Weight vector analysis of an RLMS adaptive filter with on-line auxiliary path modelling

A fully adaptive active noise control system for ducts that produces a unique adaptive model of the auxiliary and error path transfer functions has been developed. An input microphone is used to measure the undesired noise and provide an input signal to an infinite impulse response (IIR) adaptive filter that provides an output signal to a noise-cancelling loudspeaker. The IIR filter uses the Feintuch recursive-least-mean-square (RLMS) algorithm and enables the direct path and feedback from the loudspeaker to the input microphone to be modeled simultaneously on an online basis. Good cancellation is obtained with narrowband or broadband input signals, with unique coefficient vectors for the IIR adaptive filter obtained only for signals that have a sufficiently rich spectral content. Changes in the primary plant or error plant are readily modeled. Relatively long weight vectors with a high degree of resolution are required to model the response of the system accurately and completely