Hi there. I followed this tutorial (https://calibronic.com/tinyml-stm32-audio-recording-for-speech-recognition-part-1/), but I'm currently facing a problem. I am using an STM32H750 with the following parameters:
DFSDM clock is 120 MHz
DFSDM CLKOUT is 120 / 39 = 3.072 MHz
FOSR (Filter Oversampling Ratio) = 70
IOSR (Integrator Oversampling Ratio) = 1
Filter type: FastSinc
DFSDM input clock: 3.072 MHz
RightBitShift is 0x03
Target: 16-bit PCM (int16_t)
Sample rate is 44 kHz
With these settings, the data I get can be played back correctly.
Now, I want to change the filter type to Sinc4 while keeping all other parameters the same. I now have two problems:
The data obtained with Sinc4 is unusable and cannot be played.
After switching to Sinc4, I still have the bit depth set to 16 bits, but setting RightBitShift to 0x03 is incorrect, and the data cannot be played.
How should I calculate the RightBitShift value for the Sinc4 filter?
I tried the calculation method described in the tutorial, but it didn't work for me.
I'd also like to understand: if I change the DMA length from Word to Half Word, can the HAL_DFSDM_FilterRegularStart_DMAmethod be replaced with the HAL_DFSDM_FilterRegularMsbStart_DMAmethod?
