Indigo2 non-Impact Power Supply Repair - Hints welcome

mistermusixx

New member
Jan 7, 2022
6
1
3
Hello everyone here,

I'm an engineer trying to repair my neighbor's Indigo2.
Symptoms are, machine only starts until playing the startup sound, but no graphics appears, and the LED stays constantly amber.
From previous experience with aging (PC) power supplies, I opted for taking out the PSU of the Indigo2.
Espscially as my neighbor mentioned that the system needs switching on TWICE since quite a while to get started at all.
On opening, I only identified one electrolytic capacitor with a little bit of brownish residue, and changed that one right away.It's on the +12V apparently.
Only now as I get to the 5V, it has a very poor quality, with ~1V undershoot and ~+1.5V overshoot, so there is certainly something not good any more.
See attached scope shots.
However, adding just some capacity to 5V (4700µ Low ESR), to (temporarily) compensate for eventual capacity loss, didn't change anything. 5V seems also reasonably stable and sustained when swithcing off; a connected HDD runs for a couple of seconds after mains unplugged. So I guess the 2200µ cap's are not at fault.
Now, in lack of schematics, I would be grateful for some hints which components to address next, and what to try to get the regulator back on track. Still I guess it's some faulty capacitor, but I just can't change them ALL on suspicion. Maybe it's a smaller one having lost its HF capabilities?
Thanks in advance for any useful hints!

mistermusixx
 

Attachments

weblacky

Active member
Jan 13, 2020
141
37
28
Seattle, WA
Those of us doing PSU repairs (me - Indy and Tezro for now) haven't gotten to Indigo2 PSU yet due to the fact that there are literally 5 different PSUs!!! However there is only one non-impact PSU (so good for you). These Indigo2 PSU require a MINIMUM load to stabilize, so place something on the 3.3v line and also the 5v to make sure it's regulating. It will NOT BE STABLE without any load...it's not an ATX PSU. Seriously, Elf tested this.

Also while you're not wrong about the PSU needing work, you're likely wrong about the PSU stopping the system. While most old SGI PSUs do output bogus Power Good signals (it's not telling of good power), if the chime started...it passed basic POST.

I'd suspect something else before hand. Most new people don't know that ALL SGIs output a serial console while booting (literally bootloader) for diags by default. While Indigo2 and Indy use Mini-8 connectors you can go to any old PC recycler and get an old MAC modem cable and splice that. Pinouts for terminal interface are in the SGI indigo2 manual!

Use an old XP machine with hyperterminal and hookup a hacked together serial cable and get some output. You'll see where it's stopping. I'm going to suggest you may have a graphics card issue or a graphics plane issue.

That's not to say you don't have a PSU issue. But I'd stop on the PSU work, reassemble, do a bench load (reasonable) test and then place it back in the machine and do a serial hookup (without keyboard and mouse) and see what it says it's doing.

Then troubleshoot based on that. Maybe you're right in that it's the PSU. But MOST of the old SGIs rely HEAVILY on 3.3v and 5v (even for graphics). 12V is really for the Drives. They are odd that way. So if you have good 3.3v and 5v lines...you should be booting enough to see serial output and see where it stops.


I'd also say that reseating connectors won't be a bad idea as well. Also...remove ALL drives for testing a poor HDD can do odd things. So boot totally driveless to main PROM (BIOS) screen.
 
Last edited:
  • Like
Reactions: Elf

Elf

Storybook
Feb 4, 2019
589
164
43
Definitely agree with what weblacky said about testing under load, and that the PSU is not necessarily the culprit for the overall symptoms you are seeing with the system. As recommended above, you might want to hook up to the serial console and see if there is any output? If you can get some sort of PROM console over serial it may give a better explanation of what is going wrong.

With regards to the PSU, if you are still getting switching noise when under load and you've replaced (or provided an equivalent to) all the likely electrolytic capacitors, it wouldn't really be the regulator (really, switching controller) at fault. The switching controller is going to be involved in a much lower frequency feedback loop, and modulating pulse width or frequency accordingly to maintain a target voltage. However -- absent a time scale to really say -- what you are seeing looks like impulse noise / ringing on each switching cycle which would be downstream of the controller.

