Previous demonstrations have shown that broadband noise reduction in feedforward active noise control systems is limited to 23-25 dB. This is suspected to be the result of nonlinearity, constraints in adaptive algorithms and causality, and the extent of these factors on the noise reduction limitation was investigated by carrying out an experiment in an 1-D duct with absorptive ends. The results show that both nonlinearity and causality contribute to the noise reduction limit. The latter issue affects the noise reduction performance because the secondary path is non-minimum-phase, such that its inverse is unstable. This limitation can be reduced by putting an extra secondary source. The Filtered-x Least Mean Squares (FxLMS) algorithm is found to be performing poorly under a coloured reference signal, both in terms of convergence speed and the optimal filter sought. This weakness can be relieved by using other algorithms.
展开▼
