The collected works of Rhodamine

Just wanted to introduce myself, I've been lurking on the Nekochan boards for a while and finally got around to joining.

I'm based in the UK and have been interested in Silicon Graphics / SGI since 1993, when I sat in front of an INDIGO at university and was won over by their sheer style and power. Since 2002 I've been collecting, reaching about 11 machines as one point. Space constraints have forced me to rationalise to 6 machines and following a recent house move I'm slowly bringing all my machines back up.

Nice to meet you all, Adam

I've recently aquired an ONYX2, but have managed to get it into a state of confusion. I'm using a stand alone indigo2 as a console, so I haven't been able to pull off any log information to post on Nekochan. Machine is 4x300MHz with IR2 and IO6G.

Machine currently does not have an OS or harddisk, so was attempting to check machine whilst new disk was on order. Running a terminal into the diagnostic port shows a clean boot (again will try and post actual listing when I have downloaded from my indigo2 > PC).

Running terminal into TTY1 (console) on the IO6G, also shows clean boot through to the PROM menu. So based in this I set the environment variable 'setenv console g' to check graphics.

After reboot, this is where the trouble started. I get the flash screen showing 4 lines of information (screen res, Infinite reality, etc), but don't get a graphical version of the PROM menu (as I would have expected, based on an INDY or indigo2). Should I expect one?

Console shows the message 'Installing Graphics Console...' and does not return to the PROM menu. MSC shows initsaio. I've searched the Nekochan boards and notice that this problem has occured with other peoples machines, but haven't found an obvious answer. I waited over an hour to see if this message would time out, but it didn't.

I removed the keyboard and mouse and rebooted in an effort to regain console control, but no luck. Still tried to bring up the graphics console and doesn't give access to the PROM menu, without access to this menu to change the environment variable back I don't see how to force it back. Any thoughts?

To understand if there were any hardware issues I entered POD mode (not sure what the best way of doing this is, there isn't a 'pod' command on the MSC interface, but forcing an 'nmi' seems to get me there, but does through up some exceptions - I'm I doing this right??) .

Followed the 'go cac, clearalllogs, etc' route. Should it take a long time to perform the 'go cac' comand, takes about 3 minutes on my machine, is this normal. 'clearalllogs' also caused an exception the first time around, but after a few repeats it managed to work. Subsequent reboot didn't show any messages in the log file.

At the moment I'm at a loss as to how to get the console back, is there any way I can reset environment variables through POD?

All thoughts would be greatly appreciated, Adam

In Bristol, where abouts are you?

very good, made me chuckle.

In the great scheme of things that's pretty close - just a quick jaunt on the M5.

foetz wrote: first try it pure: viewtopic.php?p=129565#p129565

then this one: viewtopic.php?p=129514#p129514
in particular the last paragraph there.

Useful threads, many thanks. I agree, keep it simple is definately the way forward.

On a practical note, is there a prefered way to remove cards. At the moment I release the CPOP connectors (if used), then chassis fastenings, and slide the card out by an inch or so to ensure all connectors are dis-engauged, but don't actually remove the card. My logic being that the card and connectors are less likely to be damaged either physically or electrostatically by doing this. What does everyone else do?

I will post updates as I progress - hopefully will be informative for the next ONYX2 newbie.

foetz wrote: and welcome here btw

Many thanks for the welcome, Adam

Sounds ominous...

The front stages of PSU's tend to be perminently powered, even when the computer is switched off - this is to allow electronic control of the PSU. Only unplugging the PSU from the mains/wall socket or using the mechanical switch will de-energise these stages. There is a lot of electrical stress on the PCB and these front end components, which could result in strange sounds (crackings, hissings, clicking). What you are hearing is probably the result of the PSU 're-stressing' as the mains voltage is re-applied.

In my experience these things rarely get better and ultimately lead to some form of catastrophic failure, but it is sometimes worth getting a vacuum cleaner out and removing as much dust and debris from the PSU as possible, as this all increases the probability of electrical breakdown within these high stress stages.

Cheers, Adam

hamei wrote: The power supply itself could explode, but as long as that little interface chip is okay there's probably a hundred dead Fuel power supplies to cannibalize.

Very true, as long as the PSU has decent crow-bar protection and doesn't blow anything when it goes, then they are probably one of the 'easier' parts to replace/fix.

I'm not familiar with the specifics of Fuel PSUs or their failure mechanisms, but sounds like a bit of a design flaw.

Cheers, Adam

I bought a dual R12k 300mhz octane processor about 10 years ago, it was stupidly cheap. It's been sat on the shelf ever since. About 6 months ago it got damaged when it got dropped, my own stupid fault. There is a small region on the corner of the main PCB which is damaged as shown in the picture.

I recently aquired an octane with a dual R10k 195MHz processor and wondered if it was worth trying the damaged one, or is there a strong chance I would damage the rest of the system.

Do I just cut my losses and bin the damaged item?

Cheers, Adam

Geoman wrote: remove that chipped edge or bend it right and try it out.
these modules are ridiculously robust. the worst thing that could happen is octane not starting at all.

Thanks for the advice. Wasn't really able to clean it up, even bent PCB is pretty hard. Plugged it in and all powered up and reported in hinv. So looks like it is working. Thanks once again.

Regards, Adam

Went to fire up my indigo2 r10k impact today, and nothing, dead. Suspect the PSU as this seems to be a common problem on this forum.

Have removed the PSU, but would like to do some diagnostic tests first. Does anyone know the cable pins outs or know a link to them? If it does turn out to be the supply, first stop will probably be to replace the electrolytics!

Cheers, Adam

chicaneuk wrote: I swear the IMPACT power supplies are the flakiest crap!

