ECHOES AND TARGETS 315 
where /’; is the fraction of time that the system is on 
target during the scanning procedure. Ordinarily this 
scanning loss amounts to approximately 10 db in an 
average radar system, requiring a signal perhaps 10 
times as large as the necessary amount for detection 
while searchlighting. It is important that this formula 
be used only. where scan-to-scan integration takes 
place. 
Discussion 
While this paper has specifically been limited to 
noise considerations, it seemed reasonable to hope that 
the same general considerations could be applied in 
determining the visibility of signals in various types 
of clutter, in particular the simpler types which are 
echoes from rain and snow. If the mechanisms involved 
were more thoroughly understood, the fundamentals 
of the problem would be understood too and could be 
put together in a coherent form. 
The shape of the response curve has been considered 
by the author and is known to have some effect, but 
the experimental approach to various shape factors 
has been rather limited. In the work presented here 
the response curve of the receivers involved has been 
that of a so-called double-tuned circuit, whose ampli- 
tude response is proportional to 
[eae 
where w is the frequency difference between the fre- 
quency under measurement and the center of the band. 
@ is the % bandwidth. The difference between this 
response curve’s performance and that of a multiply 
narrowed, synchronously tuned, intermediate ampli- 
fier, which has Gaussian response, was not observable 
experimentally. Theoretically also, there is little dif- 
ference. It is felt that the considerations may not apply 
in extreme cases of sharp-edged amplifiers or in single 
single-tuned circuits but that in other cases the same 
answers do apply. 
The question was raised as to the dependence of 
signal threshold on pulse recurrence rate. In all the 
other parameters the visibility of the signal is pro- 
portional to the signal energy. The author found that 
for a given average power the visibility is distinctly 
better if you concentrate more energy into each pulse 
and separate the pulses by longer intervals. In other 
words, the threshold is proportional to the energy per 
pulse but inversely proportional to the square root 
of the repetition frequency. This settles a disagree- 
ment between two groups, one of which believes visi- 
bility would be found independent of pulse repetition 
rate and the other that it depends on average energy. 
The answer lies between the two views. In this matter 
of visibility it is interesting to recall that the first suc- 
cessful radar, which was giving ranges up to 25 miles 
in 1936, had a receiver bandwidth of about 200 ke and 
a pulse length of 5 psec, a combination which lies on 
the peak of the maximum visibility curve. The pulse 
length on the radar screen Was about 3 mm. The curve 
for optimum visibility peaks at 1 mm and does not 
decline very rapidly for longer pulses, so that, too, was 
near the optimum value. The first production radar for 
use in the fleet had a pulse length of about 3 psec and 
a bandwidth of about 300 ke, which is again_on the 
peak of the visibility curve, and its visible pulse was 
about 2 mm long on the screen. This was of course not 
entirely accidental but was fortunate, nevertheless. 
The preproduction model of this radar was built in 
1938. 
The author discussed the effect of fluctuating sig- 
nals in scanning as distinct from the steady signals 
which had been employed in the experiments described. 
In the case of signal fluctuation, it is necessary first to 
define the signal amplitude in such a way that analy- 
sis is applicable. Employing the average value as a 
criterion, the visibility of fluctuating signals may ac- 
tually prove greater than for steady signals. If the 
peak value of a fluctuating signal is taken as the signal 
threshold power, the visibility is probably poorer than 
for a steady echo, but it is felt that the result would 
be again essentially independent of the scanning speed, 
as long as it is high enough to cause pulse-to-pulse and 
scan-to-scan integration. The limit, however, may 
occur at 20 rpm instead of 10 rpm. 
One variable has been omitted which has proven 
puzzling. This is the target speed. What really con- 
stitutes a scan-to-scan integration ? If the target moves 
the distance of one spot diameter in a scanning period, 
is it still integrated ? It would seem to be so integrated 
provided the observer is able to perform as an aided 
tracker, i-e., can appreciate a change in linear motion. 
If it is not integrated, one would expect to find a 
difference in signal threshold depending on the target 
speed, probably in direct proportion to the square root 
of target speed. Some experiments have been made on 
simulated echoes of this variety, and there was some 
indication that targets of higher velocity are definitely 
harder to see, but this cannot be considered quantita- 
tively established. Signal fluctuation, however, is im- 
portant, and it is felt that, in general fluctuating tar- 
gets with cross sections defined on the basis given in 
the following paper are harder to see by perhaps 2 db 
but that this estimate is not affected by any arguments 
about scanning. 
There was another inquiry concerning the explana- 
tion of the Watson effect observed occasionally at very 
close ranges when the background noise was so large 
the normal signal could not be detected. This con- 
sisted of an inverted signal smaller in amplitude than 
the background noise which could be observed to a 
range of almost 100 yd in sets which had a direct wave 
extending to 3,000 to 4,000 yd. In these cases the signal, 
instead of appearing as a small inverted V, showed up 
as a small upright V, approximately 1% the amplitude 
of the initial noise. This effect had been often reported 
