SE/30 32Mhz clock and UH7

[mejs]

Hi folks,
I will try to summarize this situation into a question, but I need to share some context first.

I own two original SE/30 boards. One was battery bombed, while the other was in great physical shape, but with a never identified issue which prevented it from successfully booting into any system. As part of troubleshooting, I replaced this board's F258 and F253 muxes, and the FPU. This replacement didn't fix my issue, but it correlated with me starting to experience bizarre issues with the 32Mhz clock which have now plagued me for months, across multiple logic boards, analog boards and power supplies. The issue effectively results in the clock signal coming off of UH7 fluctuating in voltage, and affecting other components as well. It *feels* like a short. But I've been unable to figure it out. This is how this issue manifests itself on the screen:

https://mastodon.vladovince.com/sys...254/748/083/086/original/245efedad420c7d9.mp4

Here is anothe example:

https://mastodon.vladovince.com/sys...292/603/195/174/original/d9b5914a07146ae8.mp4

This is what this looks like on the scope of the 32Mhz clock off of UH7. Eventually it stops moving and goes low. At that point the image would fully collapse too.

https://mastodon.vladovince.com/system/media_attachments/files/111/486/690/839/802/565/original/b4b20ebb735b45a5.mp4

I've been referring to it as "collapsing video", but I've collected at least a dozen of instances of failure -- most often I can't get any chime, but I'll get the stripes for 5-10 minutes until it collapses. Other times I might get bad or even good chime, then stripes. I have gotten more interesting simasimac patterns:

Regardless, it always results in the clock falling apart.

Here's where things get weird. This issue has now followed me across 4 different boards.

Board 1: "good" board that wouldn't fully boot -- first started exhibiting symptom

Board 2: battery bombed board, it could get stripes but nothing more. After I switched out its clock and UH7 with the "good" board, the problem started happening on this board as well

Board 3: Bolle board #1 - I tried to build a new board using the battery bombed's components. The same issue happened

Board 4: Bolle board #2 - frustrated with this issue, I got extra boards to experiment on. I tried to build a minimum viable circuit to test with, with all new components. This exactly same thing happens as soon as any other chips are installed other than UH7 and the Glue chip:

Very early on I suspected the analog board and power supply. I have two SE compatible chasis:

1. SE/30 analog board and PSU that came with the battery bombed board. In good shape, I recapped the analog board
2. SE analog board which I recapped a few years ago

Same issues happen in exactly the same way with both of these analog boards/PSUs.

My SE board work perfectly fine in both of these chasis.

At this point I am at a complete loss. The fact that I can replicate this on a brand new board, with all new components baffles me. I am grasping at straws, but I am presumably making a mistake here, somewhere, and perhaps I'm making the same mistake over and over again, or I'm ignoring something that should be obvious.

Any ideas are very welcome. I've been documenting this saga in a lot of detail over at Mastodon: https://mastodon.vladovince.com/@mejs/111008723163117281

And finally for the 2 concrete questions. S

- Should it be possible to incrementally test this like I've been doing with board #4? I've been building it one component at a time, until I run into the issue (I run into it very quickly). Does my plan with the mostly empty board make sense?

- How do the SE and SE/30 compare in terms of what state of PSU/analog board they need? Is it theoretically possible that my analog boards/PSUs have issues, but the SE which boots fine with both is more resilient?

Thanks everyone!

 

[croissantking]

Wow, this is quite an impressively tricky puzzle. It just doesn't seem plausible that you'd have the same problem with the clock crystal across multiple original boards and Reloaded boards. First thing I'd ask myself is, what's the common factor across all of these boards when I test them out?

- How do the SE and SE/30 compare in terms of what state of PSU/analog board they need? Is it theoretically possible that my analog boards/PSUs have issues, but the SE which boots fine with both is more resilient?

Very unlikely but I think it's important to question everything at this stage, including your house(!). Why not try to bench test your boards with an ATX supply and custom harness?

You could try a different oscillator on there. One of my Reloaded boards has this type:

 

[croissantking]

Also, have you looked at your voltages while the board is powered up? You could monitor it on the +5v at the oscillator.

 

[mejs]

Wow, this is quite an impressively tricky puzzle. It just doesn't seem plausible that you'd have the same problem with the clock crystal across multiple original boards and Reloaded boards. First thing I'd ask myself is, what's the common factor across all of these boards when I test them out?


Very unlikely but I think it's important to question everything at this stage, including your house(!). Why not try to bench test your boards with an ATX supply and custom harness?

You could try a different oscillator on there. One of my Reloaded boards has this type:

View attachment 14213

Thank you for this!

I did at one point suspect the house wiring, but it does seem unlikely since I live in a new apartment building. I also tested running from a UPS on battery, with same results.

I'll be testing with an ATX supply later today. The 5V input is solid, including off of the oscillator. I may try a different type, thought this issue has been occurring with a handful of different through hole ones, and when I test them outside of the board they all work fine.

 

[croissantking]

Let us know how it goes with the PSU.

Looking at your purple Reloaded board I see the ADB chip is installed wrong. Guessing you caught and corrected that? Maybe post a pic of it as you had it when you'd finished assembling it.

 

[croissantking]

Another idea is to run the oscillator outside of the board and then run a wire with the clock signal to the board.

Regarding your bombed board, you say the problems began when you swapped the clock crystal with the one off your original board. Did you try putting back the bombed board's original crystal again?

Something really doesn't add up here...

 

[mejs]

Let us know how it goes with the PSU.

Looking at your purple Reloaded board I see the ADB chip is installed wrong. Guessing you caught and corrected that? Maybe post a pic of it as you had it when you'd finished assembling it.

