schwarz–ChristoffeI transformation,

Fig. 2. Shielded coupled strip transmission

1
line.

w =s c [(z - X,)(Z – Z2)]-’M + B (23)

leads, after the determination of the constants, to the trans-

formation equation

h
w = — — arc cosh ~<. (24)
n-

If we substitute first w= – UI and then w== – UZ using the

(:) corresponding
following
values z= —r and z= —r’, respectively,
relationship is obtained for r’/r:
the

1 cosh2 ~
I I-t h
I
— .— (25)
—-—-———— — 4, r 7TU1
P
‘1 -U2 -u,’ W2.0 coshz —
h
(b)

Before substituting this value into (22) we have to consider

that (22) yields the characteristic impedance for the whole
plane in Fig. l(a). Consequently, one half-plane, into which
(c) the cross section represented in Fig. 3(b) was transformed,
Fig. 3. (a) Even mode. (b) A quarter cross section. has twice this value. For Z,v,n the factor 2 is eliminated
(c) Mapping into a straight line. throughout, and after substituting (25) into (22) the follow-
ing expression is obtained:

dcoshwosh?+’
z even=

‘1” [ 2d’”’h%’cosh?-’
‘ 1
(a)

‘1’=”
———————
f ~-{W2=0
w -U2 - u{

(b) The calculation is similar for the odd mode, We pick out a
quarter of the transmission line (see Fig. 4(b)), which has a
characteristic impedance of 2 Z.cIcI, and transform the bound-
ary into the x axis, applying the transformation equation

h
Fig. 4. (a) Odd mode. (b) A quarter cross section. w = — —arc sinh~—z. (27)
(c) Mapping into a straight line. r
264 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, MAY 1969

103
8
6
t
z
E’ L

24 \ I 1 r--L
I

30T-

1 t ,

21

10” ~ I I I I I I 111 I I I I I I 11I 1 1 I I I 1111 I I I I I I 111 I I I I I I lx

10-3 2 L 6 8 10-2 2 L 6 8 10-’ 2 4 6 8 10° 2 4 6 8 101 2 4 6 8 102

Fig. 6. Characteristic impedances ZOeven and ZO cdd for shielded coupled striP transmission lin@-

If, again, we substitute w= – UI and w= – u, with the cor-

responding values z= — r and z= —r’, respectively, the fol-
lowing relationship is obtained for r’/r:

sinhz ~
r’ h
— — (28)
r 7TU1
sinh2 ——
h
‘%/’n[2dc0sh3
‘oro<z%”
It makes no sense to apply more accurate formulas for
Substituting this expression into (22) the following approxi-
graphical representations. In Fig. 6 the curves were calcu-
mations are obtained fOr Zodd:
lated from (26) and (29). Instead of the symbols U1 and u,, w
and s, the characteristic quantities introduced in [2] and
marked in Fig. 2 were used:
dsinhwsinh?+’
Z.dd
.3 s
U2=W+—) Ul=—. (31)
‘;’n [ 2 ‘@inhwsinh?-’ 1 ‘
2 2

Comparing the curves of Fig. 6 with those of Cohn [3], the

following em be recognized: in the middle area of the dia-
gram, which alone can be compared, there is agreement. In
the transition between the two Z ranges of the approxima-
‘odd
“:/’n[2u’sinhwsinh%l tions no break is to be seen. There are apparently no difficul-
ties in computing the curves even for extreme parameter
foro<zodd<~” values, as shown for very low and high ones, Just there the
accuracy of the approximations even rises and the error is far
By letting the value UI in (26) approach zero, twice the below l&3. Roughly the following is valid: starting from the
value of the characteristic impedance of the microstrip line impedance Z= 307r ohms and multiplying this value by 2N
(see Fig. 5) must be obtained. By substituting b = 21A the fol- where N=O, -11,-12, . . 0 yields impedances with a greatest
lowing simple impedance is obtained for the line of Fig. 5: relative error of the order of ( 10–3)2’N’. The computation can
IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. MTT-17, NO. 5, MAY 1969 265

be carried out by means of small digital computers or with the whole diagram. In this case it is favorable to use one of
less accuracy even on the slide rule. the approximations for K/K’ given in ( 15), (17), {19).
As in the above example, other cross sections given by
polygons with alternating electrical and magnetic walls can
be transformed by Schwarz–Christoffel transformation into REFERENCES

