40 BELL SYSTEM TECHNICAL JOURNAL 



coder only (without passing through the compressor), and the output was 

 taken directly from the common output of the two decoders. Thus the six 

 odd channels took turns transmitting the signal while the six even channels 

 produced the horizontal center-line. Uniform quantizing steps may be seen 

 along most of the trace, but are obscured near the ends by defocusing of the 

 test oscilloscope. 



A similar pattern, obtained with the compressor and expandor included 

 in the transmission path, is shown in Fig. 25. Here tapered steps may be 

 discerned, as well as a small amount of non-linearity due to residual imper- 



Fig. 25. Output of expandor (vertical) vs. input to compressor (horizontal). 



fections in the companding. The effect of the channel peak choppers is not 

 included in this pattern. 



Two measured overall input-output characteristics appear in Fig. 26, for 

 the case of a typical single channel and for five channels patched in tandem 

 through 17-decibel pads on a 4-wire basis. The latter simulates a possible 

 extreme case of a long circuit in which for some reason it is desired to decode 

 to audio at four junction points as well as at the final terminal. It should 

 not be confused with a series of spans between which the PCM pulses are 

 amplified or regenerated without decoding. In the latter case, of course, the 

 overall audio characteristics are independent of the number of spans. 



Audio Noise. Quantizing was found to be the only significant source of 

 noise in the received audio signals. Noise levels measured in the absence of 

 speech are shown in Fig. 27. The measurements are given for various num- 



