SGI: Hardware

Folks,

today I got my hands on a Power Challenge L Deskside in near mint condition. The previous owner made a clean shutdown but left the machine connected to the mains in order to avoid problems with faulty capacitors. When I entered the machine room in order to grab this gem, the Challenge showed a POKB-Error in the System display. Not too good I think. We switched the circuit breaker on the back of the system to "Off" and I hauled the machine over to its new location. After reassemby of the FLDs (three HDDs and a CDROM), plugging in the power connector, switching the circuit breaker to "On" and turning the key to the "I" position nothing happens. Absolutely nothing. No fan noise, no LED's, no System Display. The PSU seems to have some Indicator LEDs for DC and AC - both dead. I checked the cable, it's OK. I'm new to the Eveready machines so please apologize for some noob questions: Did I forget anything in order to power this machine on? I checked the Challenge Owner's Guide but I didn't find any additional steps. Is there a chance that the PSU died after showing this POKB-error and disconnecting it from the mains? I think dis- and remounting the PSU could be worth a try but I can't find any hint how to remove the back cover of the system. Again, apologies but I checked the Challenge related posts on Nekochan for some advice or a link to a SGI document but to no avail. It would help me a lot if anyone here would be able to provide me a link to an appropriate maintenance guide or if anybody would be so kind to point me in the right direction.

Thanks!

Cheers
HDC

The PSU fan should spin up as soon as the circuit breaker is closed. The back plastic panel is removed by pulling straight back - it will snap off and the plastic clips are fairly robust. The screws will then be obvious. You can poke at the circuit breaker and mains lines going into the PSU (sometimes called off line switcher) with a multimeter.

Once you get the OLS working you can then track down the POKB error. POKB means a 1.5v or 3.3v regulator has faulted... but I would only worry about it if/when it happens again. (POKA is 5v or 12v fault). More details here http://www.talisman.org/~erlkonig/misc/ ... l#POKAPOKB

This may sound dumb, but are you sure that the breaker that feeds what you are plugged into is switched on? The ofline supplies are a set of single phase supplies so one should have come up. I have the manuals for these and will get some info on the POKB power issue as I suffered with the same failure on an Onyx. It was from a damaged IO4 board IIRC.

Were any boards moved/ removed before you picked it up?

Here is a good start from sgidepot
http://www.sgidepot.co.uk/chalonyxdiag/ and I have attached the pdf.

and the users manual....

http://techpubs.sgi.com/library/tpl/cgi-bin/summary.cgi?coll=hdwr&db=bks&docnumber=007-1732-050

I have several paper manuals and the hardware developers manual if you need more info.

John

Cracking! mopar5150 an guardian452, you guys made my day. Thank you. I'll post an answer to your suggestions asap after breakfast with my familiy.

Cheers!

HDC

Ok, here's an update regarding my attempts to resurrect the Power Challenge:

The machine is still dead. Sigh.

Thanks to the kind support I received it was no problem to remove the back panel and to check if the OLS was correctly seated. I think there's no problem with the physical connection of the OLS to the backplane.
After that I poked the cables at the power receptacle unit with a multimeter. The circuit breaker works, no beep if switched to "off" , beep if switched to "on". If the switch is in the "on" position I can measure 230V at the electrical contacts wired to the cables which go into the backplane.
I have no idea how the connectors from the backplane to the OLS are wired. Anyway, 230 V is routed from the receptacle through the circuit breaker into the system and nothing seems to reach the OLS because there is no sign of activity from the OLS. No Fan, no amber or green LED. mopar5150, is there a chance that you have some infomation about the wiring of the backplane? My next step would be to check if I can detect any voltage on the input connectors for the OLS. If there is any suitable voltage on these pins I think most presumably the OLS is dead, am I right here?
Does anyone know what I can check in addition?

BTW, I removed all the boards (IO4, MC3 and the IP25) and reseated them.

I've seen many failing SGI power supplies, but most of them died with at least some olfactoric effect - I suspect that there is something wrong with another component that I haven't taken into consideration, maybe associated with the transport. The machine was alive just prior to the transport and is totally dead after the haul. And I hauled it with all loving care I could give, believe me. So, if anyone has an idea what can be done to bring this wonderful machine back into production, please let me know.

Again, thanks for all the support.

