IWR6843AOP: IWR6843 3D People Tracking: Target height increases with range

Part Number: IWR6843AOP
Other Parts Discussed in Thread: IWR6843

Hi TI team,

I am using the IWR6843 3D People Tracking demo in a wall-mounted configuration and observing a systematic issue where the estimated target height (Z coordinate) increases as the target moves farther away, even though the target’s actual height is constant.

Additional observation:

  • Targets with height above ~1 m are estimated reasonably well

  • Lower-height targets (e.g. below ~1 m) show much larger distortion, especially as range increases

Key characteristics of the issue:

  • Height error grows roughly linearly with range (e.g. reasonable at ~2 m, significantly taller at 6–8 m)

  • The effect is stable and repeatable (not random noise)

  • I tried tweaking the sensor tilt angle in the configuration, but this did not resolve the distortion

Setup details:

  • Wall-mounted sensor, height ~1.6 m, tilt ~20°

  • 3D People Tracking demo with dynamic/static CFAR enabled

  • compRangeBiasAndRxChanPhase currently disabled (default values)

 

Questions:

  1. Is this range-dependent height distortion—especially affecting low-height targets—a known consequence of missing RX channel phase calibration, particularly impacting elevation AoA estimation?

  2. Can TI confirm whether enabling compRangeBiasAndRxChanPhase with values obtained from the xWR68xx OOB calibration is expected to correct this elevation bias?

  3. Are there additional recommended practices (e.g. elevation bias correction, tracker-level height constraints) for stabilizing height estimates in wall-mounted IWR6843 setups?

Any guidance on best practices to mitigate elevation bias in 3D people tracking would be greatly appreciated.

Thanks in advance.

  • Hi,

    We have received your post and the investigation will take some time. Thank you for your patience.

  • Apologies for the delay here.

    1. Is this range-dependent height distortion—especially affecting low-height targets—a known consequence of missing RX channel phase calibration, particularly impacting elevation AoA estimation?

      I think this is more likely on the post-processing end rather than the RF end.

    2. Can TI confirm whether enabling compRangeBiasAndRxChanPhase with values obtained from the xWR68xx OOB calibration is expected to correct this elevation bias?

      compRangeBiasAndRxChanPhase when properly configured will improve the device's angle estimation, which includes angle in azimuth and elevation. Note that this improvement gain is more prominent in non-uniform antenna geometries like the ISK rather than square ones like the AOP.

    3. Are there additional recommended practices (e.g. elevation bias correction, tracker-level height constraints) for stabilizing height estimates in wall-mounted IWR6843 setups?

      I would look at dynamic2DAngleCfg and dynamicRangeAngleCfg closely and adjust parameters there to help with angle estimation. Note that there is processing time and memory size constraints, so increasing performance here after a certain point may require tradeoffs in other areas.

      https://dev.ti.com/tirex/explore/node?isTheia=false&node=A__AcT8axbWVHGuwuh7Yx5.NA__radar_toolbox__1AslXXD__LATEST