Partial-update adaptive decision-feedback equalization

In this paper, we consider partial updating of decision-feedback equalizers (DFEs). Application of data-dependent partial-update techniques to DFE is complicated by the fact that the feedback filter regressor has entries with discrete amplitudes, making magnitude-based sorting impossible. We present a novel selective-partial-update algorithm for DFEs that treats feedforward and feedback filters separately and utilizes past decision-directed errors for feedback filter coefficient selection. In addition, a new partial-update scheme is proposed that switches between selective partial updates and periodic partial updates to make effective use of training periods in telecommunication applications. Simulation results are presented to corroborate the effectiveness of the new partial-update DFEs in dealing with realistic frequency-selective and time-varying channels at reduced complexity.