SGI: Hardware

My RM5200 O2 has had all the usual problems.
a) The occasional "oh no you touched me now I won't boot without the jumper trick".
b) The occasional ECC failure. Generally near the top DIMM slots.
- Looks like one DIMM indeed has a missing decoupling cap
- Always suspected thermal issues near the top slots.

My O2 hasn't been powered on or even plugged in in a long while.
Instead of admiring the box, I figured it was time to clean the DIMM slots to try to isolate/fix ECC errors so I could enjoy the O2 again.

Plug it in and now there's nothing on the power LED at all.
Which of course implies PSU trouble.
http://techpubs.sgi.com/library/tpl/cgi-bin/getdoc.cgi?coll=hdwr&db=bks&fname=/SGI_EndUser/books/O2_OG/sgi_html/ch04.html
And no, none of the obvious things helped. (Removing modules, reinserting logic board, the jumper trick, etc.)

I disassembled the PSU. It's Rev N.
There's nothing immediately obvious. No scorch marks. No leaking electrolyte. No bulged/popped caps.
But from experience I know that electrolytic caps go bad often without visible indication.
And this O2 PSU is clearly designed to bake its caps.
I'm no PSU design or repair guru, but I've fixed my share of game console and arcade board/PSU.

This thread was most intriguing:
http://forums.nekochan.net/viewtopic.php?f=3&t=16726251&p=7353975

because when I trigger the PSU via the blue wire (i.e. fan running), unloaded and disassembled, the yellow wires read 5.6V instead of 3.3V.
The other 3.3V rails are fine.
The +12V, -12V are fine.
The red +5V is a little high (5.40V), but again, this is under no load so it's fine.

So is the O2 smart enough to see this (i.e. is there an Over-Voltage Protection circuit?) and that's why it's not powering up?
There's a whole row of diodes/regulators on the second heat sink. Is it fair to assume the yellow wire 3.3V rail is being generated from the 5V rail? (i.e. one of these could be dead and letting a full 5V through instead of 3.3V)
Thanks to the compact design, the final electrolytic caps are also so close to these power transistors/diodes you can't even read most of their part numbers, and again, that means the PSU is designed to bake its caps, which also isn't fun. (This is also a common problem in arcade audio boards and a few consoles like TurboDuo.)

Is anyone here adept at analyzing and repairing O2 PSUs? I could always resort to buying one off ebay (like the Rev A1s I often see. yeesh, would you trust that revision knowing there's a rev N?), but then I'd have the broken one lying around forever, without the heart to add more junk to a landfill that could be repaired.

That is most unfortunate.

The biggest problem is whether the PSU is at fault or the logic board, since if the system is dead how can you tell if it is one or the other? The PSU apparently gives the correct voltages apart from the yellow wires but other than hooking up to a scope and comparing it with a known good working PSU, it's hard to figure out if the PSU is flaky in itself.

What's the story on the logic board? Any dark traces or burnt components?
And did you do any modifications to the PSU, like replacing the fan?

The easiest approach is to source a known working O2 and start swapping components. I would first make sure that the logic board isn't at fault by inserting it into a working machine.

Hi, nice to see someone else taking an interest in the O2 hardware.
The yellow wires are the main logic supply, where most of the power is. Unlike, say, an Intel box, the O2 regulates its logic supply in the PSU and does not have any point-of-load regulators. If they are as high as you measured, then that supply circuit has failed (most likely from a shorted pass transistor, but there are other failure modes in a switcher you need an oscilloscope to find). Good luck.

Dexter, yes, that's exactly the complication.
I used to have 3 O2s, including a rack-mounted Weather Channel O2, and several other SGIs.
But even I have pared down to just one O2 and one Indy.
So swapping parts isn't going to work, and these days I'd prefer to repair things first before resorting to buying extras off ebay (call me crazy).

