Data Corporation's Series 610 four- track incremental digital cassette 

 tape recorder. The complete package, which included all the electronic 

 circuitry for the individual transducers plus the tape recorder, was 

 housed in a watertight, 6-inch-diameter PVC cylinder 5 feet in length. 

 The system was designed to be operated manually or in a completely 

 automated mode, thus requiring only periodic tape changes (Fig. 26) . 



In its automatic mode, the system was activated when the windspeed 

 reached or exceeded a preset value and stayed there for at least 1 min- 

 ute. At this point, a single 17-minute sample of all the inputs was 

 taken. Each 68 minutes following this, another 17-minute sample was 

 recorded if the wind was still above its preset value; if not, the system 

 was shutdown lontil the windspeed increased. Each 17-minute record con- 

 sisted of 2,048 samples, taken at 0.5-second intervals, of all 13 chan- 

 nels plus a clock channel. Twenty-five of these records could be record- 

 ed on a single cassette tape. 



8. Data Processing and Analysis . 



The initial step in the data handling was to transfer the data from 

 the individual cassettes to seven-track magnetic tape by means of the 

 Sea Data reader. These tapes were then converted to a computer compati- 

 ble format on the University of Washington's CDC 6400 computer. The 

 histograms for all records plus the basic statistics, i.e., the minimum, 

 maximum, mean values and standard deviations as well as the transmission 

 coefficients based on these standard deviations, were then computed and 

 tabulated (App. G) . A digital filter, with a cutoff frequency of 0.05 

 hertz (Gold and Radar, 1969) was applied to the transmitted wave data 

 prior to these tabulations to remove tidal drift. The transmission 

 coefficients given in these summary sheets are a ratio of the standard 

 deviations for the transmitted and the incident wave gages . 



In the initial conversion, the data were checked for reader errors. 

 These points were smoothed using a linear interpolation between the pre- 

 ceding and the following good data points. Following this, the data were 

 checked for extreme values. Data points departing from the mean by more 

 than five standard deviations were smoothed in the same manner as were 

 the reader errors. In no case did the number of errors warrant elimina- 

 tion of a complete record (greater than six bad points). Record FH 11-1, 

 however, had bad data for channels 3, 4, 5, 7, and 8. This record was 

 run manually while calibrations were being made, and the affected chan- 

 nels were not connected properly at this time. The final edited data 

 were then stored on magnetic tape. 



The autospectra for all the wave data for all records were computed 

 with a more complete analysis of the force and acceleration data applied 

 to the more desirable events. 



Digital filtering techniques were used prior to spectral analysis 

 on all the wave and force data. The procedures used follow those given 



54 



