Fifth International Congress on Sound and Vibration

This paper presents a simple adaptive algorithm based on the adaptive notch two-taps filter [1] and the filtered-X LMS algorithm [2], to be used in active noise control applications, The system consists of two adaptive algorithms working together. One algorithm estimates on-line the error path transfer fi.mction as proposed in [3] and the other cancels a i%equency narrowband of the residual signal, The convergence factors of the two adaptive algorithms determine the spectrum width of the cancellation and the stability of the control system, The adaptive filter used for estimating the error path has also two taps; it is only necessary to estimate the module and the phase at the center frequency of the cancellation bandwidth. The system can be generalized to a configuration of multiple cancellation frequency narrowbands and a multiple notch filter control system. This adaptive algorithm can be used in active noise control applications to cancel only frequency narrowbands and when it is not possible to get any reference signal. INTRODUCTION. In some active noise control configurations it is not possible to get a cleun reference signal of the acoustic noise because: a) of the working environmental conditions of the noise source, b) there are many changing noise sources or c) because there is an acoustic feedback between the control signal and the reference signal. The classical adaptive algorithms, the filtered-X LMS algorithm [2] and the filtered-U LMS algorithm [4], need a reference signal to cancel the acoustic noise around the error microphone. The electro-acoustic transfer fimction between the control output signal and the error input signal of the algorithm has to be on-line estimated with other adaptive algorithm, to get a stable convergence of the main adaptive control algorithm. It is possible to design a very simple active noise control, without any external reference signal, to cancel some specific spectrum narrowbands of the external acoustic noise [5, 6]. This active noise control is based on the adaptive notch filter proposed by Widrow [1] and on the filtered-X LMS algorithm [2] as shown in figure 1. The algorithm is inside the dashed square. When the electro-acoustic transfer fimction between the control signal and the error signal, represented by EL(z), does not exist (I%(z)=1), the algorithm is the adaptive notch filter proposed by Widrow, ............ ................................................................................................................4 # “... :“”O)(J ...