the presence of the other target, plus noise, when the target separation is that amount, i.e., 2OrmS. This definition bears no relation to a conventional resolution measure and I do not know how to relate it to this discussion. Dunn et al. [5] also base their conclusions upon tracking. The nonlinear operation of their system is made evident by its thresholdtype performance; at a separation of 0.75 beamwidth, they are completely merged, while at 0.85 beamwidth, they are completely separated [5, fig. 451. Rarely if ever do linear systems exhibit such precipitous behavior; hence it is diffkult to relate the 0.85 beamwidth statement to the foregoing discussion. In any event, the beamwidth unit used probably is the one-way beamwidth whereas the reported measurement was made with a two-way system. Thus the value which would correspond to my units would be 0 . 8 5 4 = 1.2 beamwidth.In summary, regarding Dr. Skolnik's first point, the differences in reported resolving power either vanish or diminish considerably when differences in definitions or criteria of resolvability are accounted for.In retrospect, I believe that the measure upon which I based my conclusions was somewhat conservative: a 3 d B or larger dip in the angular response of two targets 90 percent of the time (k = 0.707 and P = 0.9). A smaller dip (larger k) and a lower confidence level (smaller P) may be more realistic, raising g to perhaps as much as 0.5 -0.7. The higher value gives a resolving power of one beamwidth. This value of g is about twice the value used in the paper. However, even so large a change does not alter the basic conclusion of the paper, which relates to Dr. Skolnik's third point.His third point raises the question of the effect upon the resolving power conclusions of using a less conservative measure for grossly unequal targets. This is the important question. The grossly unequal target case was the thrust of the paper, for identification of objects in microwave images requires that wide dynamic range of target echoes be preserved. The effect is found by introducing a larger value of g into [ 1, eq. (lo)] which, for large target strength ratio b/a, gives the angular resolving power, when the signal-to-noise ratio is large, aswhere A/L is taken to be the nominal beamwidth, and n characterizes the aperture in terms of the number of successive convolutions of rectangular apertures by which it may be represented (n = 1 for a one-way rectangular aperture, n = 2 for a one-way triangular aperture and for a tweway rectangular aperture, n = 4 for a two-way triangular aperture, etc.). Over a wide range of target strength ratio (20 -40 dB), doubling g for n = 1 approximately halves the resolution. For n > 1, however, the effect is much less pronounced; for n = 4 it is a 10-percent change. Since n 2 in conventional monostatic radar, the theory as presented in the paper is not particularly sensitive to the choice of g. Synthetic aperture radar, on the other hand, is a different story. Unless a taper is introduced in the signal processing, n = 1 a...
