A posteriori analysis of adaptive blind equalizers

For adaptive filters that employ training signals, the relationship between the a priori and a posteriori error signals can be used to quickly and easily characterize the stability and robustness of any given adaptive algorithm. In this paper, we extend these analytical techniques to Bussgang-type and Godard/CMA blind equalizers. We derive conditions on the nonlinearity and step size parameter to guarantee L/sub 2/ stability of the adaptive filter coefficients, and we study the resulting behavior of the algorithms in a deterministic context. In addition, we give conditions under which an algorithm that is based on an a posteriori blind error criterion is L/sub 2/ stable.