If it still persists when the PSU is loaded and you want to chase it down, the next things I would look at would be the switching diodes or any kind of ceramic capacitors that would be there for higher frequency filtering. I don't see any kind of time scale for reference in the scope captures, but it looks a very brief ringing that would be too high frequency for electrolytics to filter out. The electrolytics have their lowest Z (impedance) at lower frequencies (Hz to low kHz), and it tends to increase dramatically as the frequency goes up. Electrolytics generally provide a filtering response at lower frequencies and are paired with ceramic capacitors are able to provide low Z at higher frequencies (e.g. up to MHz or even GHz). Ceramic capacitors (i.e. surface mount MLCCs) don't fail as often or as predictably over time as electrolytics, but if they do it is usually by stress and cracking from what I have seen.

Aside from capacitors, in switching power supply designs the diodes involved in switching tend to dissipate a lot of heat and can go bad; a failing diode (or one that was a bit too slow) could probably cause some ringing as indicated. There may be many diodes -- especially smaller ones -- in the design, but you would be primarily concerned with the fast recovery or schottky power diodes in the switching path. You can see the prominent diodes Dr and Df (ignoring the package diode on the Q1 MOSFET) in this simplified forward converter topology. Lo and Co provide output filtering to reduce the sort of noise you are displaying, and Co would likely be a parallel series of both electrolytic and ceramic capacitors.

1641608880210.png
 
Last edited:

mistermusixx

New member
Jan 7, 2022
6
1
3
Hello all,

thanks for the hints so far. Some comments:
- Time base is actually given on the bottom of the scope shots; so the first one is taken at 2us/div, showing the repeat rate at 125kHz or so (=switching regulator fequency); the 2nd shot was taken at 20ns/div to better measure the over/undershoot, in AC mode.
- Shots were taken with 2x older HDD drives as Load on 5V (and 12V). I haven't seen a requirement for load on 3.3V too, but if required, I have to see how to get this handled (I'm not yet equipped with a programmable load, which sure would be nice to have; still I'm not a professional repair shop ;) ).
- Serial console will come out on Serial1? 9600 8N1?
- I'll probably have a look on the mentioned diodes before reassembly. Thanks for highlighting this.
- Based on the Troubleshooting section in the manual, my neighbor already opted for organizing himself a spare Graphics card. However, in the existing system there is a 3-board-stacked graphics card assembly in place, while the ones currently in reach are only 1-board or 2-board stacked ones. I suppose these should still be detected on startup, but he may (only) be losing some perfomance/features then? (What may be acceptable, only the system works again...)

Thanks for assistance, much appreciated!

mistermusixx
 

mistermusixx

New member
Jan 7, 2022
6
1
3
Hello all,

thanks for the hints so far. Some comments:
- Time base is actually given on the bottom of the scope shots; so the first one is taken at 2us/div, showing the repeat rate at 125kHz or so (=switching regulator fequency); the 2nd shot was taken at 20ns/div to better measure the over/undershoot, in AC mode.
- Shots were taken with 2x older HDD drives as Load on 5V (and 12V). I haven't seen a requirement for load on 3.3V too, but if required, I have to see how to get this handled (I'm not yet equipped with a programmable load, which sure would be nice to have; still I'm not a professional repair shop ;) ).
- Serial console will come out on Serial1? 9600 8N1?
- I'll probably have a look on the mentioned diodes before reassembly. Thanks for highlighting this.
- Based on the Troubleshooting section in the manual, my neighbor already opted for organizing himself a spare Graphics card. However, in the existing system there is a 3-board-stacked graphics card assembly in place, while the ones currently in reach are only 1-board or 2-board stacked ones. I suppose these should still be detected on startup, but he may (only) be losing some perfomance/features then? (What may be acceptable, only the system works again...)

Thanks for assistance, much appreciated!

mistermusixx
 

Attachments

mistermusixx

New member
Jan 7, 2022
6
1
3
Hello,

I now reassembled the unit, as you suggested, after some more sanity checks.
I then hooked up a laptop to Serial port1 at 9600/8N1, and managed to obtain the attached logfile. As you see, I was able to boot the machine fully up into the OS, however graphics is failing (see error message in log).

What I can also say is, that PS/2 does not work, and both mouse and keyboard are dead. I measure no clock on PS/2 ports, and so no LEDs (NumLock) appear.
I wonder whether this is a consequence of the graphics failure (does graphics card serve the PS/2 board?), or is this a functionality (failure) on the baseboard?

Thanks again!

mistermusixx
 

Attachments

weblacky

Active member
Jan 13, 2020
141
37
28
Seattle, WA
When it comes to basic post, graphics are graphics. Until the OS loads the system doesn’t care. After OS load, it cares very much! Graphics’s are detected and drivers are installed at OS INSTALL, not OS run. So the OS willl freak out if you change the graphics from IMPACT to something else. If you change from Max/high/solid impact…it won’t care. If you change from impact to non-impact it will very much care and normally require OS reinstallation.

