Analysis of the quantization effects of LMS complex algorithm in digital adaptive filters

Abstract

A novel theoretical formulation is developed for the errors due to the digital implementation of the LMS algorithm in adaptive filters. An analytic expression is determined for the mean-square error at the output of the adaptive filter. This expression is valid both for transient conditions and for the steady state. The convergence conditions of the algorithm are studied and a very simple and useful expression is obtained for the residual mean-square error. Simulation results are presented to validate the proposed theoretical model.<>