You mention that your FastDSP core within your CODECs lead to best in class ANC performance. Can you please tell me a little more about how this is achieved?
In a headphone usecase, hybrid (feed forward + feed back) active noise reduction requires very low latency at very low power consumption levels. This is very hard to achieve with FFT-based spectral subtraction techniques - and impossible to implement at ultra-low power levels. Our proprietary FastDSP is a highly efficient time-domain based IIR engine taking advantage of highly stable cascaded biquads as a means for implementing adaptive flilters for ANC. Our solution is capable of adapting filter coefficients on the fly at very low power levels and achieving very wide bandwidth noise cancellation. As a matter of fact, it easily exceeds the physical limitation of most industrial designs.
As the algorithm is based on headphone use-case, I guess that it doesn't include a compensation filter to eliminate the loudspeaker backpropagation to the reference microphone. But do you think it can be done on your FastDSP so it can be used on more general scenario?
Thank you for your question. As you point out the algorithm is a Hybrid ANC algorithm with a FF filter that targets cancelling the primary path noise and a feedback filter that cancels the residual secondary path noise. There are adaptive features that will compensate for the fitting of the earbuds where there is not a perfect seal and leakage will occurs. Designing and tuning a stable ANC solution that holds up in all corner cases is a challenge and while it is easy to tune in a lab based scenario we are lucky to work with our partner Austrian Audio to optimize for corner case scenarios.
I'm working on a different usecase, ie. fan noise with a homemade hybrid FIR algorithm where latency is much more higher which is not such a problem for in-duct control.
But I also would like to evaluate some tunable off-the-shelf solutions for no-duct fan noise control which share some similarities with headphone control (except for the absence of passive noise attenuation between ref mic and control source).
Is it possible to test the algorithm with the ADAU1761 on a SHARC Audio Module board or should I go for another one?