Oh wow, I absolutely did not catch that. It would have been when I first started testing the purple board, which did not go well. I eventually did remove all the chips, including the wrongly installed ADB one. I completely missed where pin 1 was on this chip vs my other board where the text is rotated by 90 degrees. It's possible this caused shorts and worse damage of components. Thank you for noticing this because this is something that I completely missed.

Another idea is to run the oscillator outside of the board and then run a wire with the clock signal to the board.

Regarding your bombed board, you say the problems began when you swapped the clock crystal with the one off your original board. Did you try putting back the bombed board's original crystal again?

Something really doesn't add up here...

Yes, I did swap back, and I also swapped with new crystals. One thing I'm solidly sure of is that the crystals are not the issue here -- the problems start once more components get added to the circuit (something I just tested by adding a crystal, which exhibited my same symptoms, onto a fresh board with just the caps, ferrite beads and the crystal)

Lastly I did do the test with an ATX PSU -- same thing happens, so I guess we can at least reliably confirm it isn't the analog board/PSU.

I do wonder now with evidence of at least one component incorrectly installed, is it possible that I just permanently damaged a lot of the chips I've been trying to reuse? I'm thinking for example of the Glue chip that is currently installed in my green board. That chip was previously on both the battery bombed board and the purple board with the incorrectly installed ADB chip. What if it's also damaged now?

 

[croissantking]

I do wonder now with evidence of at least one component incorrectly installed, is it possible that I just permanently damaged a lot of the chips I've been trying to reuse? I'm thinking for example of the Glue chip that is currently installed in my green board. That chip was previously on both the battery bombed board and the purple board with the incorrectly installed ADB chip. What if it's also damaged now?

Unlikely you killed more than just the ADB chip itself, although possible.

i'd recommend installing PLCC sockets on your purple board. I did this on two of mine and it made things so much easier to debug.

 

[mejs]

Hey folks, figured I would follow up here and confirm that I finally resolved this and successfully built a functional SE/30 on a new board, using parts from both of these original boards + a new CPU. I'm working on a blog post which I'll publish at https://blog.vladovince.com, but in the meantime this is a TLDR.

• the root cause of my recurring UH7 trouble (ie. clock/hsync/vsync etc.) must have been my shoddy soldering skills. Over the course of trying to correct this issue on multiple boards, I must have had bad solders and close almost-shorts countless times and at different locations. UH7 was particularly difficult because the pads are so close. Eventually, a friend recommended I try PLCC sockets, primarily to test components. But that also helped me with soldering, because I was able to actually see my work. I also found this post, which effectively advised the same: https://68kmla.org/bb/index.php?thr...p-sound-horizontal-stripes.45751/#post-511412
• additionally, I was definitely dealing with multiple bad chips, likely as result of prolonged work on this board where I dealt with many shorts. PLCC sockets were super helpful here, as well as the CPU socket which helped me finally realize that both my CPUs were likely bad. After lots of trial and error I finally identified a group of good chips. At the end, most of the chips were reused, other than the CPU, SCSI and FPU. A particularly bad one was one of my GLUE chips which was partially functional due to a damaged PIN that I re-soldered. I also ended up destroying one ROM SIMM
• even though I suspected one of my UH7 chips, at the end I ended up using that exact one. I got a 5ns and a 10ns GALs (https://www.mouser.com/ProductDetail/556-AF16V8C-5JX and https://www.mouser.com/ProductDetail/556-AF16V8B10JU) but at the end I got the most reliable performance with the original. I didn't do enough testing here, so I can't 100% confirm that these wouldn't have worked, but there was a difference
• eventually, confirming that my CPUs were inconsistent ended up being a major breakthrough. It's possible I either damaged them with all the work on boards with shorts, or while desoldering them (which was difficult and physically intense). In any case, I ended up getting an MC68030RC25B replacement because it was much cheaper than a 16Mhz 68030
• while working on this, I used an ATX extension cable for easy access. On the last day, that cable started failing which threw me for a loop, because the symptoms were exactly the same as what was going on with this for most of the process (collapsing image etc.). It ended up being HSYNC and VSYNC connections going in and out. Since I was basically done at that point, I just connected the board directly but this could have easily spiraled had I not realized the cable was the issue.

This repair took 4 months, literally, and I used up 4 new boards until I finally got it right. I don't think there are many exact lessons here for others, other than realizing my work was the problem, in addition to bad components and many other curve balls along the way. I hope this does end up being helpful to someone, because the "collapsing image" problem was so difficult to troubleshoot, and there ended up being so many causes for it in my case. At the end, patience, working closely with the schematics and the ability to actually test components with sockets ended up being crucial for success.

Many folks helped me along the way - encouragement on Mastodon was super helpful, and I received more direct help from others, but I do want to shout out @techknight -- your posts and videos regarding the SE/30 were critical for me to understand how the SE/30 works. This video in particular was super useful:

And of course @Bolle ! Without your amazing board this would not have been possible.

A very long and painful log of this project is available on Mastodon: https://mastodon.vladovince.com/@mejs/111008723163117281

 

[techknight]

Awesome you were able to use my videos as an educational tool! It is why I do what I do

 

[mejs]

I finally closed the loop on this and wrote a lengthy blog post about what it took to get two working SE/30 boards out of this mess

https://blog.vladovince.com/i-get-k...gain-the-sisyphean-macintosh-se-30-struggles/

 

[Daniel Hansen]

I finally closed the loop on this and wrote a lengthy blog post about what it took to get two working SE/30 boards out of this mess

https://blog.vladovince.com/i-get-k...gain-the-sisyphean-macintosh-se-30-struggles/

Epic!