Yep, must agree

So far I've stripped it down and inspected. Input fuses are fine, no burst or leaking caps, and no obvious sign of damage. I reassembled and put it back in on the off chance that the strip down might have reseated any connectors, but still no joy.

At the moment I need to find out how to turn the PSU on, the equivalent of grounding the green wire on a PC PSU, but can't find any info.

Regards, Adam

Solid/impact. Currently the PSU is off, actually in pieces. Normally I keep it turned off at the power socket when not in use. But it has spent the last week with the power socket on, maybe that explains the problem.

PSU is zytec 060-0027-002 rev A, 229435507 rev D. Suitable for dual heading, so had assumed it would be so over specified that it would be robust. Guess I was wrong

I have taken on the challenge of repairing, currently trying to reverse engineer a schematic for the 'high voltage' board. Seems to be a more recent model than the one on Ian's website, as now has connector on the high current outputs from the transformer.

Trippynet wrote: Best of luck with it! Anything that helps aid our understanding of these and how to fix them in future has to be a good thing!

Thanks, will post schematics if/when I manage to generate them. Will be a case of doing it when I can. Got the first 10 components down, another 290 to go!

This is the newer version of the hv board for those that are excited by such things.

Regards, Adam

Hopefully it is just the caps, I will certainly be replacing them. But, I think a better understanding of these old supplies will not go amiss.

Modern electrolytics have improved significantly since the IBM issue days due to better sealing. Unfortunately that doesn't help these legacy systems.

Cheers, Adam

Info plate from the 060-0027-002 rev A. Obvious difference is the +3.5 line to the gfx is pushed up from 36A to 48A.

Regards, Adam

Just a quick update. Started to reverse engineer the hv board. Yes I know it is probably overkill, but being curious I just need to understand this PSU.

The PCB is a 4 layer FR4 construction, so does pose a bit of a challenge. Through hole components on the top layer and surface mounts on the bottom.

To help generate a net list, I sent the board to be x-rayed. Useful info, but there are so many large caps and wire wound components (with flying leads) that a lot of the detail is obscured. The big heat sink doesn't help either.

So currently striping off the large components. Will also allow me to check the suspect caps. Once this is done I'll send back for more x-rays.

Slow going, managing about 30 mins a day...

Regards, Adam

ivelegacy wrote: what about reengineer the whole PSU from scratch?
having requirements, constraints, acquiring solid skills with 400Watt PSU

Might end up being the case, but requires some understanding of what the current on is doing. Currently don't know the start up sequence.

Regards, Adam

ivelegacy wrote: What about dynamically measuring all the currents and voltages which are flowing throught the power connector into the motherboard? (which should be functionally equivalent to a black box " black box "), with programmable multimeter?

This is a good strategy, unfortunately my PSU is dead, so can't black box. Current plan is to repair, but repairing requires fault finding and fault finding requires understanding of expected response.

If I get it working I will do some black box testing. I have access to several 4 channel giga-sample scopes (not mine, but I can borrow), so would be possible.

I've built switch mode supplies up to 100W, but would struggle to get something anywhere near as fault tolerant, robust or compact as a commercial unit.

Regards, Adam

recondas wrote: I'd suggest re-installing all of the boards and using the infinite reality diagnostic, irsaudit , to see if a little more light can be shed onto the issue

I can second this, I have used irsaudit to diagnose an issue on an onyx2 desk side in the past. Display had classic mosaic fault and would then fail completely after a few minutes. Connected in through a terminal and narrowed it down to a faulty RM board. If only all the SGI sub-systems had such thorough diagnostics!

Cheers, Adam

I work in a company where ESD is considered so seriously that every employee is given compulsory ESD training with a refresher every 3 years. This includes the secretaries that will never set foot within 200 miles of our production site.

ESD can cause latent damage, which reduces the lifespan of a component, so just because it still works doesn't mean it hasn't been damaged. Also some ICs are so sensitive that they can be damaged by proximity, you don't even need to touch them. Proving ESD damage without electron micrographs of the IC die is almost impossible, so incredibly hard to prove.

Having said that, I agree with Vishnu, these are large card assemblies and probably have massive capacitance which is going to offer a degree of protection.

Regards, Adam

jan-jaap wrote: What would be truly awesome would be some sort of module design where you can connect a modern (USB) keyboard on one end, and it emulates an SGI "4D" or PS/2 interface on the other. It could probably be done using an Arduino or Teensy. Cost isn't even my motivation, I just want something that works, and works reliably, with a KVM so I can eliminate the zoo of keyboards, mice and glass tubes at my end.

I've been thinking of something similar myself. A modular KVM, based around a eurocard cage. The back plane interfacing to a standard monitor, keyboard and mouse setup. Each SGI would connect into the back plane via an interface board. The interface board coming in different flavours, a generic for more standard systems and more bespoke for the legacy machines such as indigo etc. I would hope it would allow switching between at least 6 systems and being bespoke shouldn't suffer from some of the issues that other people have found.

Great plan, do I have the time to actually spec, design, prototype and debug... Nope. Maybe one day

Cheers, Adam

Eeek, I would have thought the metal chassis would be earthed. Sounds like there is an earth issue, I would be concerned about sorting that out first.

I used to have a washing machine that felt fuzzy when when you touched it. Turned out to be an earth fault and the whole chassis was live, saving grace was the thin coat of white paint!

Cheers, Adam

I'm a physicist with 30+ years experience coding. Mostly fortran77, various forms of basic and the last decade a language called IDL, but no formal teaching in coding. Just self taught starting with commodore basic on the VIC20.

I recently asked a software engineer friend to streamline some code at work, had been cobbled together over 8 years with various authors and constantly changing requirements. The code that came back worked fine, but I couldn't understand any if it, was all pointers and objects.... Not scientist friendly

Regards, Adam