Evaluation of a Gaussian HF Channel Model

The modeling of an HF communication channel as a randomly time-varying linear filter has proved useful in the analysis of data transmission systems and in the study of ionospheric phenomena causing dispersive effects. Such a model may be evaluated by comparing properties determined experimentally with those based on the hypothesized model. By assuming that the channel output is a zero-mean Gaussian random process when a CW signal is transmitted, theoretical probability distributions have been derived for phase change and log-envelope change that reflect secondorder channel properties. These distributions depend upon the joint distribution of the original process at two instants in time and thus provide a more sensitive check on the hypothesized model than first-order properties. Experimentally measured phase-change, logenvelope-change, and envelope properties have been compared with those based upon the hypothesized model with close agreement. Estimates of correlation coefficient for specified lag times derived from the phase-change distributions agree well with estimates made from the power spectrum of the received signal (the second estimate is independent of the hypothesized channel model). The agreement of the experimental results with those based upon the zero-mean Gaussian model demonstrates the usefulness of that model.