I doubt the logic board is really at fault regarding the PSU. (ECC errors will have to be debugged later.)
Nothing is obviously damaged (yet) or modified there, though one hopes the dead 3.3V rail didn't destroy anything on the logic board. (I am hoping the O2 has an over-voltage protection circuit inside on the logic board or the front plane? Does anyone know? That would explain why the LED refuses to turn on at all. The O2 *could* have been made smart enough to protect itself. Was it?)

The PSU fan is original and still works fine. It's literally a case of "O2 not powered on in a year, suddenly doesn't power on at all".
I've encountered this before, outside the SGI realm. 1980s Atari arcade game PSUs often do the same. I've repaired 2 of these, but usually they don't leak excessive high voltages to the output.

An OVP circuit would need to contain a pretty beefy SCR, or Transzorb type varistor component. I never saw one but they could be hiding somewhere. The LED is also controlled by logic, not directly on the rail (reflecting its 3 color states), so it could stay off due to failed logic.

I have decided to take a gamble and actually send it out for professional repair.
Naturally this is going to be more expensive than doing it myself, and, in fact, probably a few times more expensive than the system is probably worth by ebay standards.
But
A) it's a good opportunity to resurrect equipment instead of adding more junk to a landfill
B) it's always good to support smaller businesses
C) it's ALWAYS good to support a repair shop (also to always support your local video/pinball arcades)
D) it's good to know places that can service SGI gear. Or at least Sony PSUs.
E) I have about half a dozen more repair projects to start after this, including another arcade sound board. How much is my time worth?

Best of luck! I would ask the repair shop to let you know exactly what they did to resurrect it - assuming they manage to do so. It's always worth knowing this for future incidents.

One of my O2 PSUs had the connector anchor points break off. I ordered a replacement connector and the pin extraction tool. Best of luck on your repair.

I would ask the repair shop to let you know exactly what they did to resurrect it

If I were to do this myself, I'm pretty sure the 3.3V is being derived from the 5V rail (no I didn't trace it to be sure), and probably one or more of those devices near it on the secondary heatsink is at fault (various power transistors, rectifiers, etc.). The right thing to do is replace those devices, probably check any thermal interface material (if any) between them and the heatsink, and replace all the electrolytic caps for good measure.

Without a schematic in hand, I would have to desolder all of the caps first in order to even read the part numbers on the other devices. How annoying.
This company claims they can fix it and test it, and I'm sure they'll do a more thorough job than I would. That will just make life easier.


Did you see the other thread here about O2 thermals? i.e. the orientation of the DIMM slots is wrong for thermals.
That problem extends into the PSU, too.
Consider the orientation and shape of the PSU's second heatsink. There is poor airflow over those final output capacitors (near the connector), yet they are all located very close to large heat sources.

Follow up.
Now I'm more frustrated than ever.

The company successfully repaired my PSU and returned it promptly. I was quite happy.
I did one final visual inspection of the O2 mainboard, inserted the poweron jumper (breathe on an O2 funny and it won't boot without that stupid jumper), inserted the PSU, and connected all wires.
Happily, the O2 sprang to life, with all components and RAM visible in hinv!

As it was late, I opted to do a more thorough test in the morning. I disconnected all cables to power down the system.

In the morning I removed the mainboard and removed the jumper, reassembled the system, reconnected all cables, and .....no power LED.
The PSU is dead again with the same symptom (3.3V rails read 5V)

I suppose I can contact the company again. They do claim to smoke test their work. And the fact remains that the O2 DID power up once fully, so I'm sure it WAS OK for awhile.
So, what is it about an O2 that can kill a PSU just by removing the power cord? Some crazy form of inrush current from some inductor?
Are we at a point where you cannot remove the mainboard without removing the power cord and you cannot remove the power cord without killing the PSU itself?

Then my most likely assessment is that the logic board ate your PSU (again). I feel your pain, but i think it is time to track down a new logic board, since jumpering it to force it to power-on even when your PSU is repaired is suspect in itself.

I have known good spares for the O2 and you can have them for the cost of shipping. I may have an entire O2 without plastics you can have.

Pm me your name and address and I will get you a shipping cost.