12
negligible matter. A desirable aim would be to achieve
uniformity of presentation as in some other disciplines,
e.g. in electronics, frequency response curves are almost
always presented on log/log co-ordinates. In the present
case there is no clear argument for any exclusive choice
of co-ordinates; but agreement could probably be reached
on one as the normal and preferred method. Since differ
ent applications of the MTF may call for different co
ordinate systems, presentation of data in tabular form
would assist the serious user.
Symbols
Numerous different symbols are employed, thus
spatial frequency is denoted by S, N, K,t^etc. A uniform
symbolism should be agreed upon, in consultation with lead
ing optical laboratories.
Averaging over Field
Averaging the MTF over the field of a lens presents
just as many difficulties as averaging the resolving power,
indeed the difficulties are greater inasmuch as the resolving
power figures are replaced by curves. Also, phase shifts,
which in effect are taken into account in a resolving power
figure, are presented separately in an OTF measurement, and
there is no obvious way of averaging them. The averaging
problem is bound up with the general problem of the inter
pretation of MTF's, and it may be that the ultimate solution
will be found in the translation of MTF data into relative
contrast data for a specified target as discussed later in
this report.
One approach, suggested by Back, is to present maps
of the camera field showing lines along which the MTF is at
a constant value for some given spatial frequency. This
approach has the danger that the contrast information may be
interpreted as applying to general target detail and not
merely to sinusoidal targets.
It may well be that no reasonable method of averaging
MTF's over the field can be found until a method of reducing
each MTF to a single figure of merit has been agreed upon.
Similar remarks apply to phase data.
Accuracy of Present Measurements
It is commonly believed that measurement of the MTF of
a lens is much more accurate than measurement of its resol
ving power. This is not necessarily true, in practice,
(though certainly true in principle) and there is very little