PRIVATE LINE DATA TRANSMISSION 1471 



carrier terminals, in the part of the circuit where the signal is at voice 

 frecjuency. In such a case, the distortion increases with the number of 

 times in the telephone facility that the signal is modulated down to 

 voice frequency. Second order nonlinear distortion tends to develop 

 modulation products in the lower portion of the transmission band 

 which are a source of potential interference wdth the signal. 



Another impairment encountered in carrier telephone channels is a 

 slight frequency shift; that is, a 1,000 cycle input may appear at the 

 output, say at 998 cycles. This occurs because modulator and demodula- 

 tor frequencies are not identical. With independent oscillators on recent 

 systems this shift may amount to some two cycles. With older systems 

 it can run from 5 to 10 times as much. The effects of this shift are dis- 

 cussed below. The frequency shift may be avoided by working double 

 or vestigial sideband and using an envelope detector, or in some carrier 

 systems by locking the oscillators in a constant frequency network. This 

 locking may or may not result in close phase synchronization of the 

 carriers, depending on the method used to lock and the particular carrier 

 system involved. 



Still another factor is the use, or not, of "compandors." A compandor 

 compresses the range of speech volume in the impressed line signal and 

 correspondingly expands this range at the receiver. This raises the line 

 signal level during periods of low speech power, and lowers it during 

 periods of high speech power without, in principle, affecting the final 

 received level. The effect is to reduce the final noise in periods of low 

 speech power, and increase it during periods of high speech power. A 

 listener is less perceptive to the noise during high speech power levels 

 than low. By this means, it has been found that the telephone circuit 

 can be engineered to some 23 db more noise (and also crosstalk and simi- 

 lar forms of interference) than it can without the use of a compandor. 



In the case of data signals, however, the influence of noise in causing 

 error is not very much different whether the signal is marking or spac- 

 ing. Thus, there is no "compandor advantage"* (indeed there is a cer- 

 tain disadvantage as pointed out earlier), and facilities that have been 

 engineered to be entirely satisfactory for voice transmission are effec- 

 tively some 23 db more noisy for data transmission. As a practical mat- 

 ter it appears desirable to remove compandors from circuits used ex- 

 clusively for data. 



A short listing is presented here of the various types of message facili- 



* Perhaps a simpler way to think of it is that all possible "compandor advan- 

 tage" has already been obtained in a data system by using the best combination of 

 amplitudes for mark, space, start, etc. 



