fghijk

Practical Considerations for Measuring Roundness.

Introduction.

In a previous article I described a number of approaches for measuring roundness. The methods
considered included the use of callipers, dial gauges, co-ordinate measuring machines and dedicated
roundness instruments. In this article I present some of the practical considerations for measuring
roundness. Whilst the article is written from the perspective of using a dedicated roundness instrument,
many of the considerations will be equally applicable regardless of the measurement method employed.

In order to focus on the more general problems of precision measurement I have assumed that the
instrument is calibrated etc.

The scale of the problem.

Components measured on roundness instruments vary tremendously in terms of size. At one extreme
one might be concerned with the measurement of a hypodermic needle, whilst at the other extreme one
might be interested in measuring the roundness of a large bearing or engine component that could be
many hundreds of millimetres in diameter. As we shall see later in the article the radius of the
component and its general size and structure have important bearings on the measurement of
roundness. However the first thing to address is the scale of what we are interested in when measuring
roundness.

Roundness is defined in the ISO standards as the radial separation of two concentric circles that just
enclose the profile of the component. The majority of components measured on standard roundness
instruments would be contained within a cylinder of 100mm diameter and 100mm height. By
comparison with the size of the component, the deviations in roundness are usually smaller by several
orders of magnitude. Roundness tolerances of less than 1 micron (0.001mm) are quite common. For
high precision components the required tolerance might be an order of magnitude smaller, meaning that
the scale of the roundness tolerance would be 6 orders of magnitude smaller than the diameter of the
component.

Most people can easily visualise what a millimetre looks like. However, visualising a micron is a much
harder problem - mainly because it is below the lower limit of visible sizes. A micron is a truly
microscopic size - if you want to see things on the micron scale, you will need to use a microscope.
We can however gain an appreciation of the size of a micron by comparing it with things that we are
familiar with. This is quite useful because some of these things might interfere with the measurement!
Typical 80gsm photocopier paper for example is about 100um thick. Food wrapping film is typically
about 10um thick. Fine dust particles range from a few tens of microns to a few tenths of a micron, as
do bacteria! The grease film left by a fingerprint might well be a few tens of microns thick. Particles
of tobacco smoke range from about 0.01um up to 1um.

Contamination.

From the above discussion on scale, it is easy to see that the first consideration for measuring a
component for roundness at the sub-micron level is to ensure that it is clean! Dust particles, grease
films and the like can swamp the results that we are looking for. As a rule, if you can see something
on the surface in the area that you are about to measure, then this will probably affect your results
significantly. Particulate contamination will often show as a spike. When using a stylus based
instrument, this might well get pushed out of the way by the stylus. Often dust particles or grease will
transfer to the stylus tip itself, so it is important to make sure that this is also kept clean. Grease films
tend to alter the overall roundness but do not necessarily have any definite characteristics. However,
repeated measurements will tend to show a change in roundness as the stylus tip cuts through the film.

a division of Page 1 of 5
Temperature.
fghijk
It was noted above that the roundness of a component might be 6 orders of magnitude smaller than its
dimension. One obvious problem that this creates is that of thermal expansion. This is more normally
considered as an issue when measuring dimension, however if the component is not homogeneous (for
example it might have a hole in it), then clearly differential expansion should be considered as a
possible source of error. It should be noted that the expansivity of steel is about 10 parts per million
per degree Celsius. This means that differential expansion might well affect the shape of the
component at the level we are measuring. In order to avoid this, it is necessary to ensure that the
component, measuring instrument and environment are in thermal equilibrium. This is often done by
allowing the component to acclimatise to the temperature of the measurement room before it is
measured.

Of course temperature effects can also cause problems for the instrument. In particular the gauges used
on these instruments tend to be sensitive to changes in temperature. It is important therefore to ensure
that the temperature does not drift during the measurement period.

Component handling.

