• State Resolved
  • Locked Locked
  • Replies 6 replies
  • Answers 2 answers
  • Subscribers 29 subscribers
  • Views 1972 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.

AWR1642: Compute Threshold value of CFAR from TdB

Biplab Raut
Prodigy 40 points
Part Number: AWR1642

Hello,

The question is about how to compute threshold value required in CFAR and what is the normal TdB range used?        

According to "MMWAVE SDK User Guide", CFAR finds the valid signal detections as per below condition:-

CUT > Threshold_scale + (noise sum / 2^x) , where this calculation is in log2 scale

For AWR16xx, the threshold scale is computed as Threshold_scale  = 256 x numVirtualAntennas x TdB / 6. Here, 256 is for the Q8 format conversion and numVirtualAntennas is because the cell values are summed over 2DFFT values for all antennas.

I want to know what should be the normal values of TdB used in above calculation for Threshold_scale and why TdB is divided by 6?

Secondly, if the 2DFFT values are in floating poinnt and not in log scale, and 2DFFT values are averaged across numVirtualAntennas , then could you confirm if the below CFAR condition is correct:-

CUT > Threshold_scale * (noise sum / 2^x) , where Threshold_scale  = TdB / 6 Is it correct?

Thank you.

  • 0
    TI__Mastermind 22155 points
    Hello Biplab,

    Have you seen the documentation of the mmw demo application . The documentation covers the details of the CFAR, peak grouping etc. Please look at the documentation in following path "$mmwave_sdk_01_01_00_02\packages\ti\demo\xwr16xx\mmw\docs\doxygen\html" .

    Thanks,
    Raghu
  • 0 in reply to Raghunandan Kamath
    Prodigy 40 points
    Hello Raghu,
    I have checked the documentation under doxygen\ folder and it does not have detailed information about CFAR and thresholding.
  • 0 in reply to Biplab Raut
    TI__Expert 4720 points

    Hi Biplab, 

    Please look at  and let us know if it answers your query. 


    Regards

    Anil

  • 0 in reply to Anil Mani
    Prodigy 40 points
    Dear Anil,
    Thank you for your inputs.
    The post suggested by you explained about why log scale and factor to 6 is used to divide to convert from log10 to log2 scale.
    But in the first part of my question, I also wanted to know what is the normal values of TdB for normal usecases?

    And in the second part of my question, I wanted to find out if the 2DFFT values are not converted to log scale but represented in average of floating point values across numVirtualAntennas,
    then is it correct to calculate "threshold scale", Tcli as below:-
    For AWR16xx: Tcli = 10(TdB/ 20) Is it correct? (here, 2DFFT float values are averaged after combining from all virtual antennas in linear scale)
  • 0 in reply to Biplab Raut
    TI__Expert 4720 points
    Hi Biplab,

    We typically require the SNR to be between 18dB and 12dB based on the requirement on the demo. For example, if we needed to detect an object of a given RCS at a certain distance, we use the radar equation to generate the expected SNR of that object, and we program that SNR.

    For the second part of your question, I think your formula is reasonable. However, remember that the output of the 2D FFT is complex, and when you average across channels, you must average using the energy of the FFT.

    Regards
    Anil
  • 0 in reply to Anil Mani
    Prodigy 40 points
    Thank you Anil for your clarification and confirmation.