Enhanced Convergence Distributed Arithmetic based LMS Adaptive Filter using Convex Combination

Abstract

In this paper, we present a new enhanced convergence distributed arithmetic (DA) based pipelined least-mean-square (LMS) adaptive filter. It is based on the convex combination of two adaptive filters which gives the best of convergence speed and steady-state error. The overall cost of proposed filter is reduced by realizing both the adaptive filters using DA with serial implementation of look-up table (LUT). Compared to the best existing scheme, the proposed filter involves significantly less number of adders and registers without any multiplexers in LUT. The convergence performance of proposed filter is improved by selecting two step-sizes, in the orders of $O$(1/ N), where $N$ is filter order. Hence, it provides a low complexity approach to design DA based pipelined adaptive filter for improved convergence. Application Specific Integrated Circuit (ASIC) results show that a 16th order proposed ADF with 4th order base unit occupies 21.98 % less area and consumes 17.4 % less power as compared to best existing scheme.