[timob0]

Hi,

I‘ve been reading a while on this forum, now registered to ask for advice on an issue with my green Octane that out of the blue doesn‘t power on anymore. I used it a few months ago and shut it down cleanly. The other day, when I pressed the power button it just didn‘t do anything - no light, fan, nothing.
Checked the obvious - power cord, wall socket etc., without luck.
One thing that may play a role is that last time it worked, there were issues with sticky harddisk heads that could be resolved with pulling out the disk and plugging it back in. Could this have damaged the PSU? What would be good points to check to isolate the problem? I‘d replace or recap the PSU once I‘m confident that helps.

Since this is my first post, let me quickly introduce myself: I live in Germany, work in software engineering and I collect/repair vintage computers since several years with the objective to restore them to usable, original condition. I focus mainly on 90s Unix boxes (Sun Sparc 10, HP 9000/712, SGI Octane) and relatively rare early portables (Apple Mac Portable, IBM PS/2 P75).

Cheers!

[robespierre]

One thing that may play a role is that last time it worked, there were issues with sticky harddisk heads that could be resolved with pulling out the disk and plugging it back in.

!!!
Never heard of a problem like that. The idea is that cycling the head lock solenoid will forcibly knock them loose from a sticky buffer?

Could this have damaged the PSU? What would be good points to check to isolate the problem?

It should not have caused too much stress to the PSU, at worst an extra amp or so on +12V. Is it a heavily loaded system (I mean, maxed out DRAM, PCI shoebox, dual head or V12 gfx)?
I think there are some test points on the frontplane for standby power measurement. No standby power would be a sign of PSU failure.

[timob0]

One thing that may play a role is that last time it worked, there were issues with sticky harddisk heads that could be resolved with pulling out the disk and plugging it back in.

!!!
Never heard of a problem like that. The idea is that cycling the head lock solenoid will forcibly knock them loose from a sticky buffer?

Could this have damaged the PSU? What would be good points to check to isolate the problem?

It should not have caused too much stress to the PSU, at worst an extra amp or so on +12V. Is it a heavily loaded system (I mean, maxed out DRAM, PCI shoebox, dual head or V12 gfx)?
I think there are some test points on the frontplane for standby power measurement. No standby power would be a sign of PSU failure.

Should have been more clear re. the disk - of course I didn't hotplug it. Here's more detail on what happened:
- Powered machine on, fans, light and video come up, however when it tried to spin up the disk it couldn't and "power cycled" itself, i.e. fans, light, video went off, then it started again with the same symptoms
- Powered the box off with a long pwr button press
- Pulled the disk out, shook it gently and plugged back in (had seen this with other old hdds as well, heads seem to stick to the platters, preventing them from spinning up)
- Octance started up normally after that and I used it for the day and shut it cleanly down

Re. the HW configuration, it's fairly basic - Single 300MHz CPU, 1.2GB Ram, single 9GB hdd, SCSI tape drive, SE graphics (w/o TRAM), no PCI cage or shoehorn. 

Now, several months later, it's dead - nothing happens on pressing the pwr button and no evidence of standby power - no leds or similar. (Power cord and wall outlet are fine btw.)
I also tried to pull the disk before powering up since that seemed to be an issue before, but still nothing. I believe the fact that not even the PSU fans spins up indicates some problem with the PSU itself or maybe a short in the machine. I just can't imagine how that should happen with the box sitting there powered down. My suspicion is that maybe the PSU capacitors have reached the end of their life and need to be replaced - I just fixed a 30+ years old wall charger that couldn't hold a steady voltage by recapping. It had clear signs of electrolyte leakage and some corrosion on the PCB already. Got it back to working fine with new caps. Is it as easy with the Octance PSU? 


Re. the standby test-points: Do you happen to know where those are and what would be a good voltage reading?

[weblacky]

Just from a slightly different perspective, I had a SIMILAR thing happen with an Octane and a HDD I bought online (NOS). I bought an entire disk-ON-TRAY module for an Octane/Origin online (NOS). So the disk hadn't been ran in like 20 years...

I put in an Octane and it started and immediately the disk SCREAMED, SCREECHED, etc and then released and ran. I think the motor was like partially seized/hard lubricant but it stopped doing this after like 15 seconds. Oddly, it seemed to come back to life and work (so far...don't trust it much). But the Octane itself didn't falter.

My point is that yeah, the drive momentarily drew max current and winched itself into spinning. Now this isn't a short...so it's not the same as a real shorted drive (if that's in fact what you had/have) so while it might have been a contributing factor and a stressor, I do think you know have an additional problem. But I'd agree this drive incident likely hastened an aging issue.

I've not gotten to rebuilding Octane PSUs yet, but from my personal experience with them they sometimes have a faint "click" when you try to power-on but nothing happens. If you hear this, then you have we are currently calling the "click of death". It it means the PSU has gone or is protecting itself from a short or overvoltage condition. This is likely just the PSU but not 100%. You'll not know until you get a replacement PSU to break the tie.

There isn't documentation on how to bench start these PSUs yet, so outside tests aren't known at this time. SGI PSU prices have been rising lately (within the past 3 years) so it's not the greatest time to buy, hopefully we'll have rebuild services for this PSU model in a year or so. I'd say you'll likely need to purchase something in the interim but be aware that anything you purchase will be just as old.

Also while my experience is limited on Octane (just got some!) I've never heard of an SGI (correctly) rebooting on a disk issue... normally they just issue a bunch of retries and then complain on an error screen...they don't actually self-reboot. So unless someone has seen this and can correct me, what you observed was likely a PSU restart due to overdraw or other condition.

