Digital data transmission over an HF channel

The thesis is concerned with detection, estimation techniques and a method of the adaptive adjustment of the equaliser, for use in a 4800bit/sec synchronous digital transmission system operating over a voice-band time-varying HF channel. Two main impairments are additive Gaussian noise and inter-symbol interference (ISI), which can be very severe at times. All techniques considered here are algorithms or processes that operate on sequences of sample values. Modern digital modems normally operate in this way, and the techniques described are of direct application to practical systems, and could be implemented using the new technology of high speed real-time digital signal processing (DSP). The performance of the various systems that employ the above techniques are obtained using the computer simulated model of three types of HF channels. The ionospheric propagation medium, the characteristics of HF channel and the signal distortion introduced by the channel are first described. The thesis then presents a suitable base-band model of the HF channel for computer simulation of quadrature amplitude modulation systems. A suitable method for the adjustment of the receiver is described next. This method is suitable both for the adjustment of a conventional decision feedback equaliser (DFE), and also for the adjustment of a linear feedforward filter that is employed ahead of a near-maximum likelihood (NML) detector. This method uses a minimum phase (root-finding) algorithm (MPA) to convert the channel response from being non-minimum phase to at least approximately minimum phase. The results of computer simulation tests of this algorithm are then presented over different types of HF channel models. The results demonstrate the algorithm's capability to make the channel response minimum (or near-minimum) phase. Various NML detectors, derived from the Viterbi detector, are discussed. Each detector is here preceded by an adaptive linear filter that is adjusted adaptively using an MPA. The performance of these detectors is compared with the conventional DFE, whose tap-gains are adjusted adaptively using an MPA, and the detector which gives the best compromise between performance and complexity is selected for combined receivers. These results are obtained using perfect estimation of the channel response. The estimation techniques studied in this thesis include both new and conventional estiniators, which are based on the least- mean-square (LMS) algorithm or recursive least-square(RLS) algorithm. The estimator provides an estimate of the sampled impulse response (SIR) of the channel, necessary for the NML detector or MPA. The performances of these estimators are compared using computer simulation tests. The results also demonstrate that the simpler LMS algorithm with adaptive step size gives a comparable level of accuracy with the more complex RLS algorithm. Finally the most promising of the detectors and estimators are connected with an adaptive equaliser, using an MPA, to form a new combined receiver. The details of the combined system structure with its computational complexity are given. Extensive computer simulation tests have been carried out on the different arrangements of the combined system including DFE, when all the functions of detection, estimation and MPA are present, in order to find the most cost effective system in terms of performance and complexity. A considerable reduction in the equipment complexity can be achieved by allowing a long period between successive adjustment of the adaptive filter and estimator.