• State TI Thinks Resolved
  • Locked Locked
  • Replies 5 replies
  • Answers 2 answers
  • Subscribers 32 subscribers
  • Views 1105 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.

OPT9221: Motion Artifact

Michael Kelly
Prodigy 30 points
Part Number: OPT9221
Other Parts Discussed in Thread: OPT8241,

We are observing artifacts when imaging moving objects and require some assistance in understanding how to eliminate these. Using the Voxel viewer and the TI OPT8241 eval kit with default parameters and an unambiguous range of 3 m, depth maps appear to be computed at 4 different exposure times (see attached images below of a fan blade). There is a parameter mentioned in the data sheet (lumped_dead_time) which potentially could be used to eliminate the distribution of exposure gaps; however this parameter does not appear to be supported in the most recent parameter set that is exposed through Voxel viewer. Questions:

1. Is the parameter lumped_dead_time supported for the OPT9221, and if so, could it be used to eliminate the issue we are observing?

2. Are there other parameters or approaches that you would recommend so that we can get depth measurements estimated over a single time interval rather than 4 as we are observing?

  • 0
    TI__Genius 15005 points
    Michael,

    The OPT8241 sensor uses several exposures to construct one frame of data - the minimum is 4 exposures (4 quads with 1 subframe). (With the default parameters, we have either 4 quads and 4 subframes, or 6 quads and 2 subframes for the calibrated profiles).
    It is not possible to use a single exposure with this sensor, and this does cause motion artifacts as with the fan blade image.

    Lumped dead time can indeed be used to minimize the gap between quads. However this is deliberately disabled on the OPT8241-CDK-EVM for compliance with laser safety.
    This restriction is enforced in the VoxelSDK and also in the EVM firmware. The sensor and chipset do support lumped dead time and can be used without this restriction if you build your own derivative designs.

    Best Regards,
    Anand
  • 0 in reply to Anand Gadiyar
    Prodigy 30 points
    Hi Anand.

    Thanks for the detailed response. This all makes sense. We would like to reduce the gaps between quads as much as possible and were wondering if this could be accomplished by reducing the readout time. The binning and ROI options have been tested, but neither appears to significantly reduce the gaps. Is there another possible approach?
  • 0 in reply to Michael Kelly
    TI__Genius 15005 points
    Hi Michael,

    Please refer to section 7.3.3.1 of the OPT9221 datasheet for an explanation of the frame structure. In this, the frame rate and number of quads and sub-frames determines the total quad time. The reset time per quad is fixed for the sensor, and the readout time per quad depends on the ROI. The integration time per quad is programmable. Whatever remains is the dead time per quad.

    The only way to change the dead time per quad without using lumped dead time is to increase the frame rate.

    Best Regards,
    Anand
  • 0 in reply to Anand Gadiyar
    Prodigy 130 points

    Hi Anand,

    As Michael has observed. the fact that successive quad exposures are separated by a delay causes significant depth error when imaging moving objects.

    Does binning have any impact on the delay between successive quad exposures?  If so, can binning be used to improve depth accuracy on fast moving objects?

    Best regards,


    Dave

  • 0 in reply to Dave McMordie
    TI__Intellectual 2030 points

    Hi Dave,

    Binning does not change the timing between quads. Even ROI will not change the timing unless you use lumped dead time mode. But I am sorry to say that the use of lumped dead time mode is disabled on the CDK due to laser concerns. 

    Regards