Robust loudspeaker equalization based on position-independent excess phase modeling

A well known problem in loudspeaker equalization is that mixed phase design of the inverse filter causes residual "pre-ringings" in the equalized system, due to the spatial variability of loudspeaker- room transfer functions. A common strategy for robust and perceptually acceptable equalization is therefore to use minimum phase filters only. In this paper, a method for cautious mixed phase equalization is proposed. By analysis of a set of room transfer functions, it is concluded that some non-minimum phase zeros are insensitive to receiver position, and can therefore be robustly inverted. The method improves upon a minimum phase equalization by extending the minimum phase model with a robustly invertible all pass link. Validation measurements show that the time-domain aspect of equalization is improved throughout the spatial region of interest, while pre-ringings are kept at a very low and prespecified level.