MotionSP FFT q15 or q31 instead of float32

[Chubs]

Hello Community,

I am trying to implement a basic FFT computation on our STM32G0 MCUs using MotionSP for a 16-bit accelerometer signal.

I am not sure why does MotionSP by default uses float32 for FFT when the original sensor signal is 16-bit?

Can I then simply reconfigure the data type in the MotionSP by changing the float32 to q15?

Kindly guide, as we do not have FPU to compute float32.

Also, does a 16-bit sensor output requires a float32 FFT?

@Eleon BORLINI 

[Federica Bossi]

Hi @Chubs ,

You can still try using q31 or q15, but you need also appropriately to modify the other parts of the code configured to handle float data.