CARRIER TELEGRAPH METHODS 505 



cut-off low-pass filter because of the steeper wave front of the signals deliv- 

 ered by the former low-pass filter, resulting in a lower signal bias when the 

 detected amplitudes of the marking and spacing signals differed. 



When the average carrier frequency of a two-source arrangement was 

 varied, the marking and spacing frequencies moved toward one side of their 

 respective pass bands. As they approached the cut-off frequencies of the 

 band-pass filters, telegraph distortion occurred due to suppression of the 

 carrier and adjacent components. The narrow-band two-source arrange- 

 ment was more sensitive to variation in the average carrier frequency than 

 the normal-band two-source arrangement, and the reason is obvious. 



Bias at 60 Words per Minute 



When tested with reversals at 23 d.p.s. (corresponding to 60 w.p.m.), the 

 arrangements also ranked in the same order from the bias standpoint as from 

 the distortion standpoint, as the mean carrier frequency was varied. In 

 Figs. 8 and 9 the distortion due to carrier frequency variation consisted 

 mainly of bias, except in the two-source arrangements, where the received 

 marking and spacing pulses were substantially equal on reversals so that 

 there was little bias. (Bias measurements referred to here and below have 

 not been shown graphically in order to save space.) 



The on-off, two-source, and one-source two-band arrangements were 

 fairly insensitive to carrier frequency variations since the loss vs. frequency 

 characteristics of the channel filters changed but slowly near the middle of 

 the transmission band. Since the frequency-shift arrangement had a dis- 

 criminator which was sensitive to frequency changes, drifting of the average 

 carrier frequency resulted in a rise in detected current in one half of the push- 

 pull detector and a reduction thereof in the other half, thus causing serious 

 bias in the differentially combined rectified waves, since no frequenc>- com- 

 pensator was provided. It is outside the scope of this paper to describe such 

 a compensator, but it is no more complicated than the level compensator 

 used with an on-off arrangement. In the single-sideband arrangement the 

 carrier was located at a point on the filter loss vs. frequency characteristic 

 where the slope was steep. Consequently small frequency changes produced 

 large amplitude variations in the operating current of the receiving relay. 

 Since there was no compensating change in the biasing current of the relay, 

 large bias variations resulted from small changes in carrier frequency. 



Distortidii at 110 Words per Minute 



Figures 10 and 11 give distortion for the arrangements at 46 d.p.s. or 120 

 w.p.m. when the mean carrier frequency was varied. The arrangements 

 ranked in the following order when the distortion was 20 per cent : two-source 

 without limiter, two-source with limiter, linear on-off, one-source two-band, 



