• State Resolved
  • Locked Locked
  • Replies 2 replies
  • Subscribers 98 subscribers
  • Views 315 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.

Column swapping on a DM648

Jason Zutty
Intellectual 385 points

Hi -

I am working with a DM648 and the provided video_preview example.  My problem is that whenever I make an attempt to "move" parts of the image around, such as stretching and shrinking, shifting over parts, or most recently using the image library sobel function, I get thin columns swapping places all over the image.  Each column is about 4 pixels and seems to be swapped with its neighbor.

What might cause something like this to occur?  The image is still recognizeable, however the artifacts are very visible.

 

Thanks for your help,

Jason

  • 0
    TI__Mastermind 41665 points

    I think this may be a result of a silicon errata, in particular advisory 1.0.1 from the 1.0 silicon revision as discussed in the errata, what silicon revision is your particular DM648?

    The errata itself relates to the endianess of the video ports relative to the device and manifests itself by putting out every 8 bytes in reverse which would appear in the output to be horizontal columns of reversed video, this should be a bit wider than 4 pixels (8 data points actually though it may not appear that way in a monitor). I haven't done much work with DM648 but I suspect if you are using 1.0 silicon that this is the issue, as if you were just passing the data from input to output the endianess should not matter (both video in and video out match endianess) but as soon as you started processing the video it would not necessarily look right anymore depending on what your algorithm is doing.

  • 0 in reply to Bernie Thompson TI
    Intellectual 385 points

    Good call, I am using the 1.0 revision.  After reversing each set of eight bytes it now works perfectly!