Title:Infinite non-causality in active cancellation of random noise

Abstract: Active cancellation of broadband random noise requires the detection of the incoming noise with some time advance. In an duct for example this advance must be larger than the delays in the secondary path from the control source to the error sensor. In this paper it is shown that, in some cases, the advance required for perfect noise cancellation is theoretically infinite because the inverse of the secondary path, which is required for control, can include an infinite non-causal response. This is shown to be the result of two mechanisms: in the single-channel case (one control source and one error sensor), this can arise because of strong echoes in the control path. In the multi-channel case this can arise even in free field simply because of an unfortunate placing of sensors and actuators. In the present paper optimal feedforward control is derived through analytical and numerical computations, in the time and frequency domains. It is shown that, in practice, the advance required for significant noise attenuation can be much larger than the secondary path delays. Practical rules are also suggested in order to prevent infinite non-causality from appearing.