• State TI Thinks Resolved
  • Locked Locked
  • Replies 3 replies
  • Answers 2 answers
  • Subscribers 30 subscribers
  • Views 190 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.

AWR2944: A question about angle interpolation

Yang yiying
Prodigy 10 points
Part Number: AWR2944

Tool/software:

Angle interpolation method is used and the azimuth FFT num is 48 in 2944 SDK,while calculate the angle in function DPC_ObjDet_estimateXYZ.

My question is——why angle interpolation method can improve the Azimuth FFT precision from 48 to 256? How does it work? Could you provide some reference documents?

I wonder if I change the Azimuth FFT bin from 48 to 96, what the Azimuth FFT precision will be?

Looking forward to your relpy,Thank you.

  • 0
    TI__Mastermind 20770 points

    Hi,

    We perform quadratic interpolation around the peak and get the offset of the interpolated peak. Using quadratic interpolation one can determine the parabola passing through the given three points. In this case, the three points are the azimuth FFT peak, peak - 1 and peak + 1. This is a method to estimate the values of the curve between the given three points. Increasing FFT size, will increase the time taken by HWA to perform FFT. So, instead of increasing the FFT size to attain finer peaks, we estimate the location of the peak using interpolation.

    By performing the quadratic interpolation, we get the fractional azimuth index of the peak. In AWR2944 demo, we use LUT size of 128 (cossintable.c). The scale factor would be 128/Azim_FFT_size = 128/48 = 2.67. Resulting peak index would be 2.67 * interp_peak_index. In this way, even if we are not performing large size FFT, by performing quadratic interpolation, you can get a better peak estimate.

    Note: We perform 32-point Azimuth FFT in the SDK (mmwave_mcuplus_sdk_04_07_00_01).

    You can search about quadratic interpolation on the internet and you should be able to find multiple resources.

    Regards,

    Samhitha

  • 0 in reply to Samhitha Bobba
    Prodigy 10 points

    Tnank you for your reply. And I want to know what value should the scale [ LUTsize / Azim_FFT_size ] be set to be appropriate?

  • 0 in reply to Yang yiying
    TI__Mastermind 20770 points

    Hi Yang,

    You can experiment with default settings where we use 32-point azimuth FFT and LUT size of 128. Later based on your observations, you can change the configuration. Do remember that increasing Azimuth FFT size will increase the time taken to process one frame data. On the other hand, if you increase the LUT size i.e cossintable, you will be consuming additional memory.

    Regards,

    Samhitha