510 BELL SYSTEM TECHNICAL JOURNAL 



tween the marking and spacing current amplitudes on the Une, causing a 

 corresponding reduction in sideband power. If the quadrature component 

 had been completely ehminated it is possible that 1998 cycles would have 

 been found to be a more favorable carrier location than some of the other 

 closely adjacent frequencies. However, no attempt was made to design the 

 filters for the theoretical single-sideband requirement that the transfer 

 admittance of the vestigial sideband should be complementary^ to that of 

 the other sideband near the carrier frequency. 



Tests of Received Level Variations 



Any transmission path is likely to have level variations caused by tempera- 

 ture and weather changes in case of wire lines and by fading in case of radio 

 links. Each arrangement was therefore tested for susceptibility to level 

 changes by var>'ing the artificial line over a wide range. As the artificial 

 line was made of resistances, the effect of equal fading over the entire fre- 

 quency range was thereby simulated. 



Distortion at 60 Words per Minute 



In Fig. 13 is shown the total distortion at 60 w.p.m. vs. level change for 

 different arrangements using the same type of filter. As may be seen from 

 Fig. 13, the arrangements which had the same loss characteristic ranked in 

 the foUowmg order at 20 per cent distortion and 60 w.p.m. as regards stability 

 when the line level was varied: two-source with luniter on a par with one- 

 source two-band, frequency-shift, level compensated on-ofif, two-source 

 without limiter, Imear on-off , and single sideband with — 6 db spacing carrier. 

 The range of levels over which the first three arrangements mentioned above 

 were stable, in the absence of interference, was largely a function of the range 

 of the limiter, which was over 80 db. The on-ofif arrangement including a 

 level compensator had a range of approximately 40 db, but it should be re- 

 membered that a level compensator is effective only for level changes slow 

 compared to the signal speed. The range of the two-source arrangement 

 without limiter depended on the accuracy with which the marking and 

 spacing halves of the receiving circuit were balanced. In the arrangement 

 tested this range was about 30 db. The linear on-ofif arrangement without 

 level compensator was ver}' sensitive to level changes because the operating 

 current in the receiving relay varied without a compensating variation in the 

 bias current. The single-sideband arrangement had greater sensitivity than 

 the linear on-ofif arrangement because a variation of 6 db in the line current 

 of the single-sideband arrangement was accompanied by a 15 db variation in 

 relay operating current due to the use of increased gain and a large grid bias 

 in the receiving d-c amplifier. 



