21-24kHz Data Acquisition with Matrix Voice


I’m working on an application to record signal with frequencies from 21kHz to 24kHz. I have few questions:

  1. Does Matrix Voice support 96KHz sampling rate? I have recorded signal using HAL code and the results seem to be aliased. If not, what is the maximum sampling rate that it supports?

  2. Is the filter on FPGA core adjusted based on input sampling rate from HAL code? Should we make changes in the FPGA core if we want to support 21-24KHz?

  3. Do you know the frequency response of microphones on Matrix voice? Do they support more than 21kHz? Datasheet does not cover this information, I was wondering if you have tested them.

I appreciate if you can help me with these questions.

Hello @mehrdadh,

Apologies for the delayed response! Addressing your questions below.

  1. The MATRIX Voice is currently having some DC offset issues with the 96 KHz sampling rate. We will be doing some testing to figure out why that is.

  2. The FIR filter coefficients change based on the input sampling rate from HAL as shown here. 21-24 kHz audio should be possible to sample with the available sampling rates, but you may need to create additional filters for good sound quality.

  3. We have not tested audio frequencies beyond 21 kHz so don’t know what the response will look like. Theoretically, the mics should be able to record (sampling rate)/2 audio.

You can check out the datasheet for frequency response upto 21 kHz. I am including it below for your convenience.



I consulted with our hardware engineers and our microphones are actually able to record beyond 21 kHz. Theoretically, they can record up to (sampling rate)/2 audio frequency, so even with 48 kHz sampling rate, you should be able to record up to 24 kHz audio.

You may have to implement whatever extra processing that may be necessary (like compensation filters) to support those higher frequencies.

We currently have a 128th-order FIR filter implemented in our FPGA.

We haven’t tested beyond 21 kHz audio ourselves though, so we don’t know what the audio response will look like there.

Apologies for the inaccurate response earlier. I am still learning! I will edit my previous response to reflect the circumstances better.


1 Like