518 BELL SYSTEM TECHNICAL JOURNAL 



there is probably little use in complicating the modulating arrangement to 

 produce sinusoidal shift, merely for the purpose of simplifying the sending 

 filter. 



In the multi-channel (on-off type) voice frequency carrier telegraph used 

 in the Bell System plant, the carrier currents of the different channels have 

 frequencies which are odd multiples of an 85-cycle base frequency, and the 

 channel filters have corresponding midband frequencies. Even order modu- 

 lation of these carrier currents, which occurs to a certain extent in the line 

 repeaters of the system, results in the production of interfering frequencies 

 which are even multiples of 85 cycles. These products fall midway between 

 the pass bands of the receiving channel filters and the loss which they en- 

 counter in these filters greatly reduces their efifect. In a telegraph system 

 having this channel frequency arrangement, but designed to operate on a 

 frequency-shift basis, with the carrier frequencies shifted over a large portion 

 of the channel frequency bands, even order modulation products originating 

 in the line repeaters would, to a much greater extent, lie in frequency ranges 

 freely passed by the receiving filters; and the effect of such interference would 

 be correspondingly greater than in the on-off system. 



Noise Tests 



One way to judge the relative noise sensitivity of carrier telegraph arrange- 

 ments is to subject each to measured amounts of noise on the line and then 

 to compare the resulting signal distortions. Resistance or thermal noise 

 was used in these tests because it is the most general kind of noise. It con- 

 sists of a superposition of rapidly recurring random impulses, some of which 

 may overlap. No tests were made using impulse noise such as caused by 

 lightning, ignition, or sharp static, because it was thought that resistance 

 noise tests would suffice. Impulse noise, when considered in a strict mathe- 

 matical sense, consists of isolated pulses of very short duration and the com- 

 ponent frequencies are so phased with respect to each other that their 

 amplitudes add arithmetically at the instants of occurrence of the pulses. 

 Atmospheric disturbances range all the way from isolated pulses to grinding 

 static caused by dust storms which has characteristics approaching those of 

 resistance noise. It is difficult to choose a representative type of impulse 

 noise for testing. Another reason for not testing with impulse noise was that 

 theoretical considerations^^ indicate there is not much difference in the 

 advantage of frequency-shift over on-off methods whether the disturbance 

 is of the impulse or resistance type. 



In order to compare the sensitivities of the different arrangements to re- 



'^M G. Crosby: "Frequency Modulation Noise Characteristics", Proc. I. R. E., 

 Vol. XXV, No. 4, April 1937, pp. 472-514. 



