530 BELL SYSTEM TECHNICAL JOURNAL 



fairly general conclusions are warranted. With the use of other types of 

 filters, relays, etc., somewhat different results would doubtless be obtained. 



Relative advantages and disadvantages of frequency-shift, on-off, and 

 single-sideband methods of carrier telegraphy were deduced from measure- 

 ments utilizing the same channel filters and covering a range of signalling 

 speeds, noise, interference, and other variables. 



A frequency-shift arrangement with amplitude limiter is substantially 

 unaffected by non-selective level changes and relatively insensitive to noise 

 currents. However, mterfering currents from similar flanking channels are 

 greater than when using the other methods. It has bias instability when 

 the mean carrier frequency drifts. None of the arrangements tested had 

 any compensation for drifting of the mean carrier frequency. However, 

 it is known that automatic compensation for this may be obtained by various 

 special methods not here described. 



A channel using the on-off method is less sensitive to frequency drift, 

 slightly faster for a given band width, and cheaper in terminal equipment 

 than that using the frequency-shift method. The interfering currents from 

 similar flanking channels using the on-off method are quite small. Weak- 

 nesses of the on-off method are greater sensitivity to noise and level changes. 

 However, on good wire lines, when a level compensator^^ is used, these 

 weaknesses are unimportant, and the on-off method is satisfactory. Of 

 course, these weaknesses become important on radio circuits when noise is 

 strong and when fades are too rapid or too severe to be overcome by the 

 level compensator. 



The greatest speed for a given band width is attainable by the single- 

 sideband method. Unfortunately this method is poor from the standpoint 

 of interchannel interference, is the most susceptible to noise, and, unless 

 special compensating devices are used, is the most sensitive to carrier fre- 

 quency drift and level changes. 



Several arrangements were also investigated which utilized approximately 

 double the band width of those mentioned in the preceding paragraph. 

 Among these, the two-source method is the best of all the methods herein 

 mentioned from the standpoint of insensitivity to changes in carrier fre- 

 quency, whether or not a limiter is used. If a limiter is used this arrange- 

 ment ranks well in its ability to withstand non-selective level changes. 

 Two-source arrangements are sensitive to differences between the. marking 

 and spacing levels, but some advantage is obtained by the use of a limiter. 

 The distortion vs. speed characteristic of the two-source method without 

 limiter is about the same as that of the linear on-off method utilizing half the 

 band width of the two-source method. As previously explained, the use of 

 a limiter with this method causes distortion and materially reduces the max- 



