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.

TMS320F28379D: Max fft size

Part Number: TMS320F28379D
Other Parts Discussed in Thread: C2000WARE

I have read in the forum that some TI staff said: 

"that Controlsuite fft libraries, limit FFT sizes from 2^5 through 2^10, which is not correct. This range only reflects the values for which pre-computed twiddle factor tables are available. As we have seen, FFT sizes through 2^14 work properly, however 2^15 does not work properly, and I have indicated in the JIRA ticket that this information needs to be present in the user's guide."

I would like to know if this issue is solved and what is the max complex FFTsize possible. Is it necessary to do something on the memory map?

Regards

  • Fransciso,

    This information still holds, and there has been no update. 16384 RFFT was the max that seemed to work, but that was assuming the necessary RAM was available. 

    Thanks,

    Sira

  • Thank you Sira, but I asked for COMPLEX FFT, the post above, stated that limit for REAL FFT.

    Is it the same or half?

    Does f28379D have enough the RAM for it?

    Is it possible to extend that FFT with a custom made function? If so, do you have an example?

    Best regards

  • OK, I'll look into it. It's the holiday weekend in the US, so I will get back to you tomorrow.

    Thanks,

    Sira

  • My thinking is for the Complex FFT, it would be half that size, to account for real and imaginary coefficients. But again, the original constraint wasn't related to size - there was enough room for it, still it didn't work, and this is what we have not spent time troubleshooting. So I can't say for sure that 16384 CFFT won't work. I'll be curious to know if you tried it. I would certainly expect atleast the 8192 CFFT to work.

    What kind of custom made function do you have in  mind? Our examples are all out there for you to see in C2000Ware.

    Thanks,

    Sira

  • Sira Rao80 said:

    My thinking is for the Complex FFT, it would be half that size, to account for real and imaginary coefficients. But again, the original constraint wasn't related to size - there was enough room for it, still it didn't work, and this is what we have not spent time troubleshooting. So I can't say for sure that 16384 CFFT won't work. I'll be curious to know if you tried it. I would certainly expect atleast the 8192 CFFT to work.

    No harm intended, but I think that work should be on vendor's side.

    Sira Rao80 said:
    What kind of custom made function do you have in  mind? Our examples are all out there for you to see in C2000Ware.

    A C programmed FFT function. Do you have something similar?

  • Ok, I'll have to make time for this. It may take a couple of days. I'll see what I find.

    Our FFT implementation is in C28x assembly for optimization purposes. I don't believe we have a C version. I will check that as well. Pretty sure there will be C versions you can find online and run without issue on our devices.

    Thanks,

    Sira

  • Thank you Sira, that would be fantastic!!

    Looking forward to hear from you about both issues.

  • Unfortunately I didn't find the time to test what I planned on testing this week. I am not sure I will find the time either, since I am not the sole owner of these libraries. I will have to consult internally. If you give me an email address, I can discuss this separately with you.

    As for the C FFT, it should be at the following path in C2000Ware

    \C2000Ware_2_00_00_02\libraries\dsp\FPU\c28\reference\C\source\fft

    I will close this issue and take this up via email with you.

    Thanks,

    Sira

  • Thank you for the interest Sira.

    Could you send me a private message, so I could give my email address to you?

    Regards