Hi Bruce,
I was looking through my notes, and about the time of your postings last year, I had run a test on a T3500 running Windows 7 for hours and hours with no errors. The test used a hardware serial loopback, and compared the incoming serial data with what was sent. The quantity and content of the data always matched.
But, I couldn't get permission to re-image the machine with Windows XP. Evidently, I didn't ever post my results to the forum -- sorry about that.
A couple of thoughts since this is still an issue:
* I can see if I can get a hold of a similar system and retry with 1028 bytes instead of 1024 bytes. Can you send your camera file and any code snippets (are you using LabVIEW or C? which version?) that send the serial config data?
* Another thought... can you change your application to send the data in smaller chunks (like 128 or 512 bytes), with 1 ms pauses in between (e.g. send start bit + 511 bytes of data; pause 1 millisecond; Send 512 bytes of data; pause 1 millisecond; Send 1 byte of data + checksum + stop bit). The fact that it doesn't work between Dell computer models suggests a low-level PCIe timing dependency... I know that you're sending large amounts of configuration data, so the pauses aren't optimal, but it might help us figure out the root of the problem. Does the behavior change if you increase the pause to something ridiculous like 100 ms? Does the behavior change if you use really small chunks (like 8 bytes)?
* Just to confirm... you said in your original forum post that the same thing worked with a PCIe-1429 in the T3500 computer? There are a few under-the-hood architectural changes since the 1429 that could affect serial transmission, but the fact that it only fails on the T3500 doesn't suggest a framegrabber-level problem.
-Daniel
