A joint optimization of the decision feedback equalizer (DFE) filter weights and an erasure slicer is proposed in this letter. The feedback weights of the DFE are constrained with an accurate mean square error (MSE) model as the criterion. An erasure slicer is adopted to further decrease error propagation. Two consecutive 1-D optimizations are shown as good as that of a single 2-D optimization. Simulation results verify the performance improvements of the proposed scheme.

A technique is proposed to mitigate the DFE error propagation effect in high order modulation on highly time dispersive channels. Although the performance improvement is not large with the technique alone, it can offer some advantages in terms of the distribution of error burst length. It could be considered as a primitive channel coding. When combined with an outer RS code, the performance improves significantly. Preliminary results show that this method can achieve more than 1 dB SNR gain below 10-2BER to the conventional DFE of similar interleaving and channel coding scheme.

A technique on optimal decision feedback equalizer (DFE) soft feedback slicer design is proposed in this letter. A soft slicer of M-ary PAM is given as a function of the noise variance at the slicer input. The equalizer mean-square error (MSE), which includes the error propagation effect, is used as the criterion of the optimal soft slicer design. By incorporating the slicer characteristics into the MSE and the assumption of Gaussian noise distribution at the equalizer output, the optimal soft slicer is obtained by minimizing the MSE. Simulation results corroborate the effectiveness of the method.