MULTICHA NNEL MODULA TION SYSTEM 23 



to build up above a threshold when the system is first framed. With this 

 time interval thus fixed, each clock position is maintained for about one 

 millisecond. Hence the time required to frame the receiver varies between 

 one and eighty-three milliseconds, depending upon the epoch at which the 

 system connection is established. 



FCM Receiver. The received PCM signals of a 12-channel group are 

 filtered from the frequency multiplex while in the intermediate-frequency 

 state and then detected to the original signal band, as indicated in Fig. 6. 

 They consist of rounded pulses, nominally sinusoidal in shape, but more or 

 less distorted by transmission defects and accompanied by noise and inter- 

 ference. 



These signals are supplied as input to the PCM receiver shown in Fig. 10. 

 They are first sliced in amplitude, the slice being taken at approximately 

 half the average or noise-free pulse height. Code groups of seven pulses 

 are then routed alternately to two decoders, which handle even and odd 

 channels respectively. The routing function is represented in the drawing 

 by a two-segment commutator (A) rotating at 48,000 revolutions per second. 

 Before entering the actual decoding circuit, the pulses are again sliced to 

 secure very great uniformity, and are gated with 0.4-microsecond, 672- 

 kilocycle pulses from the receiver timing equipment. Immediately after 

 the arrival of the seventh digit of each code group of these standardized 

 pulses, the decoding circuit produces a voltage on its low-impedance output 

 lead proportional to the quantized amplitude represented by the code. As 

 in the case of coding, details of the decoding process will be reserved for a 

 later section of this paper. 



The decoded amplitudes are available only momentarily; therefore it is 

 desirable to sample each one at the proper time and store the result as a 

 charge on a holding capacitor. This sampling process is represented by 

 switch B in one decoder and B' in the other. Here the switch closures last 

 only two microseconds, and values are held for about 19 microseconds. 



The next step is to assemble the six samples from odd channels held 

 successively by one capacitor and the six from even channels held by the 

 other into a single time-division multiplex. Switch C performs this opera- 

 tion, rotating at 48,000 revolutions per second, and making contact alter- 

 nately with the output circuits of the odd and even decoders. 



This 12-channel multiplex signal is passed through an instantaneous 

 expandor, the purpose of which has been noted in an earlier section. To 

 simpHfy the problem of making the input-output characteristic of this cir- 

 cuit accurately complementary to that of the compressor, the two devices are 

 designed to use identical silicon units. In the expandor, however, the non- 

 linear device is employed in the feedback path of a broadband amplifier 

 rather than in the direct transmission path, thus giving the inverse character- 



