This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TMDSEVM6657: DSP lib - DSP_fft16x16 result incorrect with Matlab FFT SIZE 8192

Part Number: TMDSEVM6657

Hello.

Now I'm trying to make QPSK to 16x16 IFFT , So I put even side to real, and odd side to imag.

and adjust all of Scale.

But Output signal's is not corret with matlab function.

eg. input = [1 -1 1 -1 1 -1] (= (1-1j, 1-1j, 1-1j) )

Can you help me how to make Input for complex.

  • Dear Customer,

    From the dsp Lib package "dsplib_c66x_3_4_0_4\packages\ti\dsplib\src\DSP_ifft16x16\c66

    I could see many sample functions for fft. such as DSP_ifft16x16, DSP_ifft16x16_cn, gen_twiddle_ifft16x16 etc.....

    ----

    If your question is just how to make inputs for those fft functions, please go through the .c files which tests these FFT and produce results.

    For example, DSP_ifft16x16_d.c file is the one which tests the  DSP_ifft16x16() and produce the result in stdout. 

    This program file will have the input too which will help you understand how to frame the input...

    -----

    Check, you are able to produce the expected result in CCS for that particular function and compare the results with Matlab.

    -----

    Supporting Matlab is out of scope for Processor forum.

    ----

    Apllcation notes for dsplib: https://www.ti.com/lit/an/spracn4/spracn4.pdf

    Regards

    Shankari

  • umm, acturaly when i not consider malab result , i can't check dsp out is acurrate result and when i put all imag part for 0. that result is similaly with matlab out put But when i put imag part for not 0, dsp result is different with matlab output. I'm already considered about scalling isuue So input data Scaled to log4(nx)-1.

  • Hi Kwon,

    My suggestion is to run the any of the dsplib projects in CCS and test the ready made examples available for FFT. 

    FOR example ---- DSP_ifft16x16, DSP_ifft16x16_cn, gen_twiddle_ifft16x16

    These examples are made in such a way that it results the output in the CCS window.

    Have the inputs with different combinations as you mentioned, as non-0 for imaginary part and 0 as imaginary part etc.... and observe the output data.. In this way, evaluation can be done and narrow down the issue.

    Regards

    Shankari