Skip to Main Content
In current long-distance h.f. radio practice frequency-shift-telegraphy signals are usually demodulated by the use of limiters and discriminators on conventional frequency-modulation lines. This method of demodulation is liable to fail whenever the signal fades on either the mark frequency or the space frequency. This is wasteful. The full message is available on each frequency and failure need not occur unless signals on both fade together; if the fading is frequency-selective, as it often is, simultaneous fading may be comparatively rare. Thus there is the possibility of frequency diversity. Theoretical analysis of the case of Rayleigh-fading signals disturbed by white Gaussian noise, the fading being slow relative to the speed of signalling, shows that substantial advantages may be derived from frequency diversity. Also, the analysis leads to a mathematical specification for an ideal receiver. An experimental equipment has been made, which, while avoiding the complexity that full accord with the specification would involve, complies with it in several respects. The most important feature is that the received signals are assessed in terms of the expected amplitudes of mark and space signals, derived from observation of earlier elements; that is to say, the absence of a mark signal, which would be strong if it were present, is taken as a strong indication of space. The effective bandwidth is reduced almost to the limit by the use of linear-build-up band-pass filters to respond to the mark and space frequencies. Laboratory tests on the experimental unit gave results in good agreement with theory. At the optimum signalling speed the noisy-signal performance was only about 3 dB short of the ideal. On signals from Australia the experimental unit gave a useful improvement compared with a high-grade receiver of the conventional type, although not so large as in the laboratory tests. The difference is attributed to the occurrence on the practical radio channel of disturba- nces, such as interference and atmospheric crashes, other than white Gaussian noise. Further investigations are needed. The new technique may have other applications.