a straight line (x axis). Taking the approximations for the [1] W. Hilberg, “Uber die Moglichkeit, gewisse Wellenwiderstands-
coplanar line, the usual next step of transformation into a formeln, die efliptische Integrale enthalten, durch Naherungs-
formeln beliebig hoch wahlbarer Genauigkeit zu ersetzen,” Arck.
rectangle leading to expressions K/K’ can be omitted. When Elek. Ubertragurzg, vol. 21, no. 11, pp. 603-616, 1967.
using the algorithms, the elliptic integral theory is obviously [z] ———“Naherungen fur die elliptische Integral-Funktion K/K’ und
not needed at any point even for arbitrarily good approxima- Rek&sionen zur beliebigen Verbesserung ihrer Genauigkeit,”
Z. Arzgew. Math. Pkys. (to be published).
tions, i.e. for exact values.
[3] S. B. Cohn, “Shielded coupled-strip transmission line,” IRE Trans.
In many cases of practical applications there is still Microwave Theory and Techniques, vol. MTT-3, pp. 29-38, October
another problem. In the literature a formula is given for the 1955.
characteristic impedance of a transmission line in the con- [4] W. Hilberg, “Die Darstellung gewisser fur die Berechnung von
Leitungen benotigter elliptischer Modulfunktionen durch Rekur-
ventional form K/K’ with moduli k and k’, and one should sionsformeln,” Arch. EIek. Uberfragang, VOL 21, no. 9, pp. 451-
like to obtain good numerical values rapidly or to survey 459, 1968.

Mode Coupling Between Surface Wave

Transmission Lines

KENNETH R. COOK, MEMBER, IEEE, AND TEH-MING CHU

Abstracf—Tbe eigenvahre eqnation describing tbe mode coupling very weak coupling. The eigenvalue equations obtained by
between two Goubau surface wave lines is derived utilizing the point Meyerhoff and Tiazhelov were identical and had the limita-
matchmg method for solntions of boundary value problems. Excellent tion of requiring the radii of the surface wave lines be small
agreement is obtained between theoretical and experimental results. A
compared to the wavelength as well as the separation of the
method for obtaining the hybrid wavenumbers of coupled surface wave
two lines. Furthermore, their analyses were directed only to
lines is proposed for a finite set of TE and TM modes.
the lowest order TM mode structure.
HE TRANSFER of power between two parallel Goubau and Sharp [4] reported measured power transfer
Goubau surface wave lines has received little atten- between two Goubau lines. However, their data represented
T coupling over rather restricted values of separation. Bracey
tion in the past. Tiazhelov [1] considered the inter-
action of two surface wave lines considering only the lowest et al. [5] presented theoretical and experimental data on the
order TM modes. Under rather limited conditions, he was coupling between surface waves supported by dielectric
able to apply approximate boundary conditions of the sheets and rods.

Leontovich type and obtain reasonable agreement between The theoretical data presented in the present paper does
theory and experiment. His results were presented in terms not have the limitations of Meyerhoffs and Tiazhelov’s
of the hybrid space wave characteristics. However, for close analyses. Also, experimental data are presented for power
coupling conditions his results were somewhat in error. transfer over a much wider range of separation than previ-
Prior to Tiazhelov, Meyerhoff [2] used a somewhat different ously repolted [1], [4].
approach for obtaining the hybrid phase characteristics for We have directed our attention to the coupling between
two parallel surface wave lines capable of supporting an
Manuscript received September 25, 1968; revised December 26, arbitrary but finite number of TE and TM modes. By utiliz-
1968.This work was sponsored by the U. S. Department of Transporta- ing the method of point matching [3] a numerical scheme is
tion, Office of High Speed Ground Transportation.
generated which allows one to compute the coupled phase
K. R. Cook was with the Institute for Telecommunication Sciences,
Environmental ScienceServicesAdministration, Boulder, Colo. 80302. velocities for a set of surface wave structures supporting a
He is now with the Department of Electrical Engineering, Colorado finite number of modes. Limitations of the process are those
State University, Ft. Collins, Colo.
inherent to the point matching method and computer capa-
T. M. Chu is with the Institute for Telecommunication Sciences,
Environmental ScienceServicesAdministration, Boulder, Colo. 80302. bilities.