FREQUENCY SHIFT TELEGRAPHY 



287 



In Fig. 25 is shown the effect of frequency drift on signal bias in an AM 

 system. In an AM system little bias is produced until the carrier reaches 

 the cutoflf region of the filter. The bias then becomes rapidly negative due 

 to the increased loss and decreased amplitude of demodulated signal. When 

 an automatic gain control arrangement is used the bias becomes positive 

 due to a distorted envelope shape. The demodulated wave form deter- 

 mines the degree of sensitivity to frequency drift and depends on the band- 

 width both before and after demodulation. 



In an FS System using a linear discriminator, frequency drift changes 

 the d-c. component of the signals and thus changes the operating point on 

 the demodulated wave. The amount of bias depends upon the slope of the 

 wave front and is thus affected by the amount of low-pass filtering. The 

 effect of frequency drift on bias for a number of FS systems is shown in Figs. 

 26 and 27. If a two-bandpass filter type of discriminator is used the system 



cc 



z 



<0-20 

 ^-30 



-500 -400 -300 -200 -100 100 200 300 400 

 FREQUENCY DRIFT IN CYCLES PER SECOND 



Fig. 25. — Signal bias versus frequency drift for AM transmission in a 740-cycle band. 



is insensitive to moderate frequency drifts due to the flat pass bands as illus- 

 trated in Fig. 26. The relative shape and amplitude of the signal from a 

 linear discriminator does not change appreciably with frequency drift; 

 only a d-c. displacement occurs. This makes it desirable to have the low- 

 pass filter coupled to the output amplifier by a network which passes only 

 the useful signaling frequencies and blocks the d-c. and very slow drift com- 

 ponents. When this is done the effect of frequency drift on bias is not 

 greatly different from that for an AM system, as may be seen by comparing 

 the dotted curves on Figs. 26 and 27 with Fig. 25. 



The general method of performing this d-c. elimination is illustrated in the 

 block diagram of Fig. 29. The output from the low-pass filter is passed 

 through a coupling network which blocks the d-c. and passes the useful 

 signaling frequencies. The output of the coupling network is passed 

 through a positive feedback nonlinear amplifier which has but two output 

 conditions representing the mark and space of the telegraph signal. The 

 feedback network passes d-c. and low frequencies so as to just compensate 

 for the loss of the coupling network. The time constant of the coupling 

 network may be made large enough so that the signal wave form into the 



