DEQX Improves Speaker Sound Quality with MATLAB

DEQX’s audio-correction technology needed to enable users to measure their loudspeakers and rooms, and from those measurements create correction filters that improve the resulting sound from the speaker in a particular environment. With such a varied range of input data, their signal processing algorithms must be extremely robust.

"It is difficult to reliably design a loudspeaker that is accurate across its entire bandwidth," Mr George notes. "Furthermore, it is impossible to design an analog crossover that is steep enough to separate the interaction of each loudspeaker driver. We needed to design signal processing algorithms to correct these problems."

Finally, in developing a product where there wasn’t much existing technology in the field, DEQX was under pressure to deliver the best product in the shortest timeframe—before their competitors.

Using MathWorks tools, DEQX developed custom speaker and room correction software, part of a range of products based on the calibration correction process. This process enables speaker designers to provide levels of accuracy and clarity in speakers to achieve the best possible reproduction of audio source material.

DEQX’s development methodology involves initial research, followed by a proof-of-concept and testing stage to confirm algorithms work as expected. After reaching a satisfactory solution, they develop and incorporate a real implementation into the existing software.

During the initial stages of software development, DEQX engineers used MATLAB to develop and test algorithms that will run on their custom DSP platform. They wrote the MATLAB code in a similar way to the code that would run on the DSP. Using this method, engineers confirmed that the results were mathematically correct and avoided spending unnecessary time writing custom C code. They also used MATLAB to generate intermediate results to test the algorithms in a DSP simulator.

"We tested our ideas quickly and accurately with MATLAB," says Mr. George. "MATLAB was such an important component of our design process that I can't imagine how we would have progressed without it. Our only choice would have been to write our own version of it!"

Using MATLAB and the Signal Processing Toolbox, they implemented signal processing functions to determine which components would work on actual hardware before implementation. This helped them to evaluate the effectiveness and feasibility of their ideas before committing time to further development.

After developing their algorithms in MATLAB, engineers used the MATLAB Compiler to compile these algorithms into C++. The compiled algorithms were then integrated with the graphical user interface that they developed with C++. This approach enabled DEQX to take advantage of both languages—MATLAB for the mathematical algorithms and C++ for the user interface software.

"The integration between the compiled algorithms and Microsoft Visual Studio is seamless," says Mr. George. "We can change our MATLAB files, recompile them with the MATLAB Compiler, and link them again in Visual Studio."

DEQX has already released the first version of their hardware and software. They are using MathWorks tools to develop a more efficient version of their firmware to work at higher sampling rates.