Inversion of Loudspeaker Dynamics by Polynomial Lq Feedforward Control

Loudspeakers always introduce linear and nonlinear distortions in a stereo system. It is however possible to partially cancel these distortions by digitally pre-inverting the signal. We here discuss one of several possible solutions to this problem, namely the use of a linear inverse obtained from a measured ARX loudspeaker model. The inverse is obtained by well known polynomial methods as a H2-optimal feedforward lter. Due to numerical reasons, sound signals sampled at 44kHz will have to be separated into 3-6 sub-bands, to reduce the complexity of the model in each band to less than 70 poles and 70 zeros. An inverse is then implemented in each band and with this implementation, the method works well. The resulting performance is competitive to that obtained with direct adaptation of FIR pre-compensators by the idealized ltered-X algorithms, and a far shorter computation time is required for the lter tuning.