SE/30 Pixel 'Snow' Artifacts After 10 Minutes or More

[hfrazier]

Hi everyone, I've had a bit of issues with the internal video on my SE/30. (The SE/30 has a Radius Color Pivot and an Asante MacCon installed.) After about ten minutes of use, pixel artifacts/snow begins to appear on the screen. Over the course of an hour they get worse and worse, maxing out at the 60 minute mark. It constantly moves and seems to reflect CPU/bus activity. (The snow will sometimes form patterns with CPU activity, kinda neat actually.) Performance seems completely unaffected, and there is no snow on an externally attached monitor through the pivot. The PSU, AB and Logic Board have all been recapped, with tantalums on the Logic Board. I thought that recapping the analog board and PSU would fix the issue, but apparently not. I assume that the issue occurs with some component warming up, and wanted to see if someone may know or have any pointers as to what it may be.

I uploaded a video, as well as some photos of what's happening. Thank you in advance to anyone that may know what's going on!!

View attachment IMG_6314.mov

 

[IlikeTech]

Bad VRam?

 

[hfrazier]

Bad VRam?

That definitely could be.

Especially if a region of the RAM used as VRam. Is it or is there a standalone chip? (I replaced the original RAM with some newer 60ns simms I had.)

 

[mraroid]

I think your idea that heat will cause this issue is a good one. Vram is a good place to start. You may be able to isolate the problem by spraying some cold spray on each item.  As soon as you cold spray the item that is causing the problem, it may go away.
 
This is kind of expensive, and you can probably buy it cheaper, but this is an example of what I use:

https://www.amazon.com/MG-Chemicals-403A-Super-Spray/dp/B008UH3NB8
 
Keep us posted.
 
mraroid

 

[IlikeTech]

I think it has a standalone 64K chip.

 

[Bolle]

Are you using my large adapter without a CPU accelerator to fit both PDS cards in?

Because that looks exactly like it did in my tests when I did try two cards in the adapter without an accelerator.

That's why I decided to solder that one /BR jumper down on some of the adapters to prevent usage without an accelerator.

My guess is that the long distance between logicboard and the two PDS cards is the problem here and something on the bus is getting upset because the driving capability of the logicboard on the bus signals is limited.

Pretty interesting that it even runs for so long. Mine crashed after a few minutes when the screen artifacts started to get worse.

 

[hfrazier]

@Bolle

I am actually not using it yet, still on the hunt for an accelerator. (I knew that one day I will find one so I didn't want to miss out on buying some from you! ..unfortunately I just missed out on this one.. had auctionsniper set too low..) 

For now, the Pivot is plugged into the 90 degree adapted MacCon.

I was wondering about a potential power draw issue, or the cards causing more noise on the bus, which is probably what is happening...

But since the problem arises over a period of time and it is fine at power-on, with absolutely no hardware changes, it's got to be something funky with a voltage slowly dropping or an altered power draw/supply as something heats up.. let me try running it without the expansion cards for a bit  and get back to you all.

As a side note, after re-capping everything and measuring voltages on the PSU (with no load attached) I did notice that the -12V rail was at -9.8V from GND. I don't know though if that's a big deal or has any affect on the logic-level stuff.

 

[Bolle]

Now that's even more interesting then. With those two cards plugged directly into the logicboard I never had any issues so far.

Might have to give it another go with adapter without accelerator and two cards to see if I can find a cause for this.

It has to be something VRAM related. I suspect it might be the data lines being too weak to saturate the inputs on the VRAM. If it were for the address lines there would be more things going wrong and the artifacts would form more harmonic patterns.

It might be worth checking how D24-D31 are connected on the PDS cards ie. if they are buffered or not.

 

[pb3623]

@Bolle

I don't know if this is worth mentioning (because it seems like a wholly different issue) but remember the issues I PM'ed you on? The intermittent freezing that I thought might have had to do with PSU load, heat or the lack of an FPU on my Turbo 040? Well, it devolved to flat-out failure to chime on boot (with checkerboard) so I plugged and prayed w/ various PDS card combinations, overlooking the obvious and unplugging EVERYTHING and still had the problem...