HDC

The POK(x) errors are board-level, so try pulling boards to solve that. The system controller will shut off the system on these errors, but you should still be getting system controller power and the +48V lights, so look at your OLS module(s).

Check this out if you haven't yet: http://www.futuretech.blinkenlights.nl/chalonyxdiag/

I have never had my OLSes open, so I"m not sure where the ACOK tap is.

Yes, that's what should happen. I was able to track the electricity upriver to the wires ending in the internal connectors of the backplane. I checked the diag document but was not able to find the wiring of the connectors - I'm not very good with such things as circuit diagrams , so if anybody is able to tell me what should arrive at the pins I can track this down. Without this information I think the risk of frying the wiring, the backplane and maybe myself is a little bit too high.
If anybody did a total disassembling of a Power Challenge: Is there another connector between the cables running from the receptacle to the connectors of the OLS? I can't observe the complete pathway of the cables and I didn't find enough time yet to disassemble all the brass inside the chassis and I have the (admittedly weak) theory that the problem is not the OLS but a loose connection due to vibrations during the transport. BTW, if anybody is interested in some photos of the cabling (for documentation purposes) let me know, I can upload them.


I tried to inspect the internals of the OLS (the Power Challenge has only one OLS) but I'm still missing an idea how to open the OLS. The brass nuts are leveled with the housing and not accessible with my set of tools. Maybe I shouldn't try because the only thing I could do is an optical inspection and tinkering with such monstrous PSU devices can be painful if one does not fully understand the technology.
Anyway, I found a spare OLS for a fair price at a German reseller and ordered it. Most hopefully it works and I can start to find out which of the boards throws the POKB-error. How to get a spare part for the defective board will be the next challenge I'll post an update if anything interesting happens.

HDC

The power goes like this:

1) power inlet on the back side.
2) breaker switch right next to it must be in the 'ON' position. Sometimes, the lever off the breaker has been broken off, and it isn't obvious anymore that there's a switch at all.

The assembly of power inlet & breaker switch can be easily removed from the system for inspection. This assumes that the black panel has been taken off the backside, of course. This is easy too: grab the plastic at the bottom end and pull firmly.

From the breaker, 220VAC cables go straight to a tiny backplane that the OLS slots into. If you remove the metal panel on the right hand side you can see it:


The OLS itself has two connectors: 220VAC in, 48VDC out. No switches. The OLS can also be easily removed from the system.

IIRC, there are two LEDs on the OLS: 220VAC OK, and 48VDC OK.

With the breaker in the 'ON' position, 220VAC will go to the OLS and you should hear a fan humming inside it. If you then turn the key on the front panel to the ON position, you will experience the full force of the massive blower in the bottom of the chassis After a couple of seconds it should calm down -- a bit at least. Until the ambient temp rises and the system controller revs the blower in order to keep the system from overheating.

Oh, one last thing: the OLS of these systems were recalled at some point because they were a fire hazard . The 'safe' OLS is part# 080-0120-001. The part# is on the OLS. If you're going to have a look anyway, better check that as well.

Use a continuity tester to figure out where the mains go into the OLS. (i.e. without being plugged in). If you still need help figuring this out, I can post pictures or even a video on youtube, maybe by the end of the week. It's somewhat annoying that this is the one piece of the system (OLS connections) I don't have any pictures of, and I haven't had the ass-end of the machine apart in close to 5 years. IIRC my Onyx has the 2400w OLS because of the IR graphic but I believe the connections should be identical (it's been a few years since I had it apart, although it was up and running quake and bzflag just the other day they are fantastic machines... sadly the only monitor I have (at the moment) is an old 27" sony trinitron TV. Great TV, but the resolution is low for a computer screen )

FWIW I run mine on 120v, there are pros and cons either way, I've run it on 240v before but not recently... EU doesn't have the double-tap breaker panels like here so good luck with that

Running an Onyx IR off a regular, residential circuit shouldn't cause any problems at all in any EU country. No hacks required. We Europeans don't believe in 110V

I have two dedicated 16A 220V circuits in my new computer room. Each has their own integrated RCD so they can't take out (part of) the rest of the house. These are 'slow' breakers, so I can push them back in without having to disconnect half a dozen systems first to reduce the switch-on surge. That's a bit more extravagant than the other circuits in the house, but hey I have a third (220V 16A) circuit for the air conditioning, and the lights are on an existing (fourth) circuit. It's pretty common over here to put anything that consumes 'significant' amounts of electricity on it's own circuit. I think I must have nearly 15 breakers in my panel, not counting RCDs and the primary breaker.

