CARDIOVASCULAR SOUND 



687 



-160 -120 -BO -40 40 



CYCLES PER SECOND 

 FREQUENCY RELATIVE TO BAND CENTER 



FIG. 4. The pass-band characteristics of five filter setups tested 

 to determine empirically the optimum system for study of 

 heart sounds. Defined in terms of width at 6 db attenuation 

 (indicated by the horizontal line), these are /I = 9 cps, B = i6 

 cps, C = 23 cps, Z) = 34 cps and £ = 71 cps. [Courtesy of 

 Williams & Wilkins (12).] 



consecutively higher frequency bands. In fact, the 

 vibrations being analyzed are in toto progressively 

 altered in frequency and passed over a single constant 

 filter. The process of "heterodyning" is performed at 

 the approximate level of 15 kc. The filter setup of the 

 heterodyne type can be understood by comparison to 

 the tuning arrangement of the ordinary radio which 

 uses the same principle. By turning the dial for 

 "tuning in" stations one changes the tuning of the 

 oscillator in the radio set. When the sum of the 

 oscillator and the pertinent sound component is in 

 perfect resonance with a particular frequency, the 

 sounds at that frequency are picked up. Only a station 



operating in the same frequency band as that at which 

 the set is operating will come through clearly. 



In the original investigations of the applicability of 

 sound spectrography to cardiovascular sound (1953) 

 five filter systems (i.e., five heterodyne modifications, 

 according to effective width of the band-pass) were 

 tested. The band-pass characteristics of each filter 

 system are shown in figure 4. The width of each band- 

 pass at 6 db attenuation is approximately 9 (^4), 16 (5), 

 23 (C), 34 {D) and 7 1 cps [E). It is evident from com- 

 parisons such as those shown in figure 5 that the 

 recordings produced by the five diff'erent filter sys- 

 tems are quite different. The filter system with narrow- 

 est pass-band by definition gives finest frequency 

 definition but is slower in onset and offset, giving 

 excessive temporal lag, improper delineation of 

 transients, and "after smudging." The filter system 

 with the widest pass-band characteristic gives mini- 

 mum time lag and minimal ringing, but also, of 

 course, by definition gives least detail in the frequency 

 dimension. It was concluded that filter C represented 

 the best compromise for display of both the temporal 

 and the tonal characteristics of cardiovascular sound. 



The use of a single variable filter has the theoretical 

 disadvantage that the Q^, or characteristic with respect 

 to ringing, varies with frequency. The temporal error 

 increases at higher frequencies. It is doubtful that 

 error significant to spectral phonocardiography is 

 introduced, however. 



A second major type of filter system which was 

 explored was the phase filter, which has the advantage 

 of providing a cleaner display of musical tones than 

 does the heterodyne system. For example, the har- 



l440^ 

 '|200- 



A 



9«0- 



TlME (sec) 



FIG. 5. Crescendo or late systolic murmur displayed by the five filter systems. Pass-band C is 

 considered the best compromise for routine use in the study of cardiovascular sound. [Courtesy of 

 Williams & Wilkins (12).] 



