A Filtered Reference – Filtered Error LMS Algorithm

Adaptive filters updated by the least mean square (LMS) algorithm are succesfully implemented in digital control systems. They are utilized for both plant identification and control purposes. In control appications (e.g. in active noise control) the output of the adaptive filter drives the plant input, and the error signal is derived only at the output of the plant. In such cases the filtered reference LMS algorithm guarantees the stable adaptation. However, the convergence of this algorithm can be very slow, depending on the plant in the control loop. The paper introduces a novel algorithm which filters both the reference and the error signal by the same filter which is designed to provide a loop gain close to the unity along the whole frequency axis. The paper illustrates the efficiency of the proposed method by examples.