BAND II //;/•// AND TRANSM ISSIOS PJIKIORM A \CE 515 



tlie frequency of tlic disturbance increases linearly, 'i'he baseband lilter 

 is excited only during the time the difference is less than I''t>. Thus, the 

 disturbance results from a series of perturbations to the otherwise smooth 

 frequency variation of the I"'M wave. The time during which tliese pertur- 

 ])ations can atTect the baseband filter is short compared with the shortest 

 |)ulse the baseband hlter can pass, e.xcept when the baseband width is greater 

 than half of the swing. This occurs at the extreme left-hand end of the 

 curve. We have not attempted to calculate the response to these transients 

 except to note that the response is a j)ulse which extends roughly 2T from 

 its point of origin, peaking somewhere near the center of this interval. If 

 we assume that the PPM pulses are closely spaced (e = 0) so that they result 

 in a wave frecjuency modulated by 8 mc, there are two such evenly spaced 

 disturbance pulses per cycle of modulation (two per 2T interval) and there- 

 fore there is an almost continuous disturbance wave in the base band filter 

 output whose amplitude does not greatly exceed its RMS value. We have 

 accordingly calculated the power sum of all the extraneous frequencies 

 j)assed by the baseband fdter, assuming the FM wave to be sinusoidally 

 modulated. The location of the CW frequency giving greatest interference 

 power was used in these calculations excej)t in the wide band cases where 

 the worst frequency appeared to be near the edge of the band. Here the 

 transient viewpoint indicated that the resulting interference in the ba.se- 

 band would be greater if the CW frequency were nearer the center. 



If the trailing edge is used to measure the time of the pulse, the princij)al 

 disturbance of this time arises from the perturbation produced at the leading 

 edge of the same pulse, and so the calculation for close-spaced i)ulses is not 

 greatly in error when applied to wider-spaced pulses. If the leading edge 

 were used the worst CW frequency for widely spaced pulses would be one 

 differing from the rest frequency by Fb and the interference would be worse, 

 we think, than that arising from the frequency worst for trailing edge opera- 

 tion. 



It has been explicitly assumed that the system is idle, but we see no reason 

 to believe that the interference would be significantly different with normal 

 activity. 



Fig. 13— PAM-FM, Fluctu.atiox Noise 



Fluctuation noise in a PAM-FM system produces the sloped noise spec- 

 trum characteristic of FM in the output of the frequency detector. The 

 noise power per cycle is zero at zero frequency and increases with the square 

 of the frequency. The baseband filter accepts only the portion of the 

 spectrum between zero and Fi,. If instantaneous sampling of the signal 

 values is used, all noise frequencies in this range are equally effective as 

 causes of errors. Use of a channel gate of maximum permissible duration 