3.3v is absolutely the line that requires minimum load!!!! It’s the heaviest used rail in the entire system!!

SGIs use type-2 PS/2 scan codes, not all PS/2 keyboards will work. Most of the 2004-2006 microsoft keyboards and mice work on them (that’s I’ve tried). Unless you’re using SGI keyboards and mouse, try a early Logitech or a Microsoft keyboards. Most PS/2 keyboards you find (new) are type 3 scan codes and incompatible with SGIs.

Also on indigo2 to the keyboard & mouse and audio are separate daughter boards. And can be repaired or swapped out. Be aware that the keyboard and mouse port contains the system identity chip, I think. That has the host ID used for software licensing
 

mistermusixx

New member
Jan 7, 2022
6
1
3
Hi again,

thanks for the hints.
However, I wouldn't have identified the PS/2 daughter board yet... The machine is the turqouise-colored "desktop" model, if there are different ones.
And swapping the PS/2 Board with the Hardware ID would be a major issue, since the SW (licenses...) must stay untouched.
Maybe anyone could help out with some photos?

Anyway, the KB and mouse are original SGI ones. And I think these are most likely not faulty. Still, the PS/2 ports have no Clock, as opposed to what I would expect.
Any idea on that one? Can this be due to the graphics failing, or should I be looking for something else?

Thanks again!
mistermusixx
 

weblacky

Active member
Jan 13, 2020
141
37
28
Seattle, WA
The teal models do have different graphics systems and cannot take impact graphics at all (don’t try). For HW testing purposes it does not matter if you swap graphics boards for any of the teal indigo2 generation (extreme, xz, etc), the OS will care but PROM booting won’t.

I’m assuming you have/had SGI keyboard and mouse that used to work on this station, won’t bother asking you to confirm the models as “slab style” granite keyboards and mice came in a dangerously incompatible PS/2 style plug designed for Indigo/onyx stations that use high voltage but a ps/2 connector that can be plugged into an indigo2 but we’re never supposed to be. So if you obtained these peripherals without being certain they were for this station, please double check their model numbers online (undersides).

The daughter board has a small chip on it that can be desoldered and moved to a new board to keep your licenses. Yes, the ports are on a daughter card, it doesn’t look that way, but it’s true. So look at the standup PCB itself with the mini-din 8 serial, keyboard, and mouse port. That’s the card.

You may find someone fiddled and a fuseable link went, i’ve never heard of this happening but then again a system could be abused in a number of ways. Also the PS/2 ports cannot be swapped. Keyboard is for keyboard, mouse is for mouse they don’t work any other way.

if your good with electronics, remove the card and examine for power issues or blown caps.

Otherwise, pickup another card from eBay (they do appear often) and swap it. If everything works, then swap the black IC on it to get your system ID back: https://wiki.preterhuman.net/Number_In_a_Can
 

Elf

Storybook
Feb 4, 2019
589
164
43
I would suspect that if you see nothing from the PS/2 ports from power on that it is not (just?) the graphics at fault. However, it is also worth noting that the PS/2 devices generate the clock signal and there is only a clock signal when data is actively being exchanged. Monitoring from boot is important, as no clock is the regular idle state, but there should be a boot init sequence of data exchanged between the host and keyboard. The clock signal is not continuous, but only present during data transmission. See attached PDF for details.

Is there also no +5V on the PS/2 ports, or does it just not init the keyboard at boot?
 

Attachments

mistermusixx

New member
Jan 7, 2022
6
1
3
Hi everyone,

OK, so I'll check for the PS/2 daughter board next time I get to the machine.
Actually I did not yet measure right from power-up, but I can state for sure that KB/mouse are original SGI ones that used to work all the time, and the 5V are definitely there. Even with the system fully booted (over serial port, still no graphics, as stated), KB/mouse are dead.
Let's see what the spare graphics unit can do once it arrives. If it is not damaged, it should possibly show something at least.
So it can take a little while before I can go further. It's still my neighbor's machine.

Thanks for hints!

mistermusixx
 
  • Like
Reactions: Elf

About us

  • Silicon Graphics User Group (SGUG) is a community for users, developers, and admirers of Silicon Graphics (SGI) products. We aim to be a friendly hobbyist community for discussing all aspects of SGIs, including use, software development, the IRIX Operating System, and troubleshooting, as well as facilitating hardware exchange.

User Menu