Help with FFT mode

Thanks for your explain Kaz. That makes things a bit clearer.

I looked closely to the way I feed the FFT and came to the conclusion that my system is correct. I also now know that the output of the FFT is a block floating point and not a signed integer like I treated it. I now have to find a way to do my Pythagoras thing with this floating point. There is a mecacore function wich converts floating points to integers, but this function only works with 32-bit floats. I got a 14 bit mantissa from the FFT and a 5 bit exponent. I will have to look for a solution here.