Development of a digitally compensated active loudspeaker

OData support
Dr. Bank Balázs Lajos
Department of Measurement and Information Systems

In the field of multimedia products, development trends have seemed to aim at the reduction of size over the past decade and that concerns built-in audio systems as well. Besides this, users have ever increasing requirements on sound quality. That often tends to meet with quite a few difficulties when using small size speakers. The most problematic points are the deficiency of transferring low frequency signals and the problem of having ripples in the transfer function. Up-to-date techniques of digital signal processing may offer solutions for all of these questions.

In this thesis, I introduce the methods which are often used to compensate the imperfections of a loudspeaker’s frequency response. I attempt to do this through the design process of an active loudspeaker with small wide-range speaker unit. First I describe a technique that is widely used to measure the linear frequency response and harmonic distortion of a loudspeaker applying logarithmic sweep signal. Then I review the filter types which are suitable for the equalization of a linear transfer function, including FIR, IIR, WFIR, WIIR and parallel IIR structures. I also introduce the basics of virtual bass synthesis (VBS) and two options for its realization.

After general descriptions, I present all the steps of designing my own active loudspeaker. First I give a review of building the wooden box and introduce the aspects of choosing the optimal parts (speaker, power amplifier, etc.). I also describe the basic parameters of the SigmaDSP evaluation board which has been used to implement all the compensation methods. After that I evaluate the results of the frequency response measurements performed using the “raw” loudspeaker. I give a description of the applied methods – such as Linkwitz-transformation and parallel IIR structure – which were used to compensate the defects of the measured frequency response. Their implementation on the SigmaDSP board is also discussed. Then I make a comparison between time- and frequency-domain VBS algorithms in MATLAB, and realize the better sounding one on the evaluation board. Finally, I evaluate the whole system that was set-up and make suggestions about the possible directions of future development.


Please sign in to download the files of this thesis.