514 



BELL SYSTEM TECHNICAL JOURNAL 



and the sinusoidal frequency modulation is nearly the correct representation, 

 the above procedure is not subject to much suspicion. The validity of the 

 right-hand portion of the curve is upheld by the fact that it is about 12 db 

 lower than the marginal fluctuation noise curve of Fig. 11. If the wide 

 swing FM interfering wave had a sjjectrum much like fluctuation noise of 

 the same i)ower as the FM wave the difference would be 9 db. 



10 



20 



400 600 1000 

 N MEGACYCLES 



2000 4000 

 PER SECOND 



10,000 



40 60 80 100 200 



RADIO SIGNAL BANDWIDTH 



Fig. 12 — PPM-FM; performance with respect to CW and similar system interference 

 for 1000 4 kc channels with ratio of FM wave to interference marginal. Relations between 

 bandwidth and audio signal-to-interference ratio v.'hen baseband is optimum for suppres- 

 sion of fluctuation noise. 



It has been explicitly assumed that both systems are idle, but we see no 

 reason to believe that if either or both were normally active the interference 

 would be significantly different for our purposes. 



Audible interference from a CW wave is caused by a disturbance to the 

 frequency of the FM wave. Let us first assume that the CW frequency lies 

 near the middle of the frequency swing range. No disturbance to the FM 

 wave occurs as its frequency passes through coiitcidence with that of the CW 

 but, as the frequencies diverge, the magnitude of the disturbance as well as 



