A dual mode LMS nonlinear data echo canceller

A dual-node nonlinear adaptive echo canceller that operates in two phases is proposed. In the first phase, it performs echo-path transfer-function identification with a standard LMS (least-mean-square) algorithm. In the second phase, the same structure identifies nonlinear components arising from digital-to-analog conversion of each data symbol and intermittently, at every Mth cycle, updates the echo-path transfer function. This dual-mode echo canceller requires only an additional small RAM (random-access memory) and the same processing power as the linear LMS echo canceller. The algorithm is superior to other nonlinear echo cancellers in both structural simplicity and convergence speed