Radar System Engineering

40 THE RADAR EQUATION [SEC.2.10
taining the signal. If the sweep length were equivalent to, for example,
500 intervals (remember that what we call an “interval” is about I/@
long), one would expect to find several (roughly, five) noise peaks in the
sweep which were higher than the signal peak. In this case we should
certainly need either a stronger signal or more integration.
Now it is the trend of th6 numbers in Table 2.1 which is of interest to
us, since any real radar problem will differ in many particulars from the
ideal process to which Table 2.1 applies exactly. If we select any column
(w = constant) we observe that the variation of signal power with n is
something intermediate bet ween 1/n and 1/fi. For large n and particularly
for w not too small, the variation is not far from l/~~. This
implies that doubling the number of sweeps integrated, other things being
equal, allows the signal power to be reduced by a factor l/~2. This
relation has been strikingly verified by the experiments of J. L. Lawson
and others (Vol. 24) in the detection of signals on the A-scope, under
conditions where n was large. On the other hand, for small n, and
especially for very small values of w, Table 2.1 would require the signal
power to vary more nearly as I/n. This is not too important in practical
radar design, for so many factors are involved in the real problem that we
cannot hope for, and do not need, a very precise answer to such questions.
We shall most frequently assume, in later chapters, that the required
signal power varies as l/~;.
It must be observed, to put the above considerations in proper perspective,
that the benefits of integration are not confined to the smoothing
out of thermal or purely random noise. A very important requirement,
in practice, is discrimination against isolated but powerful disturbances
such as transients from nearby electrical apparatus, or, very commonly,
pulse interference from other radar sets. Indeed, the frequency of such
disturbances, in most locations, renders academic our earlier remarks
about the likelihood of getting an abnormally high noise peak once an
hour or so. The fact that the desired signal occurs repeatedly allows
such isolated disturbances to be discarded easily, or disregarded. Strictly
it is not integration, but a sort of coincidence selection, that is most potent
against these scattered flashes of interference. Such selection is inherent
to a greater or lesser degree in nearly all radar systems and as a rule very
little selection suffices. In practice then, it is still the noise that we have
to combat, not the interference, if we want to reduce the minimum
detectable signal power.’
I Becausethis is true, an essentialpart of any test of the condition or quality of a
radarsetincludesa measurementof thenoiselevel,or a quantity proportionalthereto.
The determination of minimum discernible signal power under any reproducible
conditionsof observationconstitutessucha measurement.