davidrg wrote on 2026-01-24, 20:44:

JammyPajammies wrote on 2025-12-30, 21:47:

I've skimmed through and have seen mentions of VCS but I'm not entirely sure what VCS is/if it is relevant for my use case/if it is compatible with VisionAV and can't find a readily available download link right now (the github repos seem to be gone), so I am tabling that for the time being.

VCS should work for your 'source preview' type use case - this is what I primarily use it for too with a VisionAV-HD and its much nicer than the program that comes with the Datapath drivers. As you've found it is effectively unmaintained at the moment, but I believe the latest source and binaries for Windows can be found here: https://github.com/arnoo-sel/vcs - this is what I use on Windows 11, and it works fine. There is source for a slightly newer version here but its linux only.

Thank you I did try vcs, however when I switched to Input-2 'S-Video' for this card, it reported appropriate video signal detected butlocked up both the application and my operating system and I had to pull the power and restart from a hard boot.

I have found that OBS 32.0.4 seems to work with it quite well.

The video capture and audio capture are quite sharp (good input signal makes for a good capture ect..) but the software maps to the expected Video Capture Device and Audio Mixer - Video Capture Device, which is very rare. It seems to be a captrue card "built-for" or properly supported for windows 8.1 and Windows 10 - (even) rarer

The hard parts with this card seems to be;

1. Realizing the card has to be used under Windows 11, 10 or at least Windows 8.1 - or a compatible version of Linux
2. Finding the version of the Device Drivers that has the model 199 firmware included with the device drivers

For me this came from [ DriverInstall_v701.exe ] nothing earlier has the firmware - payload included with the DGC133_WIN10 device driver.

For me using the latest [ DriverInstall_V722.exe ] introduced a driver that was less stable, A.I. advice was that DataPath may have had to make some development choices that favored later cards that adversily affected older cards like this one, or they did not test as thoroughly, or cared by that time. - Which is odd since the internals filesd indicate the driver files were created in 2025, so maybe an automated build process is involved at Datapath?

I have had no contact with Datapath, and merely 'guessed' my way into this situation. The 3D printed backplane rails were a big help.

Again I am only guessing, but the s-video input looks to be going through an RC network to create a composite signal for one of the chips, so that is sad, but its still a great card for the HD-RGB, DVI-I, HDMI and SDI inputs, and the S-Video input is very handy.

The Jumper "must be removed" thing is very odd, but for me, very real.. there does not appear to be any Datapath documentation for why it needs to be removed.

There is Datapath documentation for all the other jumpers in user manuals for other cards, mostly its to individually address multiple cards in one computer.

Kordanor wrote on 2025-12-30, 21:50:

So...I don't quite see how 320x200 can have analogue noise, but 640x480 is perfect.
Is that some inherent issue of double scans?

What do you guys do to get a clean recording of 320x200?

If you are seeing 320x200 come out odd, but 640x480 is great.. then the only issue I can think is noise caused by slightly mistimed sampling (i.e. the number of clocks per scanline, and front/back porch timings).

The 640x480 video signal is very standard, since it was standardized by VESA. But 320x200 is not, so maybe that is being sampled slightly off. Not 100% sure, this is a bit of a guess. I can't think of a reason why 320x200 would have more noise than 640x480 otherwise.

The other guess is that if the two modes are sourced by a different oscillator, maybe there is a bit too much jitter in one or other clock.

One idea that comes to mind is, since you are using OBS - maybe it would be possible to develop a denoise filter into OBS, that would take into account the pixel upscaling. So if 320x200 is being upscaled to 1600x1200 (i.e. 5x6 integer upscaling), then maybe an OBS filter could be used to average all those 5x6 pixels before output. And also quantize back to VGA 6 bits per color channel, rather than 8. That should stabilize noise down by quite a bit.

I don't think it's the timings. I set the sampling to the optimal via VCS for the datapath manually, and used the values Phil showed in his video for the retrotink4k. And both results in the same issue. And as mentioned it happens with two different graphics cards. So it's not like one card is a bit off (which I think happens on some cards/resolutions).

The idea with OBS also came to my mind. My request can be found here: https://obsproject.com/forum/threads/filter-r … remover.186294/
And I also posted about it on the OBS Discord. But unfortunately there is no resonance (not even from someone asking for the same).
I am not experienced with programming OBS Filters. Maybe I could Chat-GPT Slop something together. But that sounds ... painful.