Equalization and estimation for fading channels
Thesis DisciplineElectrical Engineering
Degree GrantorUniversity of Canterbury
Degree NameDoctor of Philosophy
The main contribution of this thesis is the development of high performing, reduced complexity receivers for fading channels. Three different receiver structures are proposed and they operate without the need of any channel statistics. First, a double filtering receiver for systems employing DPSK modulation on channels with small delay and Doppler spreads is presented. BER performance results obtained via simulations and analysis show that the proposed structure outperforms the conventional matched filtering DPSK detector. Second, a polynomial predictor based sequence detector for flat-fading channels is presented. The receiver consists of a bank of polynomial least squares FIR predictors. The proposed receiver is not restricted only to systems using constant envelope modulation schemes. Analytical and simulated BER results are presented. In some cases, the proposed receiver performs only a few dB worse than an MLSE receiver with known channel statistics. Third, a sequence detector employing the polynomial based GRLS channel estimator is presented. The GRLS estimator is a generalization of the standard RLS algorithm and it requires a state space model of the channel to operate. It is shown that by using a polynomial or t-power series of the channel coefficients, a justifiable state space model may be derived without the need for any channel statistics. An analytical technique to evaluate the tracking performance of the GRLS estimator is also presented. The new analytical method may also be applied to the standard RLS algorithm with improved results. Simulated and analytical results show that in some cases, the tracking performances of the proposed channel estimator is almost as good as that of an optimal Kalman based channel estimator. BER results also indicate that the sequence detector using the proposed GRLS estimator performs just as well as one using a Kalman based estimator.