IWR6843AOP: Wrong elevation estimation

Shlomi Shvartzman

Part Number: IWR6843AOP

Tool/software:

Hi,

I'm using the out-of-box demo from mmwave_sdk_03_06_02_00-LTS, I have loaded the bin file with "_aop" suffix.

I recorded a corner at 6m with IWR6843AOP and analyzed the point cloud (XYZ data from the firmware). However, I observed some anomalies:

Y-values and range (sqrt(x² + y² + z²)) are not correlated. The range appears correct, but the Y-values are off.
Z-values get high positive and high negative values, which is incorrect. For this corner, Z should be close to zero, but I’m getting ±4m.
X-values seem reasonable.

Interestingly, I recorded the same corner at 6m, with the same radar, using the 3D People Tracking Firmware and obtained accurate results across all axes.

I suspect the AoA algorithm in the out-of-box firmware might not be calculating elevation correctly.

Could you please advise?

Thanks,

Shlomi

9 months ago

0 Nathan Block 9 months ago

TI__Mastermind 37805 points

Hi Shlomi,

What configuration files did you use?

Best,

Nate

0 Shlomi Shvartzman 9 months ago in reply to Nathan Block

Prodigy 180 points

Hi Nate,

Here is the configuration we used:

0 Nathan Block 9 months ago in reply to Shlomi Shvartzman

TI__Mastermind 37805 points

Hi Shlomi,

Can you build the AOP image using the projectspec file in

C:\ti\radar_toolbox_2_20_00_05\source\ti\examples\Out_Of_Box_Demo\src\xwr6843AOP

Or just use the image in

C:\ti\radar_toolbox_2_20_00_05\source\ti\examples\Out_Of_Box_Demo\prebuilt_binaries

I think the antenna geometry needs to be used, which the SDK demo doesn't support.

Bes,

Nate

0 Shlomi Shvartzman 9 months ago in reply to Nathan Block

Prodigy 180 points

Hi Nate,

We took the pre built binary from the radar toolbox, but we also tried to built it.

Thanks,

Shlomi

0 Nathan Block 9 months ago in reply to Shlomi Shvartzman

TI__Mastermind 37805 points

Hi Shlomi,

Thank you for clarifying. Can you try one of the demos that are derived from the OOB demo, like small obstacle detection on the IWR6843AOP? I think there was an issue with the OOB for 6843AOP.

https://e2e.ti.com/support/sensors-group/sensors/f/sensors-forum/1166918/iwr6843aopevm-recommended-binary-and-notable-differences-from-out-of-box-demo

Best,

Nate

0 Shlomi Shvartzman 9 months ago in reply to Nathan Block

Prodigy 180 points

Hi Nate,

I'll check the other demos and provide an update.

In the meantime, could you elaborate on the issue with the OOB for 6843AOP? What exactly was the problem? Has anyone looked into the root cause?

Thanks,
Shlomi

0 Josh Dye 9 months ago in reply to Shlomi Shvartzman

TI__Mastermind 21621 points

Hi Shlomi,

I just wanted to let you know that Nathan is currently out of the office. He will return and continue the discussion on this thread next week. Thank you for your patience.

Best regards,

Josh

+1 Nathan Block 9 months ago in reply to Josh Dye

TI__Mastermind 37805 points

Hi,

I was mistaken, please use an AOP configuration like those labeled AOP in this folder:

C:\ti\mmwave_sdk_03_06_00_00-LTS\packages\ti\demo\xwr64xx\mmw\profiles

Your angle estimation requires the correct compRangeBiasAndRxChanPhase line

compRangeBiasAndRxChanPhase 0.0 1 0 -1 0 1 0 -1 0 1 0 -1 0 1 0 -1 0 1 0 -1 0 1 0 -1 0

Best,

Nate

0 Shlomi Shvartzman 9 months ago in reply to Nathan Block

Prodigy 180 points

Hi Nate,

Thanks for this important clarification. I'll review it and keep you posted.

Regarding 3D people tracking, the values you mentioned are set in antPhaseRot, while the calibration values are stored in compRangeBiasAndRxChanPhase. Assuming your suggestion works, I’m curious about the specific way the calibration process is applied for AOP. Here are two possible approaches—let me know if either one is correct.

For simplicity, let’s define phaseRotAOP as a vector with the correct phases:
0.0 1 0 -1 0 1 0 -1 0 1 0 -1 0 1 0 -1 0 1 0 -1 0 1 0 -1
Similarly, we define phaseRotISK as abs(phaseRotAOP).

Option 1: Before starting the calibration, set compRangeBiasAndRxChanPhase to phaseRotAOP. Then, compute the calibration values and use them going forward in compRangeBiasAndRxChanPhase.

Option 2: Initialize compRangeBiasAndRxChanPhase with only 0s and 1s. Perform the calibration process to obtain the calibration values, then multiply them by phaseRotAOP.

Please let me know which approach is correct.

Thanks,
Shlomi

0 Nathan Block 9 months ago in reply to Shlomi Shvartzman

TI__Mastermind 37805 points

Hi Shlomi,

Let's please start by using the out-of-box demo for the AOP, since we have already established that the people counting demo works. The antPhaseRot CLI only applies for the people counting demo.

Please do option 1. When you're doing the calibration, I don't think it should take the values in compRangeBiasAndRxChanPhase into account, but keep them constant just to be safe.

Best,

Nate

Sensors

Sensors forum

IWR6843AOP: Wrong elevation estimation