A new adaptive algorithm and its implementation in MOS LSI

A new algorithm for adaptive processes is presented here. The discussion is in the context of a particular MOS LSI sampled analog realization, but the applicability is more general. The algorithm implemented is a simplified version of the least-mean-squares (LMS) algorithm, which requires modification of only one tap for each sampling period. This is shown to give reduced system complexity, but more importantly, to give greatly increased bandwidth capabilities to any serial update realization of the adaptive filter. Residual error performance of the new algorithm has been found almost as good as the LMS algorithm in two different applications, namely, parameter estimation and noise cancellation.