IWR6843ISK: Performance comparison with IWR1843BOOST

Hi Massimo

I'll need more information in order to assist you. Performance is expected to be the same on both of the boards, can you please share some of your data and pictures/video of your results? How big of a discrepancy are you seeing? Can you confirm that there is no such discrepancy in your measurement scene?

Regards,

AG

Hi Akash,

The measurements setup was quite the same :

 - radar placed on a stable tripod at ~1.5m height, ~2m from nearest lane;

- elevation : ~0°;

- azimuth : slightly tilted to point to cars at ~80m;

- 6843 CFAR thresholds are set to the same as 1843 in order to get the same sensitivity;

- the video is shaky because I'm holding the recording device.

From the following capture, you can see the main issue with the 6843 is the X spread of the targets (more than 20m when distance is >40m ), and often a gap in the center of the cluster :

Here's a capture with the 1843.

Best regards,

Massimo

Hi Massimo

I noticed in the 68xx & 18xx_traffic_monitoring_70m_MIMO_3D.cfg files there is a discrepancy with the trackingCfg with lower velocity thresholds which might be why you see less number of points with the IWR6843ISK board. I would recommend ensuring the parameters in both configs (besides profileCfg) are the same. I noticed there is a discrepancy with the CQRxSatMonitor parameters as well.

Regards,

AG

Hi Akash, thank you for your reply.

From my understanding, trackingCfg  is only used by the tracker algorithm (gtrack) whereas I do not display tracks but only the point cloud.

CQRxSatMonitor differs but seems disabled from analogMonitor 0 0 anyway.

Best regards,

Massimo

Hi Massimo

Your results are much more different than expected, there is virtually no expected performance difference between the two devices. I would recommend re-testing and ensuring conditions for both are exactly the same.

If you would like to continue discussion then I will need more information about your testing setup and procedure, otherwise I cannot comment further.

Regards,

AG

Hi Akash,

I will retest and try to capture the same vehicles with the two radars configured with the same parameters.

Please don't close the thread and allow a few weeks.

Best regards,

Massimo

Thank you Massimo this sounds good

Regards,

-AG

Hi,

I've done new measurements with the two radars detecting at the same time (6843 and 1843).

Here are photos of the setup which is similar to the previous one (elevation, azimuth, config).

The metal pole is at ~90° azimuth (from the boresight) for the two radars and shouldn't interfere.

Here is a selected (X,Y) snapshot showing my issue : it seems that the 6843 (on the left) generates detection points which are more spread, specially on the X axis than the 1843 (on the right). In addition, the density of points at the correct spot seems better for the 1843.

Finally, here is a short sequence :

I understand there shouldn't be any difference by design but I would need to pinpoint the cause of this discrepancy (test setup, signal processing chain, PCB manufacturing, chip RF performances...) .

Tell me if you need specific information on the setup.

Best regards,

Massimo

Hello Massimo,

Sorry for the delay, we were on a holiday break. Can you please confirm whether you loaded the prebuilt binaries onto the devices or if you built the projects and made any changes to the software? Also, please confirm the software file loaded on each device, as they need to be different.

Regards,

Jackson

Hi Jackson,

- I used the prebuilt binaries : traffic_monitoring_18xx_demo.bin and traffic_monitoring_68xx_demo.bin.

I've also plotted the detection points for the two radars for the same traffic with respective default config files in a similar test setup as previously.

The first one is the 1843 which seems OK.

The second one is the 6843 and you can see that azimuth is noisier.

The point density is noticeably lower than with the 1843.

The elevation estimation has a real problem.

And finally, disambiguating the speed seems problematic (there is a lot of aliasing).

I didn't perform the calibration for any of the two radars and understand it can affect angle estimation and also speed disambiguation in noisy environments.

Will the 6843 performances become similar to the 1843 with a proper calibration ?

Would you point out anything else ?

Best regards,

Hello,

Calibration certainly can help the angle estimation, perhaps make it cleaner. For the aliasing, you may need to look at the range over consecutive frames to choose the correct velocity. Please try those and see if it helps. I am not sure why the performance might be different, although really it seems fairly close. I will consult with my team for more suggestions.

Regards,

Jackson

Hi Jackson,

Calibration improved the performance of the 6843.

I have now plots similar to the 1843.

Thanks for the assistance,

Massimo