A systolic array realization of an LMS adaptive filter and the effects of delayed adaptation

Abstract

A systolic array design for an adaptive filter is presented. The filter is based on the least-mean-square algorithm, but due to the problems in implementation of the systolic array, a modified algorithm, a special case of the delayed LMS (DLMS), is used. The DLMS algorithm introduces a delay in the updating of the filter coefficients. The convergence and steady-state behavior of the systolic array are analyzed. It is shown that the performance of the systolic array is similar to that of a conventional LMS implementation for a wide range of practical conditions.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>