AWR1843AOP: Peakval variable

Hi Shoko,

As mentioned in the E2E that you shared, peakVal can be computed from the sideInfo. 

peakVal(dB) = DPIF_PointCloudSideInfo.snr/10 + DPIF_PointCloudSideInfo.noise/10

Regards,

Samhitha

Hello Samhitha, 

Thank you for quick response. 

> peakVal(dB) = DPIF_PointCloudSideInfo.snr/10 + DPIF_PointCloudSideInfo.noise/10

Which file is the above formula in? 

It is not in "C:\ti\mmwave_sdk_03_06_02_00-LTS\packages\ti\datapath\dpif\dpif_pointcloud.h". 

Please advise. 

Thanks, 

Shoko 

Hi Shoko,

This formula is not present in the SDK. You can use the DPIF_PointCloudSideInfo to calculate the peakVal. 

Shoko Sakaguchi said:

It is not in "C:\ti\mmwave_sdk_03_06_02_00-LTS\packages\ti\datapath\dpif\dpif_pointcloud.h". 

You can find the DPIF_PointCloudSideInfo structure at this path.

Regards,

Samhitha

Hello Samhitha, 

Thanks for your reply. I understand the formula is not present in the SDK. 

I have additional questions about DPIF_PointCloudSideInfo.snr, noise, peakval. 

1) I'm using AWR1843AOP with the version mmwave_sdk_03_06_02_00-LTS and I made AWR1843AOP detect the target. 

Below is the pointcloud data from AWR1843AOP.  

 SNR: 260, Noise: 482, x: 0.15753, Y: 4.97663, Z: -0.7877, PeakVal: 3172 

You said, 

> peakVal(dB) = DPIF_PointCloudSideInfo.snr/10 + DPIF_PointCloudSideInfo.noise/10

Thus, 

peakval(dB) = 260/10 + 482/10 = 74.2(dB) 

Is it correct? 

2) The PeakVal in the pointcloud data is "3172". It isn't dB unit. 

Please let me know the relation between "3172" and "dB unit of peakval". 

3) "dB unit of SNR" is 260/10 = 26dB, and "dB unit of Noise" is 482/10 = 48.2dB. 

Are they correct? 

If yes,

Is "dB unit of SNR" the level from the noise around target to the peak of target?

What level does "dB unit of Noise(48.2dB)" refer to? Please explain using range FFT waveform. 

Please advise. 

Thank you for your help. 

Shoko 

Hi Shoko,

Which application are you running? Is it the demo from the SDK? There is no member peakVal in the DPIF_PointCloudCartesian structure.

Samhitha Bobba said:

peakVal(dB) = DPIF_PointCloudSideInfo.snr/10 + DPIF_PointCloudSideInfo.noise/10

I found few instances where this formula is used. Here is the path to one of the instances, mmwave_sdk_03_06_02_00-LTS\packages\ti\datapath\dpc\objectdetection\objdethwa\test\main.c. Check ObjectDetection_Test_convertCoordinates function in the above mentioned file.

Regards,

Samhitha

Hello Samhitha, 

Thanks for your information. I'm running the demo from the SDK. 

 OK, I understand peakVal is not in DPIF_PointCloudCartesian structure, and it is in ObjectDetection_Test_detectedObj_t. 

mmwave_sdk_03_06_02_00-LTS/packages/ti/datapath/dpc/objectdetection/objdethwa/test/docs/doxygen/html/struct_object_detection___test__detected_obj__t.html

Also, I found "peakval" in the documentation of AoA using 2D FFT method. 

mmwave_sdk_03_06_02_00-LTS/packages/ti/datapath/dpc/dpu/aoa2dproc/docs/doxygen/html/index.html

By the way, could you answer my following questions? 

1) I'm using AWR1843AOP with the version mmwave_sdk_03_06_02_00-LTS and I made AWR1843AOP detect the target.

Below is the data extracted from AWR1843AOP. PeakVal is the data from CFAR detection list of above image. 

