• State Resolved
  • Locked Locked
  • Replies 3 replies
  • Subscribers 32 subscribers
  • Views 279 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.

Original question:

IWRL6432: Non-coherent integration along doppler dimension

IWRL6432: About coherentDoppler

Kei
Intellectual 750 points
Part Number: IWRL6432

Hello TI team,

I have a question about coherentDoppler which is 4th parameter of sigProcChainCfg in mmwaveL SDK 5.x.

This question is related to the thread below.

https://e2e.ti.com/support/sensors-group/sensors/f/sensors-forum/1239750/iwrl6432-non-coherent-integration-along-doppler-dimension

In the above thread, noise level increase is explained in terms of leakage factor in angle FFT.
But when I checked range-profile in visuzliser.exe, noise level increase is observed even where there is no peak.
Therefore I guess that the leakage factor is not relevant, is my understanding correct?
Please see the figure below.

I made a simple diagram below. In case of (a) which is only the maximum peak in doppler dimension is selected, SNR of range direction is 6, but in case (b) which is non-coherent sum, SNR is 1.9. Is this the basic principle of SNR degrades(noise increase) ?
If so, if we adjust CFAR threshold in each case, I think there is no difference between the two case in terms of detection performance (e.g., max detection range), is it correct? In the above thread, it is said that non-coherent sum is more robust, but I don't understand why it is robust. Could you let me know more detail of it?

  • 0
    TI__Expert 4448 points

    Hello, 

    Sorry for the delay, our teammate has been out of office. Allow me to get an expert on this. Thank you.

    Regards,
    Luke

  • +1
    TI__Intellectual 790 points

    Hello Kei.

     

    Thank you for diving into this. In the other E2E thread, I mentioned about the angle leakage as one of the root causes. But as you mentioned, it is not the only reason. However, although I explained it along the angle domain, the overall effect here is similar to the one you depicted; when non-coherently combining the Doppler spectrum, we are also adding up the noise, which increase the noise floor.

     

    The example scenario you depicted is a good case, but it covers a very specific scenario. If we have a clear single Doppler peak in the Doppler spectrum, selecting the peak only will always increase the SNR, as you showed. But in a real world scenario, the reflections from a single point will typically have more than a peak. You can think about a typical human target. We have multiple moving parts that will create multiple peaks in the Doppler spectrum. Therefore, combining the whole spectrum instead of selecting the maximum peak only typically creates better statistical results, as we observed. If you select the single-peak only, you will loose the contribution of the other peaks. This is the main reason of the robustness I mentioned in the previous reply.

     

    In general, if you only expect a single peak per angle/range, the SNR improvement after the single peak selection is expected. But in many real world scenarios, we expect multiple peaks in the Doppler. So, we may want to add up all the reflections to create a better statistically result. 

    But since the coherent peak / non-coherent combination is a configurable option, you can select the one you found better in your particular case.

     

    Regards.

    Muhammet

  • 0 in reply to Muhammet Yanik
    Intellectual 750 points

    Hi Muhammet,

    Thank you for the detailed explanation. I understand.

    Regards,
    Kei