A restructured decision feedback equalizer for facilitating the LMS algorithm

The least-mean-square (LMS) algorithm used to adapt the feedforward and feedback filters of a decision feedback equalizer (DFE) is often poorly conditioned. The result is that convergence is slow and misadjustment large. We propose a restructuring of the DFE that will improve the performance of the adaptive algorithm without changing the efficacy of the equalizer. In particular, we propose two similar designs neither of which requires a matrix inversion or matrix multiplication. One takes advantage of some a priori knowledge of the channel and the other draws on simple channel identification as part of the adaptive process to facilitate the algorithm