View the original post on the IUSSCAA message board here

A Signal Processing Approach

By Bruce Rule - Dec 26, 2015


This posting further discusses the previously posted conclusion that "the more carefully you measure something, the more you find out about many other things."


PAC SOSUS failed to identify acoustic detections of the loss of the GOLF II Class Soviet SSB K-129 on 11 March 1968 because the acoustic energy associated with the precursor battery-associated explosions had durations of less than 0.5 seconds with important details at the 0.1 second duration level. The signals associated with the firing of two R-21/D4 missiles within their closed launch tubes had peak energy durations of less than six seconds. The signal-to-noise apparent from a single hydrophone detection at a range of about 700 nautical miles indicates all then operational West Coast arrays should have detected those events.

The SCORPION Analysis Group failed to confirm their conjecture that a signal "at about five Hz" was bubble-pulse energy associated with the collapse of the SCORPION pressure hull because they did not determine the frequency of that source more accurately. Refined analysis completed for the first time 40 years after the tragedy identified the frequency to have been 4.46 Hz which correponded to a collapse depth of 1530 feet and an energy release equal to the explosion of 13,200 lbs of TNT at that depth, a value that exceeded the combined yield of all conventional weapons aboard SCORPION. The more accurate measurement also would have refuted the Court of Inguiry final conclusion that SCORPION was lost "because of the explosion of a large charge weight external to the pressure-hull." The figure on the last page of chapter four of WHY THE USS SCPRION (SSN 589) WAS LOST illustrates the display from which the 4.46 Hz value was derived. The displayed time resolution of that figure is 0.002s or two milli-seconds.


Traditionally, the review of acoustic data for extremely short duration signals has involved aural analysis, i.e., listening to tape-recorded data. That approach will fail to identify transient energy present near or below the normal range of human hearing or that is aurally masked by energy present in other areas of the audio spectrum. Some of these "lost" transient signals - which could have durations of less than one second - may - with high temporal resolution analysis - be detectable at ranges greater than "conventional" sources.


Selected high-value data should be analyzed with temporal resolutions of at least 0.01 seconds. The normal processing trade-off (loss of frequency resolution for a gain in time-resolution) will not be of consequence because time-resoluion - in this context - is significantly more important than frequency resolution. The object of such processing is to determine if signals are present, not to characterize those signals in detail.