Breaker and O'Neil 



to a record, together with the time of the pulse. The speed of this 

 process is still limited by the rate at which the digital data logger 

 can record the final computer-compatible tape--500 frames per second. 

 To reduce digitizing time the 5.3 kHz tapes are sampled only at 

 2.65 kHz and the 9.0 kHz tapes at 3.0 kHz. Even so, a total of more 

 than 500 hours of digitization time will be required from first to 

 last. 



Pulses are detected by a modified zero-crossing method. A 

 high-frequency detection square wave, of sufficient amplitude to 

 exceed the average analogue-tape noise level, is added to the 

 analogue data. Changes in the high frequency zero-crossing pattern 

 signal a detection. As presently set, this system will accept 

 virtually all "good" data pulses but will also let some noise and 

 bottom-reflected pulses through. These are later eliminated by the 

 computer, making use of the constant 2-second keying interval. 

 The program uses a pushdown-list sieve to select subsequences of 

 records having a constant 2-second separation. These methods have 

 worked very well with the short and medium-range events. Further 

 refinements may be necessary to deal with the long-range events, 

 however. 



After this rejection of noise and bottom-bounce pulses, 

 the pulse amplitudes are computed. Figure 4 shows two actual pulse 

 envelopes and one schematized version of a pulse envelope. "D" is 

 the portion due only to direct transmission, and "I" is the portion 

 due to the interference between surface-reflected and direct trans- 

 mission. Where "D" is sufficiently broad and well defined one can 

 calculate the height of a hypothetical direct-transmission pulse by 

 using only those envelope points falling within "D". At long 

 ranges, however, "D" becomes very narrow. Moreover, pulse distortion 

 becomes more serious and S/N ratios become worse at these ranges. 

 Thus, especially at lower frequencies, estimation of a "direct" 

 pulse height may become sheer guesswork. Only the height of the 

 whole D+I+S pulse can be computed. 



Although this is inconvenient, information about the 

 purely turbulence-caused fluctuations may still be extracted from 

 such data. Clearly the lag between direct and surface-reflected 

 arrivals can be small only if the paths lie very close together. 

 Since the reflection at the surface should not (for the low sea 

 states encountered and the frequencies used) cause any significant 

 change in the structure or frequency of the pulse the surface- 

 reflected pulse should be affected by turbulence very much as its 

 direct brother is. That is, the "D heights" and the "S heights" 

 should have virtually identical distributions and spectra. 



Because of the relatively short wavelengths used, however. 



378 