Also be aware that the vast majority of PSU death occurs on shutdown and not startup. It's just as easy for the PSU is become damaged during shutdown as startup as the process is pretty complex and needs to absorb a lot. Normally diode failure in switching PSUs seems to occur at shutdown (from what literature I've generally read) and also some PSUs have "kick starting" circuitry to specially start the PSU's processes that isn't used during "run", these can become damaged over time as well and fail to start correctly).

It's really unknown until something is changed. Octane is HARD to get into and troubleshoot so the system's modularity often works against you as you try to probe it.

However something easy would be to remove the drive itself (that you've already separated from the Octane) and flip it over and do some multimeter Diode probing around the 12V power plane and see if it it's really in short or not. That might give you more info on if the drive "triggered" this whole thing or not. Not direct info but at least forensic analysis.

[timob0]

[...]
I put in an Octane and it started and immediately the disk SCREAMED, SCREECHED, etc and then released and ran.  I think the motor was like partially seized/hard lubricant but it stopped doing this after like 15 seconds.  Oddly, it seemed to come back to life and work (so far...don't trust it much).  But the Octane itself didn't falter.
[...]
I've not gotten to rebuilding Octane PSUs yet, but from my personal experience with them they sometimes have a faint "click" when you try to power-on but nothing happens.  If you hear this, then you have we are currently calling the "click of death".  It it means the PSU has gone or is protecting itself from a short or overvoltage condition.  This is likely just the PSU but not 100%.  You'll not know until you get a replacement PSU to break the tie.
[...]
Also be aware that the vast majority of PSU death occurs on shutdown and not startup.  It's just as easy for the PSU is become damaged during shutdown as startup as the process is pretty complex and needs to absorb a lot.  Normally diode failure in switching PSUs seems to occur at shutdown (from what literature I've generally read) and also some PSUs have "kick starting" circuitry to specially start the PSU's processes that isn't used during "run", these can become damaged over time as well and fail to start correctly).
[...]

Hey, thanks a lot, this is really good information! Interesting to hear that you had similar hard-drive issues - I'd assume "sticky heads" as well in this case which however the motor was powerful enough to overcome without the machine "power cycling" itself. To me this indicates that my PSU already had issues before and the high current flow when the hdd spindle was stuck forced it into its knees. Also interesting to hear that the shutdown process itself is something can damage the PSU which might have happened in my case and with the addition to the likely overcurrent situation maybe killed it.

Did some more probing and found that indeed the PSU faintly clicks when I attach power but does nothing otherwise. It does this clicking no matter if it's installed in the machine or outside. I started taking the PSU apart and did not find any visible signs of something burnt or leaked, apart from the dust inside it looks quite good.
I read from another chap who solved the same issue by fully recapping and replacing a thermoswitch and a power switcher IC. I guess I'll follow the same path for now since it's hard to get a replacement soon.

Cheers,
Timo

[weblacky]

That’s what I’d do. If you’re experienced at doing PSU work then it shouldn’t be too bad. I’ve not done an Octane PSU rebuilt so I don’t have a component list or anything. The only advice I generally give people is that newer/different caps don’t alway equal better. By that I mean your job is to find caps the same specs, better specs may or may not cause issues.

So try to research what the caps even are now. Record dimensions, markings, maker and try to find the same size, from the same maker, same family marking or the upgraded family from the cap family tables, and try to get “like” replacements.

This helps ensure that you’re hopefully getting a cap that meets very close to the original spec and therefore will run.

Finding semiconductor damage and everything is a totally different issue. However I would recommend that you carefully remove each cap and record it in a spreadsheet and once you have all your caps out you’ll have disconnected much of the circuit so that it’s easier to either remove major semiconductors for testing or attempt to find the broken semiconductors yourself, in-circuit, before installing all new caps.

But yes, that click means its brain works and is protecting the system from bad things.

[timob0]

[..]
Finding semiconductor damage and everything is a totally different issue. However I would recommend that you carefully remove each cap and record it in a spreadsheet and once you have all your caps out you’ll have disconnected much of the circuit so that it’s easier to either remove major semiconductors for testing or attempt to find the broken semiconductors yourself, in-circuit, before installing all new caps. 
[..]

Thanks, yes my plan of attack will be:
- Give the boards a thorough cleaning
- Remove and document all the caps on the 3 boards - luckily all through hole as far as I can see
- Identify and replace all the electromechanical components - those thermoswitches and relays have moving parts that likely wear out
- test all the diodes and fets for shorts, replace if I find any
- treat connectors with de-oxy stuff
- put the whole thing together and cross fingers

Plan B would be to buy a replacement from Ian once he has some in stock.

Will report back 

[weblacky]

While that’s a solid plan, it’s also perhaps too optimistic. I’d not bother with electromechanical items because most of them are emergency-only (low actual hours) and for the main relay, if you can visually identify it, sure that one has some hours.

However there are a few, “what the heck is that thing” (like around the 24v spike interface) that I don’t think anyone has identified yet.

Don’t get bogged down on the details that aren’t really core to your issues as the reality is, you’re likely not going to start it more than 300 times in the next how many years?

So I think it’s okay to leave most of the electromechanical and passives in place as the computer is basically living out retirement where it sometimes goes on trips between long naps.

[timob0]

While that’s a solid plan, it’s also perhaps too optimistic. I’d not bother with electromechanical items because most of them are emergency-only (low actual hours) and for the main relay, if you can visually identify it, sure that one has some hours. 
[...]
Don’t get bogged down on the details that aren’t really core to your issues as the reality is, you’re likely not going to start it more than 300 times in the next how many years?

So I think it’s okay to leave most of the electromechanical and passives in place as the computer is basically living out retirement where it sometimes goes on trips between long naps.

Good point, I'll focus on the caps and the semiconductors in the first round.

Cheers
Timo