SE/30 Reloaded VRAM issue

[croissantking]

I have the following graphical issue with one of my Bolle builds - vertical bands of data corruption.

It takes quite a while to show up. In this case, I’ve had the setup running for a few days.

My guess is that it’s a VRAM issue, either a bad chip or a marginal solder joint. Is it possible to deduce more just from the photo?

 

[Phipli]

The first and easiest thing to do is reflow the solder on each of the VRAM chips.

 

[Bolle]

The VRAM is only 4 bits wide each chip, so the two together form 8 bits of video data which translates to 8 pixels.
As the pattern seems to be 8 faulty columns of pixels followed by 24 good ones you most likely have a problem with addressing. If a VRAM chip was bad the pattern would repeat in a different fashion.
Check the VRAM address muxes and all the connections around them.

 

[croissantking]

I cleaned and inspected the area around the muxes and my solder work looked good. I noticed one of the legs on UE8 was showing a bit of copper, obviously slightly corroded from its previous service life so I have swapped this over with a brand new part. I’ve reflowed the F253s too. Am testing again - running After Dark continuously.

 

[croissantking]

In addition to the LS166 I’ve replaced the F253s and VRAM with brand new parts. I’ve reflowed the PALs and the LS393s. All my solder joints seem to look good.





No joy.

The issue with the graphics is intermittent but usually starts after the machine has warmed up.

 

[zigzagjoe]

You could always try the old trick of slightly warping the board while the issue is present to see if it exacerbates the issue or makes it go away. If that makes a change, you know you've got a bad contact somewhere.

 

[croissantking]

You could always try the old trick of slightly warping the board while the issue is present to see if it exacerbates the issue or makes it go away. If that makes a change, you know you've got a bad contact somewhere.

You’re onto something. While I was putting pressure on the area by oscillator/UH7, the graphics corruption appeared, and then the machine rebooted. I only managed to get it to do it once so far. Not sure how much flexing is safe!

 

[croissantking]

@zigzagjoe it’s one of the F258s (UJ4)! When I press down on it, the problem appears and when I release it, the problem goes away! Thank you so much!

 

[zigzagjoe]

Awesome! With any luck, a quick reflow, and you'll be cooking.

 

[croissantking]

It was a bad chip, in fact!

 

[croissantking]

I spoke too soon. Problem is still there. But flexing the board at that location still 100% influences the problem.

Back to the workbench. Might be a late one

 

[LaPorta]

Is it possible (I hope not) that you have an iffy trace within one of the inner layers that happens to run underneath that spot?

 

[croissantking]

Is it possible (I hope not) that you have an iffy trace within one of the inner layers that happens to run underneath that spot?

How likely is that on a new PCB?

 

[zigzagjoe]

SE/30 reloaded is a 4 layer board, the internal layers are gnd/vcc with 2 or 3 traces beyond the ground/power planes. So it shouldn't be an internal issue.

 

[LaPorta]

Hopefully not. Perhaps a random break in a trace on the top. Maybe wait till it heats up and the issue happens, then buzz out the traces in that area.

 

[zigzagjoe]

@croissantking, want to post a picture front/rear of the UJ4 area?

 

[croissantking]

@croissantking, want to post a picture front/rear of the UJ4 area?

Here ya go. It might be hard to tell anything from these pics. I’ve just fitted the new F258s, and reflowed the Bourne filters + UJ6.

 

[croissantking]

I wonder if an address line is being pulled to ground. Could I try shorting the different input lines of the VRAM muxes to find out which address line is responsible for the problematic column on screen?

Meanwhile I’ll try reflowing more components in the UJ4 area. I am confident I’ll narrow this down.

 

[Phipli]

I wonder if an address line is being pulled to ground. Could I try shorting the different input lines of the VRAM muxes to find out which address line is responsible for the problematic column on screen?

Meanwhile I’ll try reflowing more components in the UJ4 area. I am confident I’ll narrow this down.

l'd try tapping each chip one at a time with a biro with the lid on. I wouldn't add shorts.

Don't stop when you see the issue, stop when you've checked every component.

 

[croissantking]

I felt like I was close to the solution yesterday, but can’t seem to influence the problem today. It does seem so weird. I hope I didn’t just get a bad batch of PCBs, both of my other builds have graphics issues too, although a different one.