106 



The analysis of a passive listening observation may consist of determin- 

 ing the power spectrum, if the sound is approximately random noise or a steady 

 state sound. This may readily be done with a series of band pass filters feed- 

 ing one meter in turn or, for multichannel operation, several meters simultane- 

 ously. However, the usual technique now employed is to play back the previous- 

 ly recorded observation through the appropriate filters to a meter. The design 

 of the meter is somewhat arbitrary, and there are engineering s'tandards which 

 it is frequently desirable to follow so that the results are readily to be compared 

 with others. The nature of the sound will determine the design of the filters 

 required for its analysis. For many ambient noise studies rather broad band 

 pass filters suffice, whereas for some problems it may prove worthwhile to em- 

 ploy exceedingly narrow band filters to reveal significant detail such as fine 

 lines in the spectrum. In general each problem has its own filter requirements 

 and the acoustician should be prepared to use a variety of filters in the explora- 

 tory phases of a problem. 



If the sounds are transient in character, or if they vary in composition 

 even rather slowly with time, it is highly desirable to conduct the frequency an- 

 alysis so as to reveal these fluctuations. This art has been intensively used 

 in geophysical exploration and has already been discussed by Dr. Raitt. There 

 the usual technique has been to employ a multichannel amplifier with band pass 

 filters so chosen as to reveal the significant details of the spectrum for the par- 

 ticular problem. Recording is commonly done with a multichannel oscillo- 

 graphic camera. In some instances sufficient information can be recorded with 

 direct recorders using pen and ink or heat or spark techniques. After trying 

 all three of these techniques we tend to prefer the heat recorders where photo- 

 graphic recording is not required, since the hot wire pens are less of a main- 

 tenance problem than the pen and ink. However, good results with spark re- 

 corders have been obtained recently. 



The sound spectrograph developed at the Bell Telephone Laboratories 

 and produced by the Kay Electric Company as the Sonagraph and Vibralyzer are 

 somewhat in a class by themselves as analysis tools for revealing the detailed 

 variations in the spectrum of a sound with time. While the importance of these 

 instruments to underwater sound analysis merits a detailed description, the 

 basic design is thoroughly described in the literature (Koenig, 1946). The com- 

 mercial units are designed to cover in some detail the frequency region from 

 2.5 cps to 8800 cps in several steps. The length of the sound sampling analyzed 

 in one recording varies inversely with the extent of the frequency ranjge from 

 2.4 to 20 seconds. The available frequency range can readily be extended by 

 playing the original record at multiples or sub-multiples of its normal recording 

 speed. The time variation of the spectrum is presented as relative blackening 

 on a spark recording paper. This qualitative presentation is supplemented by 

 a function of the instrument which presents the spectrum quantitatively at any 

 chosen time. It has not proven difficult to supplement the basic circuitry to 

 provide a function which takes the average spectrum over a time interval whose 

 length may be continuously varied over a considerable range. 



In the year and a half we have been using these analyzers they have prov- 

 en extremely useful in revealing frequency effects in shallow water transmis- 

 sion, bottom reflection phenomena and biological sounds. In passive listening 

 the sound spectrogram appears to have the same sort of range of usefulness it 

 has in the visible speech analyses of the Bell Telephone Laboratories, and for 

 much the same reason; they provide an objective basis for comparing the qual- 

 ity of sounds. The spectrogram will be especially useful in identifying biologi- 

 cal sounds. In most cases the animals are not identified and it is often difficult 

 to reconcile the descriptions of these sounds made by different observers. 