So, after further wiggling and testing what I thought was an effective shim, I determined my BMOW ROM SIMM had loosened ever so slightly over time. I had a brass wire wrapping around the back of the logic board that I had to tighten (the loop got messed with after all the card swapping). So far, things are stable.

I'd ask if @hfrazier thinks this might be a problem or at least it's an easy thing to rule out? I'll posit that it would be a total freeze and not artifact, but what the hell.

 

[hfrazier]

Update-

-I ran the machine for an hour without the MacCon and Radius, no artifacts whatsoever. 

-After running for a hour, I plugged both the cards in and the artifacts are back. (<1 min turned off)

-Removed the Pivot, running only the MacCon, there are no artifacts.

-Installed another pivot that I have, just to rule out an issue specific to that Pivot, artifacts are back.

-Turned it off, re-seated ROM and RAM and both cards, let sit for 10 minutes, turn it back on, there are few artifacts, but they get heaver the longer it is on, just as before.

@pb3623 The ROM seems to be fine, I wish it was that!

@Bolle What chips/numbers should I look for on the PDS cards to see if they are buffered? Also, I have a Logic Pro 8 from Saleae, I could do an analog sample for an hour at a low sample rate of all the PSU voltages, or record some of the PDS lines... 

Something else: (Everything is done on an external monitor set as primary with the 50% dithered gray background on the internal.) There is a bit of a pattern for some things. For one thing, the noise only makes the pixels go ON, not OFF. (See MacsBug scree/white background) And another thing I noticed is that there is maybe a bit of a harmonic in that there are columns that don't get as badly affected about every 8 pixels. The last image is of what you will see with the force quit dialog open.

 

[hfrazier]

Something else I noticed: the white also flows over the allowable screen space on the right. Could this provide some additional insight to someone who knows the hardware better than I?

Edit: Also, I guess is does get worse after more than an hour! I had it to the point where the columns are almost white.

 

[hfrazier]

Is there an easy way to obtain an extension cable for the analog board connector so that I can run it on its side with the logic board laying flat? I think I'm going to try the air duster approach and cool components one-by-one.

 

[joethezombie]

Yes, a standard 24-pin ATX extension cable can be easily used.

 

[hfrazier]

@joethezombie Ok, thanks!

 

[apm]

I've seen something similar twice before, and as far as I know there's no verified cause yet:



The second one is also related to the presence of a PDS card, which makes me think this has something to do with loading on one or more signals. If I had to guess, I'd say it would be the 16MHz clock, and that it's capacitance rather than resistance as the cause.

If it were a data line, you would see the "snow" stand still because the VRAM is only written when the screen changes, and the data to the screen comes out a different set of pins on the VRAM ICs themselves (i.e. it's never on the main data bus to draw the screen). Same story with the address lines. Also, the fact that you see white pixels after the end of the screen area is a strong indicator it's not the contents of VRAM, since that part of the scan line is not actually loading data from VRAM at all.

The air duster cooling test is a good one. Try UE8 and each of the video-related PALs (UG6, UG7, UE6, UE7). But perhaps especially Y2 and UI6, which generates the 16MHz clock from the 32MHz crystal Y2. It may be that as UI6 (say) heats up, the strength of its output drivers fall and it becomes more sensitive to capacitance.

If capacitance on the clock line is the cause, then what to do about it is not so obvious. Buffering the clock signal at the PDS slot would be useful but would need some kind of pass-through card. I have some other ideas but will save that for actually finding a verified cause.

 

[Bolle]

On the PDS adapter we made the clock signal is buffered and I still get the snow on certain card combinations so pretty sure it has to be something else.

 

[techknight]

I wonder if the processor doesnt have enough oomph to drive that much fan-out. 

 

[hfrazier]

@apm Awesome info! Thank you!

I am going to pursue the cooling method for now and report back. My ATX connector extension cables came in, hopefully I will have some time to set it up and test this week!

 

[Bolle]

I wonder if the processor doesnt have enough oomph to drive that much fan-out. 

That's what I am thinking as well, because in my experiments once you add a PowerCache to the mix the artifacts are gone.

The PowerCache has buffers which most probably deliver more power than the bare CPU driving the bus on the stock logicboard.