MMWAVE-STUDIO: How CFAR and is performed in AWR1843 or postProC of mmWave studio?

Palguna Kumar Reddy Gopireddy

Part Number: MMWAVE-STUDIO
Other Parts Discussed in Thread: AWR1843, AWR2243

Tool/software:

I am using DCA1000 to get the data from AWR1843. I used postProc button in mmWave studio to see the plots corresponding the obtained raw data.

But I am trying to develop the script to generate the plots like rangeDoppler plot, rangeDoppler plot and pointCloud plot.

From tidueo9.pdf, I figured out that the processing of the raw data requires steps like

1) Range-FFT,

2) Dopple-FFT,

3) CFAR deptection,

4) Angle estimation,

5) Clustering.

From section 2.1 of tidueo9.pdf, I am able to perform the Range-FFT and Doppler-FFT and get the rangeDoppler plot.

But I am unable to apply CFAR. In tidueo9.pdf it is told that the CASO-CFAR detection in applied in range domain, CA-CFAR detection is applied in Doppler domain on rangeDoppler data.

Generally, CFAR is performed by comparing the cell under test (CUT) with a threshold. This threshold value varies depending on the CUT, The threshold for each CUT is obtained by multiplying the noise estimate calculated from the neighborhood of CUT with a scaling factor.

But in the miscellaneous options window of postProc in mmWave studio, it is mentioned that for CA-CFAR threshold is 18dB. But in CFAR the threshold is not constant.

So I am not able to get how the CFAR is appled by postProc of mmWave studio.

Also, in tidueo9.pdf it is told that the CASO-CFAR detection in applied in range domain, CA-CFAR detection is applied in Doppler domain. But miscellaneous options window selected CA-CFAR method only. So I am confused.

Please tell how CFAR is applied on RangeDoppler data for obtaining the point clouds for AWR1843

2 months ago

0 Kaushik Gowda 2 months ago

TI__Mastermind 20505 points

Hi Palguna,

You can refer to the CFAR source code (Used for 4 chip as opposed to just single chip) in the following folder. The underlying algorithm should be same in both the cases and therefore I recommend you go through the same.

mmwave_studio_<ver>\mmWaveStudio\MatlabExamples\4chip_cascade_MIMO_example\modules\detection\@CFAR_CASO

Regards,

Kaushik

0 Palguna Kumar Reddy Gopireddy 2 months ago in reply to Kaushik Gowda

Prodigy 131 points

Thanks. Will see it and let you know.

0 Palguna Kumar Reddy Gopireddy 1 month ago in reply to Palguna Kumar Reddy Gopireddy

Prodigy 131 points

Hi. I saw the code, But I am unable to use it since the scaling factor or probability of false alarm needed to calculate adaptive threshold is not provided. What value should I take as scaling factor or pfa?

0 Kaushik Gowda 1 month ago in reply to Palguna Kumar Reddy Gopireddy

TI__Mastermind 20505 points

Hi Palguna,

Can you please share the piece of code for reference? You should be able to follow the flow of the code and understand how these values were either set or derived.

Regards,

Kaushik

0 Palguna Kumar Reddy Gopireddy 1 month ago in reply to Kaushik Gowda

Prodigy 131 points

Here is the sample of a code used for comparing the CUT with the adaptive threshold.

if vec(j+gaptot)>K0*cellave1                % comapring the CUT with the adaptive threshold. Here 'K0' is scaling factor, 'cellave1' is the estimated noise noise power.
    N_obj=N_obj+1;
    Ind_obj(N_obj,:)=[j+discardCellLeft, k];            % (range bin, chirp number) of the detected object.
    noise_obj(N_obj) = cellave1;                        %save the noise level %estimated noise noise power of detected object
    CFAR_SNR(N_obj) = vec(j+gaptot)/cellave1;           %SNR of the detected object

Here K0 is the scaling factor and cellave1 is the estimated noise power from neighboring cells.

I understood the calculation of the cellave1. The reference window, guard window sizes are given in the miscellaneous options window.

But the scaling factor is not given anywhere. scaling factor can also be calculated from probability of false alarm, which wasn't given either.

The complete code you can get from "C:\ti\mmwave_studio_02_01_01_00\mmWaveStudio\MatlabExamples\4chip_cascade_MIMO_example\modules\detection\@CFAR_CASO\CFAR_CASO_Range.m"

For understanding CFAR-CA see in.mathworks.com/.../constant-false-alarm-rate-cfar-detection.html

0 Kaushik Gowda 1 month ago in reply to Palguna Kumar Reddy Gopireddy

TI__Mastermind 20505 points

Hi Palguna,

Please give me a week's time to review this internally and get back to you.

Regards,

Kaushik

0 Palguna Kumar Reddy Gopireddy 1 month ago in reply to Kaushik Gowda

Prodigy 131 points

ok. Please reply as soon as possible.

0 Kaushik Gowda 1 month ago in reply to Palguna Kumar Reddy Gopireddy

TI__Mastermind 20505 points

Hi Palguna,

The values of the K0 set are derived empirically based on the expected SNR and are configured in the following file:

\MatlabExamples\4chip_cascade_MIMO_example\main\cascade\paramGen\module_param.m

Regards,

Kaushik

0 Palguna Kumar Reddy Gopireddy 1 month ago in reply to Kaushik Gowda

Prodigy 131 points

Yeah. Saw that. I was wondering to know a way to decide that SNR value to calculate the K0. If you know how to decide the SNR value, like can we choose it arbitrarily or is it decided by some otherway.

I saw the miscellaneous options window in postProc of mmWave studio mentioning about the SNR value in CFAR parameters section. Is this the SNR used to calculate the K0 or some other SNR?

0 Kaushik Gowda 1 month ago in reply to Palguna Kumar Reddy Gopireddy

TI__Mastermind 20505 points

Hi Palguna,

The SNR of targets can be derived from the radar equation by plugging in device specifications and parameters (This can be obtained from the data sheet of the AWR2243 devices). Different targets have different RCS which are again experimentally derived.

With cascading, there will be an improvement in SNR by 10*log10(Number of cascaded devices) (When operating in TDMA mode)

Regards,

Kaushik

0 Palguna Kumar Reddy Gopireddy 1 month ago in reply to Kaushik Gowda

Prodigy 131 points

Thanks.

Can we use the SNR value in CFAR parameters section to calculate the K0. I mean is this the SNR used by postProc to calculate desired K0 value or some other.

(We can close this thread after this clarification.)

0 Kaushik Gowda 1 month ago in reply to Palguna Kumar Reddy Gopireddy

TI__Mastermind 20505 points

Hi Palguna,

Would it be possible to share a screenshot of the exact CFAR parameters you are referring to avoid any confusion?

Regards,

Kaushik

0 Palguna Kumar Reddy Gopireddy 1 month ago in reply to Kaushik Gowda

Prodigy 131 points

sorry for the late reply.

It is mentioned as CFAR threshold, not SNR on miscellaneous options window.

I checked that when I changed this CFAR threshold, the objects with SNR less the CFAR threshold are disappearing from the Detection and Angle estimation results plots. Is this CFAR threshold is the value used for finding K0.

0 Kaushik Gowda 18 days ago in reply to Palguna Kumar Reddy Gopireddy

TI__Mastermind 20505 points

Hi Palguna,

Apologize for the late response, I was out on vacation.

Yes, your understanding is right. You can alter this particular CFAR threshold to pick objects selectively based on the SNR.

Regards,

Kaushik

Sensors

Sensors forum

MMWAVE-STUDIO: How CFAR and is performed in AWR1843 or postProC of mmWave studio?