Self-improved error reference for adaptive echo cancellation in full-duplex baseband data communications

Abstract

A novel adaptive echo canceller for baseband data communications working with PAM (pulse-amplitude-modulated) signals is presented. The goal is to take advantage of the prior knowledge of the signal statistic to improve the error reference of the adaptive algorithm in situations of an advanced convergence state. In these cases, when the incoming signal level is clearly above the residual echo level, the adaptive algorithm cannot correctly 'see' the right error reference, the residual echo, thus slowing down the convergence. The authors use a strategy to self-cancel the masking effect produced by the incoming signal that speeds up the overall convergence of the algorithm. A modified LMA (least mean absolute) algorithm is used to keep the computational burden low.