AWR1843: USRR coordinate Inaacuracy

Hi Ayush,

Needed some more info here:

• Have you made any changes to the demo?
• Have you performed calibration on the device?
• Are you sing the TI EVM for 1843 or your custom board?

Regards,

Kaushik

Hello Kaushik,

There's no changes made in the demo.Currently running the default binary.

In terms of calibration we don't see the deviation in MRR.We also ran OOB demo and observed no deviation.Only in terms of USRR demo is giving us this issue.

We are using a custom board AWR1843 ES2.0 while testing this.

Warm regards

Ayush Shetty

Hi Ayush,

Can you use a TI EVM and see if the observation is the same in your test setup? This would make it easier to debug on my end if needed.

Regards,

Kaushik

Hello Kaushik,

I have tried the similar test with TI1843 boost board.I observe the same deviation.

No other lab is giving such issues.I checked with Out of box and MRR as well.

Warm regards

Ayush Shetty  

Hi Ayush,

Can you provide a brief description of your test so that I can reproduce similar conditions on my end?

Additionally, can you update to the latest version of this demo from the Radar toolbox which can be found here.

Regards,

Kaushik

Hey Kaushik,

Sure so you enable MRR point cloud and tracker in the GUI and at radar's boresight(X=0) walk a straight path. You should observe the path walked is straight.

Then enable disable everything in GUI and enable USRR point clouds and cluster and walk the same straight path, You should observe a path with huge deviation.

Till then I'll carry out the same demo via radar toolbox as you suggested.

Warm regards

Ayush Shetty

Hi Ayush,

I was talking more along the lines of:

• Did you use a corner reflector?
• Were there any other strong reflecting surfaces in your vicinity or the sensor's line of sight?
• Was this done in a dark or anechoic chamber?
• How far away were the walls in the room or was this performed outdoor?

Regards,

Kaushik

Hello Kaushik

I used a person as well as a Corner reflector in this sceanrio.Both are giving me the same deviation

It was carried out on a terrace where there are no walls surrounding the setup.

Hi Ayush,

Thanks for the info. Let me work on this and get back to you by end of next week.

Regards,

Kaushik

Hi Kaushik,

Any update on your findings..?

Regards,

vikrant.

Hi Kaushik,

Please check the link below:

e2e.ti.com/.../awr1843-where-are-the-calibration-parameters-in-automotive-lab-demo-mrr

This issue has been discussed but not clear whether issue is resolved or not.

Can you please check this and help me out.

regards,

vikrant

Hi Vikrant,

Appreciate your patience. We are currently working on the same. Kindly allow us to get back to you once we have more information here.

In the thread you have mentioned, only info about calibration procedure was discussed. Regarding the same, it is always recommended to perform calibration so as to avoid deviation in the detections.

Regards,

Kaushik

Hello Kaushik,

Please note:

The board we are using is AWR1843BOOST EVM

Observation with the board is same as explained in above chats by my team mate Ayush.

Points to consider:

1. The procedure mentioned in the e2e link I mentioned is not clear.(need more clarity on what code to flash on the board to perform calibration. Is it OOB or what ??)

2. It has a matlab script to use it is asking "signal processing toolbox" and for which need matlab license (trial version is not suffice).  How to pass raw bin file to script. I assume this is not fill script and only a function from the full script is shared. 

3.There  is difference between procedure mentioned in SDK and e2e forum(may be MRR demo has got slightly different requirement).

4. We tried with:

• MRR+USRR,
• MRR only and
• USRR only

But we are getting issue with only USRR target detected data accuracy but MRR target detection data in all three cases is fine.

5. The above observation and e2e forum discussion is creating multiple impression's :

• That the Radar configuration is the issue since MRR is working and not USRR.
• Or there is some firmware issue on the USRR section.
• Or the radar calibration is done/applied for MRR only and not for USRR.

6. It is said in the e2e forum that I am quoting below:

" In the mmwave SDK demo, the data collection for this calibration was implemented in the target code.
In the MRR demo, unfortunately, the data collection for calibration is not implemented in the target code.

This is why we used the dca1000 to capture the raw data and compute the calibration coefficients.

The calibration coefficients are hard coded here:

C:\data-nobackup\2022\ToolboxAutomotive\mmwave_automotive_toolbox_3_2_0__win\mmwave_automotive_toolbox_3_2_0\labs\lab0007_medium_range_radar\src\dss\dss_main.c"

https://e2e.ti.com/support/sensors-group/sensors/f/sensors-forum/1120155/awr1843-where-are-the-calibration-parameters-in-automotive-lab-demo-mrr

From the above quote It appears that the Radar Calibration is required .(but why MRR is working and not USRR).

What we are looking for:

1. A procedure without dependency on matlab license.
2. If any change to be done on the firmware side please provide it's details. Or if any patch that we can use share along with its details on how to use it.

Kindly reply with your findings/solution at earliest.

Thanks and regards,

Vikrant