688 



THE BELL SYSTEM TECHNICAL JOURNAL, MAY 1957 



I- 

 2 



UJ 

 UJ 



> 



O 



a. 

 a. 



12 



11 



10 



10 



K^ 



LU — 

 7LU 6 



cct 



LU _) 4 



cr 



D 

 t- 



LU 



Q 



II 



SIGNAL POWER IN DECIBELS BELOW FULL LOAD SINUSOID 

 15 20 25 30 35 40 45 50 



4 



8 10 



20 



40 



60 

 C 



100 



200 



400 600 1000 



Fig. 14 — Improvement selectivity, i.e., departures from the ensemble upper 

 limit of improvement due to the use of a single value of ^ rather than the ensemble 

 of Me values. The minima at 5^ = locate the signals (C) for which /i andjuc coincide. 

 All curves correspond to the case where the "dc component", eo , is zero. 



Lest it appear that this range is so broad as to offer very httle practical 

 guidance, it should be noted that (38) defines a rather narrow range of 

 characteristics in Figs. 3 and 4. The assumption that this range may be 

 realized in practice appears reasonable in \'iew of the similarity to the 

 characteristic actually used by Meacham and Peterson, which is shown 

 in Fig. 4. 



2. Practical Limitations on Companding Improvement 



(a) Mismatch Between Zero Levels of Signal and Compandor. Although 

 the present discussion has hitherto been confined to ideal compandor 

 action, it lends itself quite naturally to the analysis of a significant 

 departure from ideal behavior which may be expected to result from 

 the use of an instantaneous compandor on a common channel basis in 

 time division multiplex systems. 



It will probably l)e impractical to balance the channel gating circuits 

 (required to provide sequential connection of individual channels to a 



