A unifying study of robust decision feedback equalisers

Abstract

This paper presents a unifying technique for robust feedback filter design in the decision feedback equaliser (DFE). The robustness is needed for combatting error propagation effects which is the major limitation when using DFEs. Numerous constraint strategies for the design of robust DFEs are available in the literature, however, they have been presented in their own ad-hoc context. Consequently there is a desire to give a unified approach to robust feedback filter design. The approach in the current presentation is to first develop a mean square error (MSE) measure which is a function of the feedforward and feedback filters. This measure is then combined with a feedback filter constraint using four existing norm constraint approaches (the L1, L2, Linfin and the "two tap" norm). The use of second order cone programming allows these existing norm constraint strategies to be unified and solved in a computationally efficient manner. The evaluation is performed with respect to the impact on error propagation, through the mean and the variance of the burst error length