Component placement for the bottom side can be found here: Voodoo 2 component map
As for the copper layers, I should have them, both scans and in vector format. Need to find the files first. Will probably upload them in a couple of days.

Thank you.

Looking at my broken Diamond card, I can pillage FB6, C36, C38, and the bus switch (S3L384Q). I noticed that C38 is tantalum on the Diamond but ceramic on the Atrend. Do you think the swap of these 4 components will fix the problem with the Atrend, or is the noise created from other components which will affect any bus switch/caps brand?

Perhaps shield the switching IC with some copper tape and ground it (the tape)?

There is a sticker on the back of one of the cards which says "PowerColor". I assume this is the brand. Was it a lousy brand for these cards?

The "PowerColor" VGA cards I saw were not bad.

By the way, did you guys ever think about the possibility that switching noise from the supply voltages gets fed through to the signal?
If this is the case (which I suspect), slapping some more ceramics onto the card (particularly near the switch chip) might help.
It might be also/alternatively caused by supply noise "feedback" on the original graphics card, though. So you might want to slap some caps onto that one, too.

But swapping out the PowerColor Voodoo2 card for an STB branded one resolved the noise issue entirely. It is possible that the caps around the switching IC on the STB are better optimised. I haven't yet replaced the switching IC's cap on the PowerColor (ceramic) with the one from the dead Diamond (tantalum).

Use ceramic caps, tantalums are much slower. And who knows in which state the old tantalum is?

If you can get them easily, I´d suggest some, say, 4.7uf MLCC 6.3V for the CRTC/DACs, switchers etc.
Solder on wires on both terminals and then pull some shrink tubing over the cap and wires, leaving only the wire terminals exposed.
Then connect the cap to the supply pins the shortest way directly (usually straight diagonally).
This will even out any switching ripple very well.

Make sure the rated voltage is not much higher than the supply voltage, because the ceramics reach their nominal capacity only at the rated voltage; it's their speciality that the capacitance is voltage-dependent and decreases with lower voltage.

I know it's a pain but are you able to test this video set up in a completely different system? Also have you tried another display?

From my experience with Voodoo and Dxr3 decoder cards and their cables I am not surprised to see these problems. Imho maybe other brands might have better components or PCB design to avoid that but much might depend on both the VGA original pass cable and the monitor cable itself. Sometimes I've got similar noise on a 1080p VGA TFT with my almost new GT610 cause probably it gets some noise from some fan or externally from the cables outside the monitor.
I suppose that's why exists double shielded high end expensive wires/VGA cables far from the ones with usual monitors. I suppose like old game console with SCART and RGB cables, the wires quality make a huge difference in audio/video noise and even the good ones are often far from good beside those specifically said to be shielded in a good way from the beginning to the end of the pin connection for every wires.

Also the source VGA might improve things. I remember most Voodoo or Dxr3/Hollywood+ cards did benefit from a strong quality VGA output card like most Matrox ones when instead things got awful with some lower end ones like some nobrand S3 Trio/Virge. Also another thing imho to consider that might be obvious is to be specifically designed for those time CRT monitors and far from any LCD tech with their internal logics. I found many problems with LCD sometimes being more sensible or less about VGA output quality. But sometimes happened to find some good high end time correct VGA LCDs with cables double the size of an usual VGA cable, and that happene with an Eizo LCD of those times beside not comparable to how fast the panel tech quality advanced in those times. But probably those were still built with VGA only signal in mind and for high end professional usage.

As said I think to remember the original Voodoo to have worse results of the Voodoo II usually but still the differences existed compared to a native video card connection. The Dxr3/H+ decoder cards were much known for these problems and if the decoder card output components might have had a big impact on it, I suppose the thin original pass cable was a big part of the problem that was solved in their last MPEG2/4 PCI card.
And I'd add that different revisions/models of the same cards had different results too.. early decoder cards had lower VGA output quality from my tests compared to the latest H+ cards or latest Dxr3 (same layout/chip) models.