SNR: 260, Noise: 482, x: 0.15753, Y: 4.97663, Z: -0.7877, PeakVal: 3172

You said,

> peakVal(dB) = DPIF_PointCloudSideInfo.snr/10 + DPIF_PointCloudSideInfo.noise/10

Thus,

peakval(dB) = 260/10 + 482/10 = 74.2(dB)

Is it correct?

2) The PeakVal in the pointcloud data is "3172". It isn't dB unit.

Please let me know the relation between "3172" and "dB unit of peakval".

3) "dB unit of SNR" is 260/10 = 26dB, and "dB unit of Noise" is 482/10 = 48.2dB.

Are they correct?

If yes,

Is "dB unit of SNR" the level from the noise around target to the peak of target?

What level does "dB unit of Noise(48.2dB)" refer to? Please explain using range FFT waveform.

Please advise. 

If you don't know, please change the person in charge about these questions? 

Regards, 

Shoko 

Hi Shoko,

Thank you for being patient. I work on the processing chain and the demo, so I can help you with your questions. So, can you please help me understanding your questions.

Shoko Sakaguchi said:

peakval(dB) = 260/10 + 482/10 = 74.2(dB)

Is it correct?

This is correct.

Shoko Sakaguchi said:

The PeakVal in the pointcloud data is "3172". It isn't dB unit.

Yes, this value is not in dB. I need to check internally about the peak field in DPU_AoAProcHWA_CfarDetOutput. Can you clarify about the exact location in the code from which you are deriving this peak value?

Shoko Sakaguchi said:

"dB unit of SNR" is 260/10 = 26dB, and "dB unit of Noise" is 482/10 = 48.2dB.

Are they correct?

Yes, you are correct.

Shoko Sakaguchi said:

Is "dB unit of SNR" the level from the noise around target to the peak of target?

Yes, it compares the level of the signal of target to the noise around the target.

Shoko Sakaguchi said:

What level does "dB unit of Noise(48.2dB)" refer to? Please explain using range FFT waveform.

Noise refers to surrounding average noise value. Refer to CFAR engine section of Radar Hardware Accelerator User's Guide (Rev. B) (ti.com) to see how the noise is computed.

Regards,

Samhitha

Hi Shoko,

To convert peak value to dB scale, you can use CFARCADSP_CONV_PEAK_TO_LOG macro defined in mmwave_sdk_03_06_02_00-LTS\packages\ti\datapath\dpc\dpu\cfarcaproc\cfarcaproccommon.h. 

Regards,

Samhitha

Hello Samhitha, 

Thank you for your response. 

> I need to check internally about the peak field in DPU_AoAProcHWA_CfarDetOutput.

 Can you clarify about the exact location in the code from which you are deriving this peak value? 

-> Since I'm not coder, I don't know the exact location. Could you please ask the person who wrote the documentation of AoA using 2D FFT method? 

Regarding "CFARCADSP_CONV_PEAK_TO_LOG macro", I have some questions. 

1) Please let me know what is Qformat? 

2) "log2(| . |)" "log10(| . |)"

 I don't know the above the above "| . |"... Please let me know with some example. 

Thanks, 

Shoko 

Hello Samhitha, 

I converted peak value to dB scale using your image of CFARCADSP_CONV_PEAK_TO_LOG macro. 

Below is the pointcloud data from AWR1843AOP. PeakVal is the data from CFAR detection list. 

 SNR: 260, Noise: 482, x: 0.15753, Y: 4.97663, Z: -0.7877, PeakVal: 3172

 Output = val * 6 / 2^Qformat * 10,    If Qformat is 8, 

 Output = 3172 * 6 / 256 * 10 = 743.4375 

On the other hand, 

 peakVal(dB) = DPIF_PointCloudSideInfo.snr/10 + DPIF_PointCloudSideInfo.noise/10

 peakval(dB) = 260/10 + 482/10 = 74.2(dB)

Why are avobe two results different? 

Please advise. 

Thanks, 

Shoko