M.Sc Thesis

M.Sc StudentYedvab Eran
SubjectDriving Speakers by Filter Bank Generated Equalizing
DepartmentDepartment of Electrical and Computer Engineering
Supervisor PROFESSOR EMERITUS Arie Feuer


In this work a new technique for loudspeakers is proposed. This technique has the potential to reduce the distortions caused by typical loudspeakers. As a background, various aspects of sound reproduction in sound systems are reviewed.

The suggested technique is based on the characterization of the loudspeaker’s components by their impulse responses. Driving the components with an impulse train generated by a special Drive-Unit will force them to output translated replicas of their impulse responses. When properly weighted, these can be used to reconstruct the audio signal.

The loudspeaker’s components and the signals’ combination in the acoustic medium are introduced as an inverse transform.

The distortions caused by the loudspeaker are equalized by a preprocessing subsystem that performs the forward transform.

The problem is analyzed mathematically. The theory of Frames is briefly summarized as a background with an analysis of the frame’s and dual frame’s ability to reconstruct signals.

An analysis is given to the capability of arbitrary impulse response function joined with its translated replicas to span the space of all the signals in the same frequency band. Furthermore, the ability to span some wider frequency band space by several partially overlapping impulse responses is also analyzed.

The loudspeaker’s components are characterized as the reconstruction side in a filter bank, while the dual equalizing filters are designed as the analysis filters in the filter bank.

Four strategies are presented for the realization of the dual preprocessing subsystem. The design objective of the filter bank is to equalize the whole system’s response in the audio frequency band to be as close as possible to a pure constant delay. Simulations were conducted calculating the equalizing filter bank by a standard optimization algorithm.