• State Resolved
  • Locked Locked
  • Replies 8 replies
  • Subscribers 31 subscribers
  • Views 1909 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.

IWR1443: SNR Calculation in PostProc Tool

Enric Gardein
Intellectual 500 points
Other Parts Discussed in Thread: IWR1443

Hello everyone,

I am currently doing some measurements with IWR1443 + TSW1400. To evaluate the results I am using the Radar Studio PostProc tool as well as my own signal processing tool.

In my own tool I implemented a CFAR like described in HWA Part 2 manual ( http://www.ti.com/lit/ug/swru527a/swru527a.pdf ). So the CFAR calculates the noise (or noise + clutter) for each range-doppler cell.

Then I would think that you take the peak value of the angle FFT, because this value should correspond to the peak of the signal from your object. Now you can devide this peak value with the estimated noise(+clutter) value of the CFAR .

Unfortunately, my values are not equal to the SNR values estimated by PostProc tool. 

Could you please provide any information of the way PostProc calculates the SNR (or tell me if my understanding above is correct)?

Thank you very much,

Enric

  • 0
    TI__Guru 50240 points
    Hi Enric,

    I've asked a Radar Studio Expert to look into this and someone should have an answer for you today or tomorrow.


    Cheers,
    Akash
  • 0 in reply to Akash Gondalia
    Intellectual 500 points
    Hi,
    do you have any updates?
    Thank you,
    Enric
  • 0 in reply to Enric Gardein
    TI__Mastermind 28055 points
    Hi Enric,

    In your CFAR implementation, how large are your guard windows and reference windows? Guard windows are the bins around the bin being analyzed that are not used for CFAR, reference windows are averaged for CFAR. Do you do Cell Averaging, Largest Of, or Smallest Of. In largest and smallest of, the larger or smaller half after averaging(respectively), is compared to the bin being analyzed, and the other half is discarded.

    Which version of Radar Studio are you using, and which graph in the Post Proc screen are you getting SNR from?

    Thanks,
    Justin
  • 0 in reply to Justin Curewitz
    Intellectual 500 points

    Hi Justin,

    In my tool, you can configure the CFAR dynamically with the parameters you also have for the CFAR of the IWR1443 HWA. When I was checking my SNR values against the TI values, I used the same parameters (guard cells, reference cells, CFAR-CA(SO/GO) as the RadarStudio PostProc Tool.

    My RadarStudio Version is in the package DFP-00.09.01.06. I get the SNR values from PostProc from the table, tab "Detected Objects".

    Do you have any thoughts on the steps for the SNR calculation I described in my first post? Can you confirm that my understanding is correct?

    Thank you,

    Enric

  • 0 in reply to Enric Gardein
    TI__Mastermind 28055 points
    Hi Enric,

    Radar Studio caclulates SNR with this formula:

    SNR = 10*log10(CUT power/noise-floor estimate)

    CUT power = |CUT_value|^2
    noise-floor-estimate = RMS(reference samples)

    Regards,
    Justin
  • 0 in reply to Justin Curewitz
    Intellectual 500 points

    Hi Justin,

    thanks for your reply.

    Okay, I understand your formula.

    Basically, this calculates the overall SNR in that CUT. 

    But wouldn't it be more precise to calculate the SNR with the peak value of your angle FFT peak?

    Also, since the CFAR 2D matrix is non cohrenetly summed, you have a loss in your CUT peak value, which you would not have, if you use the angle FFT peak value.

    Am I missing something?

    Thanks,

    Enric

  • 0 in reply to Enric Gardein
    TI__Expert 4720 points

    HI Enric,

    You are not really missing anything. At the end of the day, your estimate of SNR (computed after the 3rd dimension FFT) is a better estimate than the estimate within radar Studio (which is computed after 2D-FFT and non-coherent accumulation across antennas).

    The only advantage to the computation from within RadarStudio is that we can output an SNR number as a byproduct of detection (which is done before angle-estimation). In our internal testing, we then can play with the CFAR threshold factor, until we see sufficient detections.

    The SNR that we compute during detection can be used to speed up that process.

    Regards

    Anil

  • 0 in reply to Anil Mani
    Intellectual 500 points
    Hi Anil,

    thank you very much! :-)