• State
  • Locked Locked
  • Replies 3 replies
  • Subscribers 30 subscribers
  • Views 453 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

AWR2944: The point cloud of the output is discontinuous

Cherry Zhou
Mastermind 22235 points
Part Number: AWR2944

Hi team,

Here's an issue from the customer may need your help:

During the road test, the point cloud data output from AWR2944 was passed through matlab to make a range-time image, resulting in a point cloud discontinuity issue. The R-t plot looks like this: 

Discontinuities in close proximity to the target are particularly evident. Theoretically, the signal-to-noise ratio of the echo signal should be larger when the target is close.

The test scenario is a 4 lane wide bi-directional lane. The point cloud of the target can become discontinuous for a period of time when vehicles away are waiting for the traffic light to cross.

The scene of the intersection is as follows:

During the test, it was found that the number of point clouds detected per frame was between 35 and 50 points, far less than the limit of 1800 point clouds.

The code used is the code of the SDK and uses DDMA mode with a set frame period of 200 ms. During the local-maximum detection phase, in addition to comparing the peaks detected by the cfar with adjacent units, the comparison to the angle channel and Doppler channel thresholds currently in rangeBin is enabled.

What could be the possible cause of this issue? Could you help check this case? Thanks.

Best Regards,

Cherry

  • 0
    TI__Mastermind 20605 points

    Hi Cherry,

    I would like some additional info here:

    1. Can you mention the units of your range? What does 1 unit of your y-axis map to?
    2. Can you confirm whether there was any other obstruction in the fov of the radar that could've masked or damped the reflections from the object whose detections turned out to be discontinuous? It's very difficult to just say by looking at a still image.
    3. Have you analyzed the range profile of the frames where the detections were dropped? Please confirm whether the peaks persisted in those frames.
    4. Has the customer debugged the point at which these detections were dropped in the signal chain?
      • Cherry Zhou said:
        During the local-maximum detection phase, in addition to comparing the peaks detected by the cfar with adjacent units, the comparison to the angle channel and Doppler channel thresholds currently in rangeBin is enabled.

      This could potentially do something like this. Has the customer tried tuning different thresholds using which the detections are filtered?

    Regards,

    Kaushik

  • 0 in reply to Kaushik Gowda
    TI__Mastermind 22235 points

    Hi Kaushik,

    Thanks for your support.

    Kaushik Gowda said:
    Can you mention the units of your range? What does 1 unit of your y-axis map to?

    The y-axis is in meters and represents the radial distance of the target distance from the radar; the horizontal axis represents the sequence number of the frame with a frame period of 200 ms. 

    Kaushik Gowda said:
    Can you confirm whether there was any other obstruction in the fov of the radar that could've masked or damped the reflections from the object whose detections turned out to be discontinuous? It's very difficult to just say by looking at a still image.

    The test scenario was a pedestrian bridge, the radar was fixed using a stand, the front of the radar was clear, and the camera was positioned essentially the same as the radar was. And I'm trying to get the testing video.

    Kaushik Gowda said:
    Have you analyzed the range profile of the frames where the detections were dropped? Please confirm whether the peaks persisted in those frames.

    The loss of target to the target within a range of 150 m to 230 m is more severe in the range-time image, while the target to be reached has a loss of target at all distances, as shown by the loss of target when the target is stationary.

    Is there any peaking in the cfar in these frames that need to be re-acquired for data verification?

    Kaushik Gowda said:
    Has the customer debugged the point at which these detections were dropped in the signal chain?

    Did you mean that DPC is affected by the radar power-on time? Target loss occurs at each point in time and should not be relevant. 

    Additionally, the customer tried to adjust the threshold of OS_cfar and found that 256 points were detected for each rangeBin's Cmar as defined in SDK. There is a significant risk that the memory will be overwritten and the points will be lost. This was also verified during actual testing.

    Could you help provide a set of cfar parameter configurations for external field testing? Including oskvalue, winlen, threshold, and so on.

    Thanks and regards,

    Cherry

  • 0 in reply to Cherry Zhou
    TI__Mastermind 22235 points

    Hi,

    May I know is there any feedback?

    Thanks and regards,

    Cherry