Strategy to transition from the constant modulus algorithm to a decision feedback blind equalizer

Abstract

The constant modulus algorithm (CMA) is commonly used for first stage blind equalization applications. CMA's main advantage is that it will open the eye based on the received signal only. Decision directed (DD) equalization is often used as it provides a lower mean square error (MSE) in the equalized signal. Obvious from its name, DD equalization requires estimates of the transmitted symbols in order to adapt the equalizer - the eye must be open. This paper considers an elegant method to transition between CMA and DD equalization (in a decision feedback equalizer structure). A conventional “brute force” approach to perform the transition is to determine the filter coefficients for the DD equalizer through a MMSE design criterion based on the inverse CMA filter coefficients. This may cause a large procession delay at the point of transition between filter types and requires significant computations. The algorithm in this paper is low complexity and does not require a large processing delay when transitioning from CMA to DD LMS.