Error-path modeling technique for active transformer noise control system and residual noise shaping experiments
This thesis concentrates on error-path modeling techniques with active noise control (ANC) applications. A new error-path modeling technique for an active transformer noise control (ATNC) system and an on-line error-path modeling technique for a narrowband ANC system are proposed. Real-time experiments of residual noise shaping techniques are conducted. A transformer emits a periodic and harmonic noise at dominant frequencies 120 Hz, 240 Hz, 360 Hz, etc. The modeling of the error path at these frequencies for an ATNC system is very difficult for two reasons. First, it is unacceptable to turn off the transformer. The other problem is the required excitation signals have the same frequencies as the transformer noise which becomes the interference to the modeling process. A new error-path modeling technique is proposed to solve these two problems. The new technique is verified by computer simulations and real-time experiments. Adaptive on-line error-path modeling of the narrowband ANC system is introduced to track the error-path changes. The convergence analysis has shown that on-line error-path modeling can be achieved by using the augmented predictor with sufficient time delay. A filtered-E LMS (FELMS) algorithm that allows for the shaping of the residual noise spectrum according to a predetermined frequency response is per- formed by real-time experiments in this thesis. For the narrowband noises, four difference equalization modes — cancellation, attenuation, neutralization, and enhancement - are demonstrated in real-time experiments.