Response of an FM Discriminator to a Digital FM Signal in Randomly Fading Channels

The calculation of the error rate resulting from the detection of a binary FSK signal using an FM discriminator-lowpass filter detector is described for an FM signal undergoing nonselective fading in additive Gaussian noise. The results extend the error rate calculations of Schilling et al. [1], [2] which were performed for a nonfading channel. While the technique employed is quite general, examples considering Rayleigh fading and an alternating mark-space sequence are given. Following a CCIR recommendation for Rayleigh fading, the percentage of time that the signal envelope is below a minimum level is ignored, since in this case the circuit would be completely cut off or switched to a better path. Thus, the results also depend on the depth of fade or ratio of the minimum signal amplitude squared to the average carrier power. The use of diversity is studied, and the error rate for a discriminator using predetection combining diversity is found. In this case the possibility of infinitely deep fades is considered. The analysis is based on the work of Nelson[9], [14], who considered signal-to-noise ratio performance of an FM discriminator with an analog FM signal undergoing nonselective fading. Experimental results are shown to verify the theory.