Problem while receiving data from MEMS microphone.

[rrach]

I am using the STM32F407 discovery board and Wolfson pi audio interface. I have a problem in analysing sample values received from MEMS microphone.

I wanted to plot the data sensed by the MEMS microphone, which is just sensing the surrounding environment. So, in order to achieve this, I am using an array of float type (float32_t) to store the samples continuously. After running the algorithm for a few seconds, I am plotting the values. The problem here is, when I am plotting the data, huge values (around 32000) are getting displayed on the screen. And the graph is looking like a continuous pulse waveform.

Are these true sample values? If I want to operate on these samples, can I directly do that?

Some of the threads posted earlier, related to this topic, suggesting me to convert the values into the q0.15 format and then operate on it? I couldn't get a complete understanding over there.

Do I need to do that? If yes, why we need to convert into that format?

[MikeDB]

I think you'll find the microphone is a pulse density type (PDM) and it sounds like you are feeding that directly into your algorithm without converting to PCM. I'm sure there's a PDM to PCM convertor in the library if this microphone is of this type so have a check on it first.

[rrach]

Can I2s protocol handle values like fractional numbers like 12.30, 14.5678 or any other thing like that?

[MikeDB]

No reason why not. Obviously you couldn't expect a dac to understand it but for processor to processor comms its fine
Sent from my Windows Phone