More Customer feedback:
Did you hear back from NXP? I've got another question regarding the toolboxes. I'm currently trying to get the FFT going using the Matlab toolbox and so far see some differences between Matlab and the SPT toolbox that I can't fully explain. There seems to be some documentation around the SPT but I'm not sure if I have all that exists or if there is a reference manual for SPT. I do have the SPT3.X_Toolbox_User_Guide which is very brief.
My specific problem is that I want to do a complex FFT, which I presume to have the call correct, however it looks almost like I'm doing a real FFT (almost because the two halfs don't quite match up).
This is the call:
[out, acc_shift(1), exec_time(1)] = rdx4_mex(complex(Window), input, 'WIN_ENABLED', 'MULTIPLE_COEFF', 0, 'QUAD_EXT', 0, 'NO_ADPTV');
for k = 1:round4 - 1
[out, acc_shift(k+1), exec_time(k+1)] = rdx4_mex(tw, out, 'WIN_DISABLED', 'MULTIPLE_COEFF', k, 'QUAD_EXT', 0, 'NO_ADPTV');
end
if round2
[out, acc_shift(k+2), exec_time(k+2)] = rdx2_mex(tw, out, round4, 'NO_OPERANDS_SPLIT', 'QUAD_EXT', 0, 'NO_ADPTV');
end
[WinFFT, acc_shift(k+3), exec_time(k+3)] = rdx2_mex(tw, out, round4 + round2 - 1, 'OPERANDS_SPLIT', 'QUAD_EXT', 0, 'NO_ADPTV');
(window being a Hann window scaled to 2^15, input a complex 256 entry vector and tw being the twiddle factors copied over from spt3x_FFT1 from example advanced/object_detection)
And this is the result:
(note: the first half of the SPT FFT is flipped)
I can share the example if you want to have a look at it.