I probably have at least 12 breakers in my panel here and I have just a 2-bedroom apartment. like, the clothes dryer (240v) has one, the oven (240v), a/c (240v), etc (120v), all the receptacles in each room(120v...), the bathrooms, etc, etc... it adds up.

The vast majority of distributions we did (in canada) were 600/347v. In the US it is all 400/208v, along with most of the equipment in the datacenter. so pretty much everything needed a transformer at some point. It's also not uncommon to have a 600/400 transformer, then the ups (from the US of course), then a 400/600 transformer

I never had to do anything with the residential breaker panel in the EU, but it looked like a regular single-leg 240v breaker panel to me. Of course.. 'not at home' was a different story

Anyways, I can't say for sure if it is better or worse for the OLS to be run at 120 vs. 240v. The power rating on 013-1592-002 is 2400W at 200V but only 1500W at 115V, if there is multiple RMs, for example, the system requires 240v. Theoretically, the lower current at 240v will put less stress on the primary, but if there is an insulation breakdown due to aging, it might be nasty even if it otherwise runs fine on 120v.

The only reason that the US has two breakers is because the center tap of the distribution transformer means that you can get 120V on either side relative to neutral/ground (and therefore current), so both sides need to be protected. If you have 240V without the center tap you only need one breaker because the other side would be the "ground/neutral" (similar to the way the US does 110V).

Folks,

thanks a lot for the vivid discussion and sorry for the delay of my answers. My jobs keeps my so busy that I hardly find time to visit Nekochan at the moment. I'm in my office right now and stealing some time to contribute.

@guardian452:
Yep, thx. Should have come to that myself. because I used my multimeter to beep the wiring from the receptacle to the circuit breaker and from the circuit breaker to the cables going inside the machine.

Wow, thats really nice. But let's wait for the spare OLS to arrive, maybe it's not longer necessary to do such a deep-dive.
Anyway, just for learning, the only thing that I have to measure is IF there is a connection from the mains coming from the circuit breaker to at least 2, presumably 3 (phase, neutral, gnd) of the pins of the connector, right?
Would be nice to know where the current arrives at which pin anyway, so if you want to check this and publish your findings I would really appreciate that just for documentation purposes. I'll do this as soon as I have a working OLS in my machine.

@jan-jaap
Thanks a lot for your efforts. Last sunday, during one of my attempts to get the fat lady running, I dismanteled the side panel and found the metal panel . Was interesting to see how SGI divided the power to the FDA and for the boards. As afrore-mentioned, I suspected that there could have been another connector between the mains coming from the circuit breaker and the tiny backplane the OLS slots into which perhaps was loosened because of the transport . I didn't catch that this could have been easily checked through beeping the connections. Again, shame on me.
BTW, no 400V 32A circuits in your house? - whats with the big iron coming your way?

I can't wait to get the spare OLS, most hopefully I arrvive at home early enough to do a test-run. Unfortunately the beast is residing in our living room (I have a wonderful family which accepts that I have a room- and electricity consuming hobby. Well, to be honest, my wife tries to accept...) so any attempt after 10 p.m. can be dangerous because it could wake the brood .

The former owner just told me that the machine only runs for about 5 minutes and then shuts down because of the 3.3V POKB problem. I got the machine for free, so no need to get upset about that. Am I right that it is not feasible to repair a board with a faulty 3.3 V regulator? I have only one IP25 board, one MC3 and one IO4, so I think one of my next postings will be in the "Hardware wanted" forum. Does any of you have some spare boards for a Power Challenge and is willing to sell/trade?

Oops, 8 a.m already. Have to get cracking! Thanks again to all contributors. Nekochan is a really cool place!

Cheers
HDC

Only tall racks require three phase. That's why I have desksides Saves an awful lot of space too, and IMHO a rack doesn't add anything except more of the same. I have one tall rack (my 4D/380VGX) which is rated at 32A @ 220V, but that's excessive because the PSU is around 1700VA. Runs fine on a regular 220V circuit.



