tony359 wrote on 2021-10-18, 17:53:

Hi all,

I have a Trident ISA video card which has always been a bit glitchy. That was bought as "faulty" and I believe all I found was a capacitor that needed replacing.
It's been working since, even though on some graphic the output is "striped", see picture below.

The decoupling caps that are installed are for the DIP RAMs.
There are no caps installed for the SMD RAMs.
Mounting holes seem to be there already.
I am quite confident that adding these might help, as this problem is common when RAMs are being operated with insufficiently decoupled voltages.

interesting, thanks.

You mean as highlighted on the pic below?
What type and value of caps would you recommend?

Thanks!

It looks like data gets corrupted when you write to certain spots of the video RAM. The prime suspect in this case obviously is the RAM, but other parts of the system might be unreliable, like the ISA bus. If the ISA bus is overclocked, this might be a cause, too.

I don't know about the TVGA8900D in special, but a lot of SVGA cards work with less RAM than maxed out, so it is likely that your card works with 512KB. You would need to desolder the correct two chips, but this would likely also reduce performance, as the memory width decreases from 32 bits to 16 bits. If the memory organization of the trident card in 1MB mode resembles the original VGA card memory organization (it doesn't have to), inverted characters point to a problem in the second of the four emory bytes. That's bad news for you, because most likely the chips you need to desolder are responsible for the third and fourth byte, so I guess you would have to shuffle chips around.

Another idea: If you desolder the less common 8-bit memory chips, you can solder eight standard DIP sockets into that card and use 44256-like DRAM chips (256KB x 4 FPM) that are very commonplace.

tony359 wrote on 2021-10-18, 18:08:

You mean as highlighted on the pic below?
What type and value of caps would you recommend?

What you marked in the pic below might be what tony359 intended, but this is no space for decoupling caps. Caps at that position would be between pin 1 of one DIP chip and pin 20 of the neighbouring DIP chip (likewise pin 10 and pin 11). Pin 1 is a data pin, you do not want to load the address pin with a decoupling cap! The same is true for pin 11, which is an address pin.

I suppose the installed decoupling caps are close enough to the SMD RAM chips and work perfectly.

As long as you don't have any special requirements, the general recommendation for decoupling caps is 100nf multilayer ceramic caps, but anything close to that would work, too.

I didn't take a close look, just supposed because of the placement of the holes.
You really want to make sure that you connect Vcc and Vss(ground) on the shortest way possible, to reduce/avoid voltage dips that might result in bad data reads/writes.
Usually this means placing a .1uf ceramic between the lowermost right (number n/2) and uppermost left pin (that with the highest number), these are commonly (but not always the supply pins.
The only datasheet I found of that MM512256C is behind a login barrier, and I am too lazy to find out. Maybe you want to find 😀

Personally I first would try improving decoupling, as this in my experience solves this typical issue most times, before trying more radical things like replacing chips (which might also be useless if the decoupling is borderline).

I'd be happy to try but... this card used to work (with stripes but...) so I assume something has broken? Or are you suggesting that the original design was poor so maybe a capacitor got a bit worse and now I see corruption?

I tried removing the top two chips - they're labelled U5-6 and U7-8 but the board started with corrupted graphics. I put them back swapped and the corruption is in the same place. I'll try the same with the bottom ones.

Yes, there are many causes why old borderline electronics suddenly show issues.
Maybe a very little more bus stress, maybe a bit more ripple because of PSU decay, heating turned on, or whatever.

What you did seems to at least show that the RAMs are not the issue.
I have solved this typical video RAM issue often just by slapping on some extra caps like described in my last post. There are cases where the standard .1uF decoupling is apparently not sufficient.

I found some 0.1uF capacitors in my stock - not sure if those are the right type but I gave it a go. No change.

Edit: Oh, on a different motherboard it works fine! 😁
However another similar video card on this motherboard works fine.... uhm...

(Stripes are always there but that's always been the case)

Edit 2: and now it works again on the original motherboard....... ahah! 😀 I give up.

Any ideas on the stripes? Those are still there.

Check if you have those stripes with CRT as well, if you can.

It is a common issue to have vertical stripes/gradients repeating on LCD screens with old videocards.

good point as it only happens with some resolutions - but not with another video card on this monitor. Any other videocards I tried on this monitor does not show issues.
Unfortunately I don't have a CRT monitor around...

That said maybe you're right: the stripes move around when I change the "phase" setting on the monitor. So it seems that it's not on the signal. Weird that it only happens with this card but I suppose it's not impossible.

So, to recap, today I did NOT repair the card that did NOT work but which now works and I did not fix the stripes which are not an issue.

Ok, better go to bed.

Thanks for the help all of you! 😁

Umm, you wrote that it worked for a moment...
Maybe use ceramics instead of film caps, and it is no sacrilege to use higher capacitances, too, like 0,22uF.
And these vertical bars might also indicate fluctuations of the DAC output because of insufficient decoupling. Have solved such often the same way with adding capacitances.

This definitely looks like bad RAM especially when things appear in same places every time. It looks like the RAM on this card isn't following any JEDEC standard I know of so finding replacements may not be straightforward.

tony359 wrote on 2021-10-18, 21:10:

good point as it only happens with some resolutions - but not with another video card on this monitor. Any other videocards I tried on this monitor does not show issues.
Unfortunately I don't have a CRT monitor around...

Not all cards produce these stripes on a given monitor, Trident and Cirrus Logic are notorious for this.

retardware wrote on 2021-10-19, 00:00:

Umm, you wrote that it worked for a moment...

I cannot find that - if I did, I did not mean that. It used to work, then it started showing those inversed text. Then, all of a sudden, it was fine again. I have been unplugging and re-plugging that board many times as you may imagine so I doubt it was just a bad connection. I really don't know!

The stripes "move" when I adjust the PHASE control on the monitor so it seems that they are not on the signal. I can try a different monitor of course - but I do not have a CRT. It's weird as this monitor has always behaved well with anything I threw at it. But it's a possibility of course.

I guess it's fixed "for now". I'll let you know if the issue comes back...