524 BELL SYSTEM TECHNICAL JOURNAL 



the noise was small. But the reduction in distortion was negligible, because 

 the distortion was small in either case. 



In Fig. 21, a comparison of curve F, applying to the normal-band fre- 

 quency-shift arrangement with limiter, and curve K, applying to the wide- 

 band frequency-shift arrangement with 503-cycle low-pass filter, shows 

 that at 20 per cent distortion the increase in band width and frequency 

 swing resulted in a 9 db improvement in noise-to-carrier ratio. This im- 

 provement may be explained on the same basis as the 1.3 db improvement 

 obtained at 60 w.p.m. and was larger due to the higher distortion of the 

 normal-band arrangement when operating at 120 w.p.m. When the cut-oflE 

 frequency of the low-pass filter of the wide-band frequency-shift arrange- 

 ment was changed from 503 cycles to 58 cycles, this improvement in noise- 

 to-signal ratio was increased to 11.6 db. It is apparent that tolerance to 

 noise was appreciably increased by this change of filters but, of course, this 

 was accompanied by a reduction in maximum operating speed. 



N oise-to-Signal Ratio at Receiving Relay 



Another method that was used to measure noise-to-signal ratio on the 

 on-off and single-sideband arrangements was as follows: With the carrier 

 turned off, the receiving gain was increased until the receiving relay just 

 operated on occasional noise peaks. Then with the carrier on the line, and 

 with the noise absent, the receiving gain was readjusted until the receiving 

 relay again just operated. The difference in receiving gain in these two tests 

 was called the noise-to-signal ratio at the receiving relay. When this 

 method of measuring the noise-to-signal ratio was used it was found that the 

 on-ofif and single-sideband arrangements had distortion vs. noise-to-carrier 

 ratio characteristics similar to those given in Figs. 18 and 21. However, 

 in order to express these characteristics correctly when substituting the 

 words "at receiving relay" for "on line" in the scale of abscissae, it was 

 found experimentally that it was necessar>^ to shift the curves for the on-off 

 arrangement 7 db to the left and to shift those for the single-sideband 

 arrangement 4 db to the left, relative to their present positions in these 

 figures. Such a change of scale was necessary because the noise-to-signal 

 ratio was smaller at the output of the receiving filter than it was on the line. 



Summary Of Results 



Explanation of Tables VII and VIII 



Tables VH and Mil have been prepared from the test data in order to 

 compare the various arrangements for a given amount of distortion on each. 

 Table VH describes certain of their properties for 10 per cent maximum 

 total distortion. In column A (see bottom line of the table) are listed the 

 different arrangements. In columns B and C are given the speeds in dots 



