A blind adaptation algorithm for decision feedback equalizers with fractionally-spaced feedforward filters

This paper introduces a blind adaptation algorithm for decision feedback equalizers (DFE) which employ an oversampled feedforward filter. The coefficient set of the feedback filter is obtained from a pure recursive filter which is designed to whiten the oversampled input signal in the sense that it minimizes correlations between the original signal and versions of it which are shifted by multiples of the symbol duration. This recursive filter is adapted by the introduced multichannel extension of the hyperstable adaptive filter algorithm (HARF). The forward equalizer is adapted by the standard constant modulus algorithm (CMA). Simulation results show that the introduced technique is able to equalize a class of channels for which the standard CMA algorithm fails to converge and that it also works well in a digital subscriber line (DSL) environment.