New Methods for FM Threshold Extension in GNU Radio

Abstract

This work describes new techniques for low carrier-to-noise ratio (CNR) FM demodulation and their implementation in GNU Radio. FM radio was developed nearly 90 years ago and is still in common use for commercial broadcasting, emergency services, and amateur radio. The key advantage of FM over other analog modulations (i.e. AM and SSB) is the threshold effect where received signals with a CNR above a threshold level will be demodulated to audio signals with a signal-to-noise ratio (SNR) larger than the CNR. However, demodulated FM signals below the threshold are subject to click noise, which rapidly degrades audio quality and intelligibility. With modern signal processing approaches, it is possible to build practical demodulators that reduce click noise, thus improving communication networks that employ FM.

The presentation begins with a review of the classical theory of noise in FM receivers based on the work of S.O. Rice (Rice, 1962). Next, it covers a new technique for FM demodulation that leverages the natural structure of FM signals in noise, including the treatment of large noise events as erasures. Once identified, these erasures can be removed using band-limited signal interpolation with the discrete Papoulis-Gershberg algorithm (Ferreira, 1994).

Last, an example implementation of the modern FM demodulator in GNU Radio is demonstrated and compared with commercial receivers for amateur radio communications.