Inversion of Audio Systems by Direct and Indirect Adaptive Methods

Mathias Johansson and Jonas Rutström

Master Thesis, Report UPTEC IT 99 002, Uppsala University

Abstract:

This thesis describes inversion of hi-fi systems. Direct and indirect adaptive methods are used for inversion of loudspeaker systems. Three algorithms have been implemented and tested, Filtered-x LMS, Idealised Filtered-x LMS and an optimal linear quadratic feedforward regulator.

Two fundamentally different methods for room inversion are presented, Idealised Filtered-x LMS and a direct non-adaptive solution based on measured impulse responses of the room. The work shows that it is possible to improve the quality of the sound in an audio system with the use of an ordinary PC workstation.

The techniques presented for loudspeaker inversion are successfully applied on three speaker systems of significantly different qualities. Experiments show that the Idealised Filtered-x LMS algorithm is the better of the three algorithms used for loudspeaker inversion. This conclusion is based on convergence rate, stability and robustness.

For room inversion, the Idealised Filtered-x LMS algorithm is more precise in its inverse than its non-adaptive competitor, but slower and not well-suited for large rooms. The non-adaptive method in contrast, is recommended for large rooms where an exact inversion might not be possible.

Institution:

Performed at the Signal and Systems Group, as a development work for the Project course in Adaptive Signal Processing

Thesis Advisor:

Mikael Sternad

Related publications:

Master Thesis by Daniel Botella