MULTICHANNEL MODULATION SYSTEM 7 



amplitude sounds it is necessary to transmit. Considering that consonant 

 levels may be of the order of 30 decibels below vowels, and that weak talkers 

 may be of the order of 30 decibels down from loud talkers, it is clear that ampli- 

 tudes as far below maximum as 60 decibels require at least a few steps. 



Ordinarily this would involve a large number of pulses for transmission, 

 with a consequent increased complexity of terminal apparatus, and an in- 



OUTPUT 



(a) 



OUTPUT 



Fig. 4. Relation between input and quantized output, with quantization uniform in 

 (a) and tapered in (b). 



creased frequency band per channel. A more practical solution is to employ 

 tapered steps rather than uniform ones. In this way a given number of steps 

 can be assigned in greater proportion to the low amplitudes than to the highs, 

 as shown in Fig. 4b. There results a degree of step subdivision sufficient to 

 care adequately for the low-ampHtude sounds including background noise. 

 A small penalty is paid at the upper end of the amplitude scale because of the 

 proportionately smaller number of steps available there, bringing in higher 



