- 125 



It was discovered that this modification was very important. 

 The results of the analysis of DPG data varied considerably between a 

 17 minute long record with an effective sampling interval of 0.5 

 seconds compared to the analysis of an 8.5 minute long record with 

 0.25 second sampling. Further, there are considerable differences 

 between the analysis results of the first 8.5 minutes of a record 

 compared to the last 8.5 minutes. This variability may be due to 

 sampling and the confidence limits on the spectrum. Figures VI-6 

 through VI-8 are copies of the analysis output for one data set 

 analyzed with a 17 minute record, the first 8.5 minutes of the 

 record, and the last 8.5 minutes, respectively. The 17 minute record 

 with 0.5 second sampling interval block-averages 8 adjacent frequency 

 bands compared to A bands for the 0.25 second sampled records. The 

 spectra developed from the 17 minute long record, then, is more 

 stable than for the 8-5 minute long record, yet retains the same 

 frequency resolution. Figure VI-9 and Figure VI-10 illustrate the 

 energy and non-smoothed peak-energy directional spectra for each 

 case. The analysis results most closely match other observations 

 made at the FRF when the full 17 minute long record with effective 

 2-Hertz sampling is utilized. Table 6.3 lists the results of each 

 type of analysis and compares the results with independent FRF 

 observations over a period of four days. 