LOL, sounds familiar

I was restoring the system here, it was 'naked' except for the front door which was only there to protect the system from my son and vice versa. If it has lights, buttons *and* daddy is focussed on it, it automatically has irresistible attraction on small kids

I have spare MC3's and probably an IO4, but I'm pretty sure neither of them have 3.3V regulators on them. I'm afraid the IP25 is the only one ... (and Infinite Reality, but this is not an Onyx). A fault LED should show you which board asserts the POKB line.

I pulled the OLS out of my and took some pictures. I have an Onyx, but all "everready" desksides (including challenge L) should be the same. The mains voltage comes in through the connector on the left (looking at the back of the machine), and the DC and signal comes out the right. The top pin is "hot" (what it is called in north america), the middle pin is "neutral", the bottom pin is "earth". With the breaker closed, all 3 should be a closed circuit all the way to the end of the cord that plugs into the receptacle (should check out the cord too ) I'm adding some extra pictures for posterity...

Here's the other side of the OLS:

Here is where the 48V and OLS signal goes into the system backplane:

Here's where the power for the SCSI devices comes out, as well as power for the fan and the connection to the system controller interface:


To remove the large panel covering the backplane and system controller (and whatever else is in the back of a challenge L, I honestly don't know. If you want to provide a picture that would be interesting for me). Anyways, it helps to not remove the two screws on the bottom, just loosen them halfway so they can prop up the panel once all the other screws are out.

(mine is also in the living room but I live alone, at least for now)
It sure would suck if the POKB is caused by the IP25 It could also be the system controller.

If you want to hear what the healthy system should sound like, you can check out this video here http://www.youtube.com/watch?v=W-iQ_AqEJjk (this is my old youtube channel so don't bother messaging or commenting, my current channel is /themightynutmeg )

Here is the news: Got the spare OLS yesterday. Swapped OLS's. Good news: Old OLS was dead. With the spare part the system turns on again. Bad news: Immediate POKB error. Worse: As predicted by jan-jaap and guardian452 it's the IP25. Man, this sucks!

Anyway, just to add my contribution, here are some pictures showing how the setup is for a Power Challenge L:

This is how the inside of the receptacle unit wiring looks like:

A view inside the case showing the OLS connectors:

An overview of the backplane:

The backplane with the connectors the OLS slots into and the connectors for FDA and fan PWR:


I can't provide any pictures about the back end of the board chassis because I'm not allowed to :


Next thing will be another attempt to reseat the IP25 board but to be honest, I fear it's FUBAR.

I'll post an update if anything interesting happens...

<edit>
Tested IP25 in another slot: Red lights on
Reseated IP25: Red lights on
Crap.
</edit>

<edit>
Without the IP25 board the machine resumes the boot process. But, well, without a CPU board the Challenge is not much fun. It's definitely the FW300F1 DC-to-DC converter (bottom LED red on the IP25). I've digged through Google but nowadays these parts are hard to find. So, if any of you has a FW300F1 (or a board with other failed parts but with a known good FW300F1) or a 030-1072-001 Rev A-G board and is willing to sell/trade, let me know. I know that Ian Mapleson has some of these in stock, but with prices beginning at ca. 740 EUR or 900 USD this is far beyond my possibilities.
</edit>

I will look at the other boards and see if this converter is used on any of them. If not I could sell you one out of my onyx as I only use 10% of it anyway.

FYI: I've tried to de-solder a component from an IO4 once and even with a professional desoldering station is was absolutely not happening. Those PCBs are simply too thick, you will destroy the traces on the top layer before the inside gets warm enough.

That's what I'm afraid of, too. What I was thinking about was to remove the CPU from the IP25 (although I didn't find the time to check if this task has it own risks) try to de-solder the FW300F1 (I have acces to a professional SMD desoldering station and the owner of the station offered his assistance) and to solder the new one in, reinsert the CPU and give it a try. The board is defective anyway, the only problem could be that I fry the CPU or anything else because of internal havoc caused by the soldering.


mopar150, did you mean that you are thinking of selling me a surplus non-IP25 board with a FW300F1 so that I can try to de-solder it? That would be pretty cool, thanks a lot. But AFAIR the FW300F1 is only used on IP25 boards. Is there any chance that you're willing to sell an IP25? If swapping the CPUs would be possible I would be very happy with a naked board w/o the R10k. I think that you wrote "the other boards" on purpose because you don't want to part with an IP25, but it doesn't hurt to ask. Shipping the whole thing to Germany will be another question, I think.

Cheers
HDC