THE COMPANDOR 325 



rate of change is not faster than the usual syllabic envelope, is 1/w in 

 terms of db. In terms of voltage ratios the intermediate signal in- 

 tensities are proportional to the square root of their original values if n 

 equals 2, the cube root if n equals 3, etc. 



The ideal relations postulated above cannot all be met in the 

 physical design of the circuits. The indices s and t must be the dy- 

 namic characteristics of the tube and circuit and can be held to constant 

 value only over limited ranges of operation. Equation 2 is only ap- 

 proximately true as some space current is permitted to flow when no 

 speech is passing; otherwise, impractical values of control impedances 

 would be involved. However, they do serve to illustrate the func- 

 tional operation and can be approximated sufficiently well in com- 

 mercial equipment for useful amounts of compression and expansion. 

 Figure 6 shows experimental steady-state input versus output charac- 

 teristics for devices built to have a compression ratio of 1/2 and an 

 expansion ratio of 2. 



The compressor is seen to operate substantially linearly over a 45 

 db range of inputs and the expandor over a 22.5 db range. This is 

 about as much range as can be secured conveniently from a single 

 stage of vacuum tubes. As such ranges would be entirely insufficient 

 to handle the seventy odd db range at speech intensities, it is necessary 

 to control volumes to a given point before sending through these 

 devices, rather than compress or expand first and then control. The 

 range is adequate, however, to take care of the range of signal intensi- 

 ties for commercial speech at constant volume. 



Effect of Compandor 



The compressor curve of Fig. 6 indicates that, when the input is 

 15 db above 1 milliwatt, the compressor gives no gain or loss. If 

 the levels are adjusted so that this point corresponds to the intensity 

 at point B on Fig. 2, then the line 5C indicates the controlled intensities 

 corresponding to the assumed 30 db spread of speech controlled to 

 constant volume. The new range of intensities as indicated by the 

 bracket X" Y" Z" is now finally reduced to about 15 db. Tests show 

 that a volume indicator on the output of the compressor reads from 1 to 

 2 db higher than on uncompressed speech at its input. Compressed 

 speech sounds slightly unnatural but the effects of compression upon 

 articulation in the absence of noise are negligible. 



In considering the action of the expandor it is important to note 

 that all of the improvement in signal-to-noise ratio is put in by the 

 compressor. Considering any narrow interval of speech the insertion 

 of the expandor does not change the signal-to-noise ratio. The de- 



