396 



BELL SYSTEM TECHNICAL JOURNAL 



principle. The essence of the time-division principle is that any input wave 

 can be represented by a series of regularly occurring instantaneous samples, 

 provided that the sampling rate is at least twice the highest frequency in the 

 input wave.^ For present purposes the amplitude quantization principle 

 states that a complex wave can be approximated by a wave having a finite 

 number of amplitude levels, each differing by one quantum, the size of the 

 quantum jumps being determined by the degree of approximation desired. 

 Although other arrangements are possible, in this paper we will consider 

 the application of these two basic principles in the following order. First 

 the input wave is sampled on a time-division basis. Then each of the 

 samples so obtained is represented by a quantized amplitude or integer 

 number. Each of these integer numbers is represented as a binary number 

 of n digits, the binary number system being chosen because it can readily be 



ENVELOPE OF 

 AUDIO SIGNAL 



NO AUDIO SIGNAL 



Fig. 1 — Pulses in a PAM System. 



represented by ON-OFF or two-position pulses. 2" discrete levels can be 

 represented by a binary number of n digits.* Thus, PCM represents each 

 quantized amplitude of a time-division sampling process by a group of 

 ON-OFF pulses, where these pulses represent the quantized amplitude in a 

 binary number system. 



The discussion so far has been in general terms. The principles just 

 discussed will now be illustrated by examples. 



Multiplex transmission of speech channels by sending short pulses 

 selected sequentially from the respective speech channels, is now well known 

 in the telephone art and is called time-division multiplex. When the pulses 

 consist simply of short samples of the speech waves, their varying amplitudes 

 directly represent the speech waves and the system is called pulse amplitude 

 modulation or PAM. In PAM the instantaneous amplitude of the speech 

 wave is sampled at regular intervals. The amplitude so obtained is trans- 



' This is because the DC, fundamental and harmonics of the wave at the left in Fig. 1 

 all become modulated in the wave at the right, and if the highest modulating frequency 

 exceeds half the sampling rate, the lower sideband of the fundamental will fall in the 

 range of the modulating frequency and will not be excluded l)y the low-pass filter. The 

 result is distortion. 



■• In a decimal system the digits can have any one of 10 values, to 9 inclusive. In a 

 binary system, the digits can have only two values, either or 1. 



