Complete Parallel Narrowband Active Noise Control Systems

Abstract

Conventional parallel-form narrowband active noise control (ANC) systems use multiple adaptive filters that are updated by the same error signal. This paper performs theoretical analysis to show that the convergence rate of every adaptive filter will be degraded by the residual error components from other adaptive filters. We develop and analyze a complete parallel narrowband ANC system that uses delayless bandpass filterbank to split the measured error signal, and uses individual error signals to update the corresponding adaptive filters. The foreground bandpass filters are updated by the background adaptive algorithm to tracks frequency change of the primary noise without introducing extra delay in the secondary path. A modified cost function is used to derive algorithm for the complete parallel narrowband ANC. The theoretical analysis and improved performance are verified by computer simulations using measured transfer functions from an experimental setup.