30 Lecture 2 
The approach to quick and reliable classification is not so obvious. The 
combination of (a), (b), and (d) above evidently gives rise to the possibility of 
an accurately delineated visual display of target shape; but in typical operational 
situations the ratio of the number of unwanted (or irrelevant) targets to the 
number of wanted targets is so large that the human task becomes too great, 
and reliability suffers. In the corresponding radar case, this ratio is, in con- 
trast, usually very small, and automatic detection, holding, and tracking of tar- 
gets is already feasible. Progress in this direction, however, for any but the 
most simple of sonar situations is likely to be difficult and slow. One step toward 
more reliable detection and classification is the improvement of the amplitude 
ratio of genuine signals to random signals by display integration methods, and 
we have given much attention to this. The effect of shape and size of the target 
display on the ability of an operator to detect and recognize it is another very 
relevant matter which has been under investigation. 
Classification by knowledge of frequency response of targets (c above) is 
very problematical at present, but we are trying to develop a wide-band system, 
having a 10 to 1 frequency band and constant directivity over this width, which 
can display this information and which will enable us to assess the potentialities 
of the method. 
Propagation of underwater acoustic signals is obviously a fundamental matter 
requiring continuous research in order that system improvements may be co- 
ordinated and exploited. Long-range propagation studies are beyond our facilities, 
but we have some interesting results on short-range echo formation which point 
to other limitations on the use of improved systems. 
2.2, ARRAYS AND DIRECTIONALITY 
Throughout this section, directionality will be considered in terms ofa line 
or strip array, so that only two-dimensional directional patterns are involved. 
2.2.1. Multiplicative Arrays 
In the search for higher directionality for a given size of array, the replace- 
ment of addition by multiplication in the combining of signals from different 
sections of the receiving array has some very definite attractions [1,2]. For 
instance, if a strip array is divided intotwo equal sections, and the output signals 
from these (assumed of narrow bandwidth) are multiplied together, the output 
of the multiplier comprises two parts (as far as the wanted signals are con- 
cerned), one a dc signal and the other a signal at twice the original frequency. 
The amplitudes of both these parts are functions of the direction from which the 
signal was received, although the double-frequency part is less sensitive to 
direction than the dc part. The directional response of the dc signal, plotted as 
the variation of amplitude as the array is rotated in the plane of its length, has 
a main lobe, or beam, which is only half the width of the main beam of the same 
array when used in the normal manner, i.e., with the signals from its sections 
merely added together. It is therefore attractive to regard multiplication (with 
a low-pass filter to remove the high-frequency output) as a way of doubling the 
directivity of the array. There are other advantages, too. In the ordinary (addi- 
tive) system, the output of the array is at signal frequency, and it is necessary 
