density appears to go to zero. The chance of cutting off a low frequency 

 band of swell which might actually be present cannot be ignored, nor can 

 much be done about it since it is inherent in the system to propagate 

 large errors at low frequencies. 



Aside from the protective measures taken to prevent erroneous infor- 

 mation from appearing at to = , and assuring maximum measurement accuracy 

 with the low-pass filter, there is little that can be done to establish 

 confidence in the estimated spectrum of the wave except to compare results 

 obtained from the SPLASHNIK with those obtained by a "reliable standard." 



Accordingly, a series of experiments was made in which the output 

 of the SPLASHNIK was converted into a wave spectrum by Equation [4] , and 

 this spectrum was compared with the wave spectrum resulting from measure- 

 ments made by other transducers (at the same time and physically close by) 

 which are considered to be fairly reliable standards. The philosophy of 

 this approach is simply that good agreement in spectral shape and area 

 will produce good agreement in prediction of the statistical character- 

 istics of the waves. Such a result would obviate the necessity for 

 further investigation of errors in the SPLASHNIK measurement system. On 

 the other hand, poor agreement would certainly indicate that further 

 study of the system is required. 



EXPERIMENTAL VERIFICATION 



Initial tests were made in the TMB deep basin where irregular long- 

 crested waves were generated with spectral peaks appropriate to wave 

 lengths of 15 and 20 feet respectively. The waves were measured directly 

 by a fixed capacitance probe and by the SPLASHNIK. The SPLASHNIK acceler 

 ation spectra were transformed according to Equation [4] and superimposed 

 on the wave spectra measured by the capacitance probe. The results are 

 shown in Figure 7. Although the individual spectral densities differ 

 somewhat, the areas are almost identical as evidenced by the rms values. 

 The spectral peaks are well identified in both cases. 



11 