Excessive handling of the component may cause problems in a number of areas. Apart from special
requirements due to the fragile nature of a component, the two prime considerations are contamination
(e.g. fingerprints) or induced temperature effects. These can be reduced by ensuring proper handling
procedures and minimising component handling.

Environmental effects.

Temperature is an obvious environmental condition that we should consider. As we have discussed

this above, I will not address it further here. The other main environmental factors to consider are
draughts and vibration. Both of these might adversely affect the measurement, and should be limited
as much as possible. The use of dedicated measurement areas or environmental enclosures will greatly
help in dealing with draughts. Vibration is largely a matter of instrument siting. The instrument should
be sited well away from sources of vibration such as large machinery, air-conditioning/ ventilation
ducts and preferably as far away as possible from busy roads and train-lines, flight paths etc. Most
instruments will contain some form of passive anti-vibration system. This will help reduce the effects
of vibration in certain frequency ranges. Where this is inadequate, additional active anti-vibration
systems or other isolation mechanisms need to be considered.

Work-holding.

In making a measurement, the component will have to be set-up on the instrument and effectively held
in some way. It might be that the component is of sufficient mass to allow it to be held by its own
weight. Placing the component directly on the work table of the instrument may not be sufficient to
ensure a stable measurement condition. Often the interface between the component base and the
instrument’s worktable may not be sufficiently stable, leading to rocking or movement of the work
piece during measurement. This can be avoid by simple means such as the use of three small strips of
paper used to form a kinematic mount on which to rest the component.

Where gravity is not sufficient to hold the component, a simple fixture will be required. This might be
in the form of an adhesive compound such as modelling compound or wax. Caution should be taken
when using such compounds as their adhesive properties are only exhibited when warm. The possible
temperature effects have been dealt with above. The other prime problem with these compounds is that
as they cool they tend to creep, thereby influencing the measurement. This can often be accounted for
by taking repeat measurements until such time as the values have stabilised.

a division of Page 2 of 5
 fghijk
Conversely gravity might work against you by distorting the component under its own weight. In this
case it will be necessary to support the component in some way. Careful consideration should be given
to the location of the supports so as to minimise sagging of the component. Typically one might seek
to support the component at the Airy points in order to minimise this problem.

A simple three-jaw chuck or collet chuck might be an appropriate means to hold the component. When
using such a device, it is important to avoid over-tightening the chuck as this may cause distortion of
the component itself.

The component.

Before looking at the errors introduced by incorrect component set-up, it is appropriate to note that the
component surface might also affect the measurement of roundness. Very clearly any surface defects
that might have been considered as surface finish or waviness will affect the overall roundness
measurement unless these can be discounted. This can be done by selection of an appropriate filter, but
it should be borne in mind that surface finish on the component might well affect the repeatability of
the result.

Instrument and component set-up.

In this section I shall look at some of the problems associated with setting up the measurement on the
instrument.

Setting up the gauge.

In order to prepare for measurement, the gauge will need to be set to the correct attitude
(vertical or horizontal) and orientation to access the feature to be measured. In moving the
gauge it is important to avoid excessive handling or even unnecessary movement of the gauge,
in order to ensure that all components in the measuring loop remain in or quickly achieve
equilibrium. Clearly unnecessary handling of the gauge will lead to induced temperature
effects.

The next stage is to ensure the correct stylus is being used. It is important to remember that
the stylus provides the mechanical interface between the component surface and the
instrument. The choice of stylus dimension, material, angle of presentation etc all have a
potential effect on the results. Again care should be taken when changing the stylus to ensure
that the disturbance to the measurement loop is kept to a minimum. Removing and inserting
a stylus puts a strain on the pivot mechanism. This is a precision mechanism designed to
allow free motion of the stylus at very low forces. Use of excessive force can damage the
mechanism.

Stylus crutch angle.

