Machine Vision

cancel
Showing results for 
Search instead for 
Did you mean: 

Problem report w/ solution: Glitch in PCI-1424 signals when displaying live video on Nvidia graphcs with WPF

Hello,

Here is a problem I encountered, and the solution I stumbled on.  I thought I'd share both in case others encounter it too:

On Windows XP, when acquiring and displaying a non-square image from a PCI-1424 via .NET WPF, I saw periodic glitches in the video image
(white vertical lines).  These glitches were simultaneous with large curved deflections in the FRAME_DONE timing signal, as viewed on an external oscilloscope.  The glitch frequency appeared to be a function of frame rate and aspect-ratio.

 

This only occurred when calling ‘WritePixels()’ on a `WriteableBitmap` object, or ‘Invalidate’ on an `InteropBitmap` object, and only when the object was bound to an Image control.  Video playback via GDI was problem free.

 

After many failed troubleshooting attempts, as a last-ditch effort I wiped the hard disk and installed Windows Vista.  That fixed it.

 

 

Hardware:

  • Dell T3400 workstation with NVidia NVS Quadro 290 graphics
  • Hamamatsu Orca ER camera attached to a NI PCI 1424


Software:

  • Problem on Windows XP SP3; Fixed on Windows Vista SP2
  • NVidia graphics driver v191.00 (WHQL)
  • NI Vision Acquisition 2009 software / IMAQ 4.3
  • WPF graphics on .NET 3.5 SP1, using the NI IMAQ API for .NET 3.5

Besides the operating system change, the only software change was in the Nvidia graphics driver (they have a different v191 driver for XP than for Vista)

 

 

I hope this helps someone.  And if anyone has any idea what might have been the underlying cause, I'd be curious to hear it.

 

Best,

Gabriel

Message Edited by Gabriel-r on 11-06-2009 04:02 AM
Message 1 of 4
(4,119 Views)

Hi Gabriel,

 

Is it possible for you to post some sample code for us to use to try to reproduce this behavior?

0 Kudos
Message 2 of 4
(4,097 Views)

Hi Oliva,

 

Attached is a VS2008 solution.  I removed some extraneous stuff and simplified it a bit, but it's essentially the same code as was generating the problem.

 

Hope this helps.

 

-Gabriel

0 Kudos
Message 3 of 4
(4,070 Views)
Thanks Gabriel for the feeback. We'll see if we can reproduce the behavior and find out what is causing it.
0 Kudos
Message 4 of 4
(4,052 Views)