The particular process generated by the analyzer system of the 

 SEADAC for realization of the spectral density function has been 

 described. It should be noted, however, that certain errors, which 

 arise in computation, may cast some doubt on the validity of results. 

 These errors are not large but one is still obligated to establish 

 confidence in such a new computing machine even if its principles 

 are well-known and accepted. Some of the questionable features of this 

 electronic analog computer are: (1) the creation of a discontinuity in 

 the time history of the event being analyzed when it is joined to itself 

 in a loop (splice effect), (2) the use of a nonideal filter (i.e., one 

 which is not a perfect rectangle), and (3) the method of calculating 

 running averages over the effective bandwidth of the filter as if the 

 energy were constant over that bandwidth. 



These deviations from an ideal method of analysis may be treated 

 independently (as in the last section), to assess their individual 

 effects on the outcome. It is, however, more profitable to examine 

 the effect of the aggregate of all these errors; if the total error 

 is sufficiently small, there is no reason to pursue the matter further., 

 The best method for verification of SEADAC operation is a comparison 

 of the output of the SEADAC with the known and understood results of 

 a general purpose digital computer, wherein the ntraierical analysis is 

 governed by an entirely different computational procedure than is the 

 analog method. Instead of filtering, squaring, and averaging, the 

 numerical method deals with a convolution of the original time history 

 with itself, and then a Fourier transformation generates the spectral 

 density function. The point is that the numerical technique is so 



16 