386SX wrote on 2021-09-26, 15:35:

From my experience with Voodoo and Dxr3 decoder cards and their cables I am not surprised to see these problems. Imho maybe othe […]
Show full quote

From my experience with Voodoo and Dxr3 decoder cards and their cables I am not surprised to see these problems. Imho maybe other brands might have better components or PCB design to avoid that but much might depend on both the VGA original pass cable and the monitor cable itself. I've that noise on a 1080p VGA monitor connected to my almost new GT610 VGA output cause probably it gets some noise from some fan or externally from the cable outside the monitor.
I suppose that's why exists double shielded high end expensive VGA cables and not the usual cheap ones found everywhere with every monitors. I suppose like old game console with SCART and RGB cables, their quality make a huge difference in audio/video noise and even the good ones were far from good beside some that were specifically said to be shielded in a good way from the beginning to the end of the pin connection.

Also the source VGA might improve things. I remember most Voodoo or Dxr3/Hollywood+ cards did benefit from a strong quality VGA output card like most Matrox ones when instead things got awful with some lower end ones like some nobrand S3 Trio/Virge. Also another thing imho to consider is that most of these combination of cards were designed for those time CRT monitors. I found many problems with LCD, even the early time correct ones and these solutions.
Also different monitors might have different results. 😉

As said I think to remember the original Voodoo to have worse results of the Voodoo II usually but still the differences existed compared to a native video card connection. The Dxr3/H+ decoder cards were much known for these problems and if the decoder card output components might have had a big impact on it, I suppose the thin original pass cable was a big part of the problem that was solved in their last MPEG2/4 PCI card.
And I'd add that different revisions/models of the same cards had different results too.. early decoder cards had lower VGA output quality from my tests compared to the latest H+ cards or latest Dxr3 (same layout/chip) models.

I second this.

Getting a card from a good manufacturer and high quality VGA cable can help mitigate this situation, many users here are using LCD monitors which worsens the situation regarding image quality.

Besides, I prefer using a S3 card from a good manufacturer (Compaq, ELSA and STB for example) rather than Matrox (there is more than just Commander Keen games 😉) for DOS compatibility reasons.

I'm having the same problem with a Voodoo 1. I tried different video cards with it, but the problem persisted (Matrox, Diamond S3). I replaced the electrolytic caps but I don't think it improved.
I don't have the original passthrough cable since I bought the card on Ebay. I'm using a very good quality cable that is 3 ft long. It's thick and have filters in both ends. I have my doubts that getting a shorter or the original cable will make a difference, but I don't know.

I own both a voodoo4 4500 and a pair of diamond monster voodoo II cards, is it possible to use them together?

No idea if the V4 will cause conflicts with the V2 SLI, I’m also curious if the noise will show from the pass through.

IIRC the V4 uses a newer glide dll than the V2 so it might not conflict.

I started looking at this issue again. A poster asked if the issue occurs if the PowerColor Voodoo2 is moved to other systems; yes it does.

I swapped the bus switch PI5C3384Q for that of the S3L38Q found on my dead Diamond card, as well as FB6, C36, and C38. The problem still occurs, but I am noticing that it is more transient now. If I let the system stay on for a few minutes, the on-screen noise tends to go away. It seems like the noise was a bit more consistent when I tested it out 2 years ago, but hard to say due to the passing of time. I decided to keep S3L38Q, FB6, and C36 from the Diamond card, but put backC38 for aesthetic reasons. Symptoms same as noted in last sentence.

I didn't notice any transient noise with the other 2 or 3 brands of Voodoo2 cards I tested. I don't think it is any coincidence that both of these PowerColor Voodoo2's exhibit the same behaviour. Could the PCB design alone account for this, e.g. poorer ground plane design? The only item I didn't remove was the VGA connectors. Possible? But seems highly unlikely. The VGA connectors on my other cards also don't have a shield over the back of the pins.

Another thought I had was related to the 1084 voltage regulator. My other Voodoo2 cards don't use a 1084. Perhaps this regulator creates more noise than the others? It is positioned right next to the bus switch. I noticed that the Gainward Voodoo2, which also employs the 1084, has the 1084 positioned way far away from the bus switch - on the other end of the card.