FREQUENCY SHIFT TELEGRAPHY 



283 



band with a discriminator consisting of two bandpass filters and 170-cycle 

 shift was compared with a 230-cycle band with a linear discriminator and a 

 shift of 140 cycles. The results are shown in Fig. 20. The linear dis- 

 criminator in this case appears to fail slightly sooner but shows a superiority 

 of about 2 db in the 5% distortion region. Due to the rounded character- 



so 



z 



LU 



oc 



UJ 

 Q. 



130 



a:20 

 O 



16 -14 -12 -10 -8 -6 -4 -2 



RMS NOISE-TO-CARRIER RATIO IN DECIBELS 



Fig. 19. — Linear discriminator versus two-bandpass-filter discriminator having the 

 same signaling speed capability. Effect of thermal noise on peak distortion — 80-cycle 

 cutoff low-pass filter. 



50 



<-'40 



z 



2 



o 



30 



-20 -18 -16 



-14 -12 -10 -8 -6 -4 -2 



RMS NOISE-TO-CARRIER RATIO IN DECIBELS 



Fig. 20. — Linear discriminator versus two-bandpass-filter discriminator with equal 

 bandwidths. Effect of thermal noise on peak distortion — 80-cycle low-pass filter. 



istic of the two bandpass filters used as a discriminator in the second com- 

 parison (Fig. 18) the difference in discriminators is less than in the previous 

 test. It has been found^ that for a given signaling speed capability almost 

 twice the bandw^idth is required if a two-bandpass-filter discriminator is used 

 instead of the linear type. This added band width does not appear to cause 

 any loss in signal-to-noise capabilities as to the failure point. The less sharp 

 breaking point, however, makes the linear discriminator superior for moder- 



