I’m using the Matrox 8.x drivers under Windows 10 LTSC 2019. They work well, but compatibility depends heavily on the Windows build, motherboard, and BIOS. Some Windows updates can break the drivers, so I keep the system stable and avoid updates. If the MXO2 is not detected, it’s often a bios or Windows 10 version rather than the driver itself.
The driver has to be installed in a fresh install windows 10 and has to be the first think that you install.

LOL, that's very much par for the course for a Matrox product. I still have a Digisuite that's hilariously finicky about system specs. I have the version 8 tools, but didn't know that they were compatible-ish with Windows 10. Thanks.

Follow-up on the Matrox MXO2 capture and the comparation on the Matrox MXO2, Aja Kona Lhi and the ATI AIW x1900 capture cards.

Unfortunately, I have not been able to get VirtualDub to work properly with the MXO2 or V-Mix, without quality problems. The only reliable workflow I found was capturing with Matrox’s MTXTools utility at 720x486 uncompressed, then re-encoding the files to HuffYUV. The results are actually quite good.

For this new MXO2 capture, I did the analysis by examining the first and second fields separately, inspecting the UToY and VToY views to evaluate the chroma channels.

Following FeedBucket’s suggestions, I performed additional analysis on all captures:

* UToY and VToY comparisons (copying the U and V channels to luma for chroma inspection)
* Classic histogram
* Levels histogram
* Luma histogram
* AviSynth’s “luma” histogram mode, which greatly amplifies small luminance differences and makes noise, blocking, and subtle artifacts much easier to see
* Bob deinterlacing comparisons
* First and second field inspection

After repeating the comparisons across all capture devices, my conclusions changed somewhat.

The MXO2 performed better than I originally thought. The AJA captures appear slightly softer, especially when viewed with the amplified luma analysis. The ATI AIW Theater 200 still seems to produce the deepest blacks, but overall I believe all of the tested capture devices preserved the image detail surprisingly well.

One observation on the Matrox capture: the Y (luma) not much and U V (chroma) histograms contain a few very thin isolated spikes separated from the main distribution. However, I could not identify any corresponding artifacts when inspecting the UToY and VToY views, so these spikes do not appear to have a noticeable impact on the image.

The luma histogram mode was particularly useful because it clearly highlighted the differences I was seeing in the AJA captures.

Thanks again to FeedBucket for the suggestions and for helping me look for flaws and differences between the capture cards.

Matrox MX02 :

Comparing again all captures Test 5