Alike fix_fft, this works with int8 to compute faster, here the only change are:
-
reduced size of sine wave to 1/4,
-
optimized values in the sine wave array,
-
wiped the fft reverse,
-
imaginaries values are created by the mini_FFT routine and start with zero value,
-
changed process for the values:
-
INPUT, only send sampled signal (real parts array),
-
temporary using an array for the imaginaries (same size than the real part array)
-
OUTPUT, both reals and imaginaries are recorded into the real array
r = [0 to n/2[ and i = [n/2 to n[
-