TR No. 22 
Computation of Auto Covariance Series and Energy Spectra 
The data processing was performed on the NUWS CDC 3200 digital 
computer. The FORTRAN programs are included (appendix B) for reference. 
The following were determined for each run and for i= 1, 2, 3, ..., N = 
number of square wave cycles in the run: 
1. The time t; from the start of the run (taken to be the start 
digital recording) to the completion of the ith cycle. 
2. The period T; of the ith cycle from 
West Vedic fon eG (16) 
3. The velocity u; for the ith cycle using the calibtation 
coefficient 
aot k 
? alk (17) 
The values of u. were assumed equally spaced at intervals of 19.61 cm. 
Each run was divided into samples of 500 values of welocity per sample; a 
computer printout of all of the digitized velocity data was obtained. Ex- 
amination of the data revealed that all except 7 of the 49 samples contained 
several obviously erroneous points. A section from the printout (run No. 2, 
sample No. 3) appears in table 2 which shows a typical series of values con- 
taining indicated erroneous points. 
The values of erroneous points were replaced with the values of the 
immediately preceding points. 
For each sample a straight line was fitted through the data by the 
least squarés method (18): 
UO) = a, Pp Ss; (18) 
13 
