Measurement noise in feedback active noise control systems
MetadataShow full item record
The works presented in this thesis focus on the occurrence and effects of measurement noise in feedback active noise control (ANC) systems and propose a new feedback ANC algorithm. In feedback ANC systems, instead of measuring the reference signal, this signal is generated as an estimate of the primary noise. Therefore, feedback ANC uses only one error sensor(s) to perform noise cancellation. Since the reference sensor is absent, this configuration prevents the acoustic feedback inherent to feedforward ANC systems from affecting the system. Measurement noise is an unavoidable problem in measurements on physical systems. The effect of measurement noise on feedback ANC systems is presented with performance comparison to a system without measurement noise. An analysis of the effect of measurement noise on the performance and stability of the feedback ANC system is presented. The sensitivity and complementary sensitivity functions from control theory are used to analyze the stability of a system and computer simulations are used to validate the analysis. A solution is proposed to alleviate the effect of measurement noise on feedback ANC systems. In feedback ANC systems, the multi-tonal noise used as a reference signal is synthesized from the error signal, which is generally corrupted with measurement noise. The proposed algorithm estimates the frequencies of the multi-tonal noise and generates an enhanced version of it. This enhanced signal is used as a reference signal in the feedforward ANC configuration. An analysis of the effect of the accuracy of frequency estimation on the performance of ANC is presented. Simulation results to verify the efficacy of the proposed algorithm are also presented.