• State Resolved
  • Locked Locked
  • Replies 3 replies
  • Answers 1 answer
  • Subscribers 30 subscribers
  • Views 585 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

IWR1443BOOST: how to reduce Zoom FFT processing time?

feng tang
Genius 3595 points
Part Number: IWR1443BOOST

How to reduce the Zoom FFT processing time (the total processing time is too long even with ES3 chip)? can we move Zoom FFT calculation to MSP432? If we can move Zoom FFT to MSP432, can you provide the algorithm /steps after the coarse FFT?

Thanks,

  • 0
    TI__Genius 11405 points
    Hi Feng,

    The steps after coarse FFT consist of slow DFT mode and interpolation. Please refer to previous e2e link for details:
    e2e.ti.com/.../2692341

    Thanks and Regards,
    Michelle
  • 0 in reply to Michelle Liu
    Genius 3595 points

    Hi Michelle,

    Could you provide the reference document that you use for Zoom FFT? Does the DFT that used shift the selected frequency band down to the base band? what the interpolation do in Zoom FFT?

    Thanks,

  • 0 in reply to feng tang
    TI__Genius 11405 points
    Hi Feng,

    The reference is also in previous link, re-past here: www.ti.com/.../swru527a.pdf
    Please check page 4 for the slow DFT mode details.

    Once the Peak_bin (bin number for max peak) is obtained from 1D course FFT, the slow DFT function include complex multiplication in the frequency range between the (Peak_bin -2 to Peak_bin+2). The FFT size is 16K for high accuracy. For each multiplication, the sum is saved in one HWA local memory in complex format.

    The output from previous slow DFT is then read by the MmwDemo_processInterpolation function. Each power is calculated and compared to find the fine peak index. (line 1142 of data_path.c)
    Then around this fine peak, final interpIndx is interpolated. (line 1151 of datapath.c). The interpolation is to obtain higher accuracy value based on zoom-in fft output.

    Thanks and Regards,
    Michelle