On many gauges, the stylus can be moved through a range of angles to allow components to
be addressed. This is usually achieved by some form of moving crutch mechanism. This is a
friction mechanism, and again excessive movement should be avoided. It is important to
calibrate the gauge with the stylus at the crutch angle that is going to be used for
measurement. Where this is not possible, it is important to use the same crutch angle when
making measurements of a particular type so as to avoid introducing cosine errors. Normally,
for modest angular differences, these errors are small and can be accounted for in the
uncertainty budget for the measurement.

a division of Page 3 of 5
 fghijk
Cresting the stylus.
The term cresting can be applied to more than one subject. The primary concern for a
roundness instrument is that the plane of movement of the stylus should pass through the
spindle axis. The distance between the actual gauging plane and the spindle axis is referred to
as the cresting error. On some instruments the cresting is achieved by setting the column
orientation. Variability in gauge position or concentricity of the stylus tip to the stylus shank
will affect the cresting, and these have to be controlled by careful manufacture. When the
stylus is used horizontally, it should be set so that its tip is in line with the spindle axis.
Where this is not possible, it will lead to a small error in the measurement. The error is related
to a number of factors including the eccentricity of the component from the spindle. Where
the eccentricity is negligible, the error caused by cresting will also be negligible and will
simply result in a slight phase error. However, if the instrument is being used for dimensional
measurement (i.e. radius measurement) the cresting error should be taken into account in the
uncertainty budget for the measurement.

Component set-up.
Roundness is concerned with the measurement of a two-dimensional form. Real components
are of course three-dimensional. Because of this it is necessary to eliminate variations in the
gauge position caused by variations of the component in the axial direction as defined by the
spindle axis. Put simply:
y if the component is a sphere it should be measured at the equator of the sphere
y if the component is a cylinder then the cylinder must be co-axial with the spindle
before measuring
y If the component is not a sphere or a cylinder or presents a feature that acts like one
of these, one cannot measure roundness - at least, not directly.
That list might seem restrictive, but that’s what roundness is. For other measurements it is
customary to take proxy measurements. A good example of this is the measurement of
hemispherical calibration artefacts. These artefacts are measured for “roundness” at a position
above the equator (which cannot be reached). The measurement is made, by convention, with
the stylus angled so that the stylus shaft is tangential to the surface at the point of contact.

A tilted cylinder poses an interesting problem. Theoretically if you slice the cylinder you will
see an ellipse, and it is assumed that if you measure the cylinder on a roundness instrument
then that is what you get. True enough, the form will be an ellipse - but is it the one you
thought it would be? By making several measurements along the axis of the cylinder it is
possible to establish the axis and therefore the relative tilt of the cylinder. By projecting the
data onto the plane normal to this axis you should get circles. This is not quite the case in
practice because the stylus has a finite size and the contact point between the cylinder and the
stylus tip changes with the orientation between the cylinder and the stylus. The error caused
by this effect is very small, but it can be noticeable if the tilt is large enough. This source of
error can only really be dealt with by levelling the cylinder accurately.

Eccentricity is a major concern when measuring roundness, particularly for high-precision

components. The reason for this is because measurements of the surface are made at equally
spaced angles based upon the position of the spindle, not the component itself. If the
eccentricity is negligible, then there is no difference and the measurement is a true reflection
of the roundness of the component. If the component is not concentric with the spindle, the
angular information is distorted and this affects the roundness value. The error due to
eccentricity is proportional to the eccentricity and inversely proportional to the component
radius. This means that the eccentricity becomes more critical as the radius of the component
decreases.

a division of Page 4 of 5
 fghijk
Summary.
In this article I have outlined a number of the main practical considerations when measuring roundness.
It has been seen that in order to get high-accuracy measurements it is essential to consider all parts of
the measurement process from preparing the component and instrument to setting up the measurement
itself.

Of particular concern are:

• Component cleanliness
• Component fixturing
• Stylus angle
• Environmental conditions
• Eccentricity.

Mike Mills, 14/9/06.

a division of Page 5 of 5