MULTICHANNEL MODULATION SYSTEM 13 



in time-division multiplex, however, amplitude non-linearity of itself is not a 

 factor. In this case the limitation comes on the departure from a suitable 

 attenuation characteristic and from constant delay with respect to frequency. 

 Distortion from these sources may increase the pulse duration so that excessive 

 overlap of adjacent pulses will leave less margin available for interference and 

 noise. 



Pulse Code Demodulation 



At the receiver, the regenerated code pulses are operated upon to recreate 

 as closely as possible the original signal by operations complementary to those 

 at the transmitter. Unfortunately, however, no process at the receiver can 

 undo the effects of quantization which remain as noise, so that the quanta must 

 obviously be made small enough from the beginning to permit satisfactory 

 speech quality. 



In each time-division group alternately working decoders may be used, in 

 the same way as the two coders at the transmitter. Routing is effected by 

 suitably timed gates. Conversion of a pulse code to amplitude may be ac- 

 complished by causing each pulse of a code combination to contribute an ampli- 

 tude corresponding to the binary digit it represents, and then summing the 

 contributions. When tapered steps are employed, due consideration must be 

 given to overall linearity, discussed subsequently. The resulting output is a 

 pulse-amplitude-modulated signal, which is then distributed to the channels of 

 the group by an electronic commutator. Reconstruction of the signal from 

 these distributed pulses is accomplished by simple filtering, which serves to 

 remove components extraneous to speech introduced by the sampling procedure 

 and tied up with the sampling rate. 



Production of the necessary timing pulses at the receiving end proceeds in 

 much the same manner as at the transmitter except for one thing. That is, 

 instead of being initiated by a local oscillator, the receiver timing must be linked 

 to the input pulses so that they may be properly routed. This involves the 

 problem of synchronizing or, to borrow a term from television, framing. Use 

 of this term is based upon the similarity of the sequence of PCM digits within 

 a single sampling period to the complete ordered array of television picture 

 elements. Preferably framing should be done with a minimum of time interval 

 and of band width. 



One method of synchronizing pulse systems employs a marker pulse which 

 serves to initiate a timing sequence for each frame at the receiver. Here the 

 marker pulse must differ sufficiently from the other pulses to permit its rapid 

 and certain identification. This is ordinarily done by making the marker 

 several times as long as any message pulse. In PCM, however, where digit 

 pulses are run together in many code combinations, the marker would have 

 to be very long to be clearly distinguishable, thereby cutting into channel ca- 



