Narrowband active noise control and equalization algorithms
The research in this thesis attacks some obstacles in Active Periodic Noise Control (APNC) applications to realize new system functions or to improve system performance. Active noise equalization emerges to meet human preference in psychological acoustics. Conventional APNC systems just cancel the incoming noise completely and provide the infinite-depth notch. By definition, the equalization means that the depth of the notch is adjustable and even inverse. Zero-pole analysis shows that realizing equalization corresponds with moving the system zeros into the unit cycle. Thus, a novel Pseudo Error Minimum Method (PEMM) is developed to satisfy the requirement for the Least Mean Square (LMS) algorithm convergence and achieve the adjustability of residual noise amplitudes. Four operation modes of the proposed equalizer (cancellation, attenuation, pass and enhancement) include all functions in the definition of equalization. Simulations verify the success of the PEMM. The passband disturbance in APNC has been a problem in system performance. The analysis of the open loop transfer function of APNC systems exposes the occurring mechanism of this kind of disturbance, thus deriving out the necessary and sufficient conditions causing disturbance. A Filtered-E LMS (FELMS) algorithm is developed to limit the high gain in the passband below a certain threshold. Simulations show that the FELMS algorithm is effective in reducing the disturbance while using a relatively large (I to improve the convergence rate. When the error path is complicated and strong uncorrelated components are present in primary input, the FELMS algorithm can avoid amplifying these uncontrolled components thus increasing system stability. Acoustic feedback problem is common in APNC systems because of using input microphone to obtain the reference signal. An alternative to avoid this problem is to convert the Revolutions Per Minute (RPM) signal into the reference signal. Two RPM conversion techniques are developed for the Filtered-X LMS algorithm to cancel the periodic noise from rotating machines. Computer simulations in APNC systems with four types of configuration combinations show the effectiveness of these. tracking techniques. Real-time work with RPM tracking techniques applied to 3-D APNC systems was conducted intensively. Experiments show that the noise reduction achieved by using the Direct Filtering (DF) technique is better than using the Frequency Measuring (FM) technique. The post-fact generation of the reference signal by the FM technique suffers the shaking phenomenon of the RPM signal which causes inaccuracy of the frequencies of the reference components. The DF technique is highly recommended to APNC when considering the RPM tracking technique for canceling the engine noise. In addition, some suggestions are presented for further research.