Deksor wrote on 2021-05-13, 21:35:

Can you please test this card with pinball fantasies ? the ramdac is similar to the original one on the first trident I improved (it was DT branded). I want to see if it shows the same palette problems. (FYI, the game was configured in high res mode).

Ok got the shareware of the game. I don't see any obvious palette issues. The colors of the clown table seem to match the screenshot in both normal and high resolution mode. What exactly should I be looking for?

You need to configure it for the high res mode. That's the only mode that triggered the most visible artifacts of my cards. If you don't see them I guess your card is fine.

It's just that I found this game was triggering some strange artifacts on some cards and it can run on a litteral potato. However if you find other programs/games that can trigger such visual problems I guess we could make a whole testing suite with them 😁

Deksor wrote on 2021-05-17, 19:39:

You need to configure it for the high res mode. That's the only mode that triggered the most visible artifacts of my cards. If you don't see them I guess your card is fine.

I have. I don't see anything obviously wrong

maxtherabbit wrote on 2021-05-17, 19:08:

Are these photos of an LCD or are you using some kind of capture?

Those are photos of the LCD screen.

Well some cards are working well with that game. My two et4000 are great for it, so is my now improved/fixed trident 8800. I think low-end/cheap early ISA cards are more prone to have issues with that game compared to newer ones. What is your card ?

Deksor wrote on 2021-05-18, 07:29:

Well some cards are working well with that game. My two et4000 are great for it, so is my now improved/fixed trident 8800. I think low-end/cheap early ISA cards are more prone to have issues with that game compared to newer ones. What is your card ?

It's the Trident 8900 I posted a picture of earlier in the thread

douglar wrote on 2021-05-18, 00:59:

maxtherabbit wrote on 2021-05-17, 19:08:

Are these photos of an LCD or are you using some kind of capture?

Those are photos of the LCD screen.

The jailbars on the GD5429 are far more extreme than the kind you would expect from a sampling error. I'm guessing this card is outputting a lot of high frequency noise on the RGB lines.

I swapped the electrolytic caps from my trident 8900C, aaaaand ....
https://www.youtube.com/watch?v=vAVSCNQWxdw

Epic fail, nothing changed.
There are tons of interference on the output.

What could be the culprit ?
As a reminder, this is the problematic card :

I don't think it's the motherboard or the PSU, the PSU is brand new and I'm using dotTB's ATX2AT adapter, and both the motherboard and the PSU are the same used for the Trident 8800.
So the electrolytic caps are fine and the RAMDAC is too, what's next on the checklist ?

Deksor wrote on 2021-05-18, 21:48:

what's next on the checklist ?

This is unexpected colors in the color pallet, yes? I'd guessing it is a ramdac issue.

Are you able to swap out the RAMDAC? You might need to get another Trident card to get one though.

No, the video output seems to have some interferences going on (if you pay attention you'll see some colors being shaky). It's hard to see but it's there.

I already tried to swap the ramdac for an inmos and it did the same. I also tried new caps and it was a fail.

Could the crystal oscillator be damaged or aged out of spec ?

I would really like to see a side-by-side comparison of these cards outputs on both an LCD and an SVGA CRT. I think we're all assuming that the 'jailbars' and other glitches are LCD sampling artifacts, but maybe the card's output is just really awful full stop.

I have the hardware to do that, but not quite the right setup 😒
My CRTs are in another room and I don't want to move them out of there

Most of these jailbars will disappear with CRTs though, I think they're originally lines that appear slightly on flat screens, and the scaling make it look worse.

All the recapping in the world won't fix bad power trace / analog signal layout and placement. 🙁
Noise on the output could indicate a ground loop someplace.

It makes me wonder... Assuming that many of those different legacy and vintage graphics cards used the same interchangeable RAMDAC chip, how feasible would it be to design a drop-in replacement chip (perhaps using an FPGA) that would convert the image to a digital DVI/HDMI (or DisplayPort) signal instead of running it through a DAC and converting it to an analog VGA signal? So basically a common retrofit to add digital video output capability to these old cards.

Perhaps such a project already exists?

Tiido wrote on 2021-05-03, 13:23:

The jailbars effect on LCDs and other digital devices is supposed to be due to mismatch between pixel clock the LCD expects and what is actually used by the card. Many of these cards have PLLs that do not produce exact VGA/SVGA etc. spec clocks, some have no PLL at all and create the pixel clocks by dividing some value they have around and fudge timings so that good resolution is achieved. It is why many cards have those wonky refresh rates at various resolutions.

This is the problem I ran into recently with an Oak OTI-067 VGA chipset on a Packard Bell mobo. Oak had a dedicated IC to generate the various pixel clocks for various resolutions (OTI-069). I scoped the clock signals and it had tons of phase noise.

Oak says it should have a dedicated oscillator and must be within a few millimeters of the IC. Packard Bell omitted the crystal and ran the ISA clock all the way across the board to the IC.

I made an interposer that disconnected the ISA clock and fitted a dedicated crystal. This removed all the pixel clock/phase noise on LCDs for this particular machine.

Excellent, another way to get better results ~

https://www.analog.com/media/en/analog-dialog … ber3.pdf#page=3