Mitigating Error Propagation of Décision Feedback Equalization in Broadband Communications

Inter-symbol interférence is the main obstacle in reliable digital communications. Except for the spécial partial response signaling system, ISI must be suppressed as much as possible in order to obtain reliable digital communications. In wideband digital com­ munication, equalization is one of the main techniques to reverse ISI channel. Décision feedback equalizer is a popular structure because of its relatively simple structure and good performance. This thesis is concerned with the error propagation mitigation of minimum-mean square error décision feedback equalizer (MMSE-DFE). DFE is an important part in modem broadband digital communications. In order to decrease the DFE error propagation, the main drawback of DFE, a précise mean-squareerror (MSE) model of DFE is proposed. Based on the new MSE model, DFE filter weights, erasure slicer and soft slicer can be optimized. Thèse methods are numerically simulated after theoretical analysis. In contrast to the tedious Markov chain model, the new methods are simple to implement. At the sarne time, the prédiction précision is very good in contrast to the conventional MSE criterion. Especially, the optimal soft slicer design shows that this optimization exceeds the joint DFE constrained filter weights and an erasure or a soft slicer design. The conventional DFE filter weights with the optimal soft slicer can be considered as a "true" MMSE DFE. The second contribution of this thesis is that a new kind of coding scheme is proposed to counteract DFE error propagation for high order modulation. The redundant bits in the new modulation scatter pattern can dramatically increase error propagation immunity at the same average symbol energy constraint. One of the promising features of this method is that it does not require additional decoding delay. The second feature is that it can still provide excellent error propagation robustness when the feedback weights of DFE are large compared to the conventional DFE. Overall, this thesis explores several means in the battle with DFE error propagation. DFE error propagation is traditionally difficult to analyze because of the nonlinear décision slicer feedback. The study results reveal that the introduction of DFE error