[weblacky]

So I’m at my planning stages of starting SGI PSU/PCB work, and I’ve kind of hit a small decision snag. 

I’d like group input to know what’s fair and expected here.

Let’s assume I’m working on an SGI PSU that kind of works, just likely needs new caps.  So you end up doing the caps, seems to work fine…do you move on to the next unit or do you take the time to also replace the secondary diodes, FETS, etc because they could fail?

I can subscribe to both the philosophies of “if it ain’t broke don’t fix it” and “while I’m here, I might as well take care of everything, so I don’t have to open it again”.  But these of course can be at odds.

Now I know what some people are saying, if it’s yours, do whatever you want. Okay, fair.  What about if it’s not going to be mine in the future?

I’m unaware of any documented semiconductor lifespan (outside of environmental factors), so I really don’t think a properly size diode or FET wears out.  But maybe it does?

Let me be clear, I’m not talking about you sending me a PSU and you paying for a fix job (though I might entertain that mode of operation once I have more experience and parts on hand, to have a fast turnaround). I’m talking about selling you a PSU that works and has had all its caps (even the huge filters) replaced and all the parts appear to work as of time of sale.

I may accept trade-ins as money off another PSU (like $75-$150 off, based on model).  But a refreshed Tezro PSU would be like $450-$500, as the PSU itself (as NOS) would be like $250-$325) and that is just an old PSU that’s not had it’s bad passives replaced.  So take the entire unit and fresh/fix into account.

So that’s the question for any of you in these relevant industries.  If you’re rebuilding an item, and really you want to do work on once, and sell/hand it off to another person (who I’m sure isn’t interested in coming back to get this stuff “fixed” in less than like 7-10 years) . Would you remove fully working parts (semiconductors) and replace with modern (possible higher-specked) replacements and chance that these new parts won’t cause their own issues.  Or would you “fix” the problem the PSU has and clean it and move on, like a car mechanic, leaving old and working parts in place?

This sort of sits in the “used car” sales sort of question, they are used, they’re working because they were fixed…but if you don’t preemptively replace a lot of stuff, failures from age may happen in a few years because that part wasn’t broke when it was with the dealer.

What sounds fair to you all?  Should I speak in terms of “tiers” of rebuild? 
Like “Fixed/refresh” (caps and broken parts) vs “fully rebuilt” (refresh + parts that MAY fail in the near future) versus “remanufactured” (replaced vast majority of parts plus stripped and re-applied electroplating to the case with new stickers/labels)?

Obviously, prices would also reflect this effort, but it doesn’t seem fair to charge more for one PSU that had more work involved, and less for the same PSU with less work needed (assume both PSUs are identical).

Given the required removal of heatsinks and glue and such, replacing the semiconductors is much more work.  What would you all expect?

Should I just give some prices and do a group poll?  I can understand how someone wants to pay less for a repair job and says, “just fix it”, they own the item after all.  But a fixed item being sold to a new owner…I’m unsure.

Thoughts?

[Raion]

So as far as your pricing goes, it's a bit too high for me to consider it. If only because I anticipate that Elf's PSU replacement for the Indy will be cheaper. For something like an Origin or Octane, 400-500 I could /possibly/ stomach. But not the Indy. I get there's a lot of guess work that goes into this, as well as time, equipment etc. I respect that. But I feel at least I represent a normative member of the community - I have a limited income to spend on the hobby. I can't spend thousands on machines, nor hundreds on repairs.

TL;DR: You've priced me out of it, and I suspect others too. I was gonna do O2 PSU recaps eventually, and for that I was gonna charge $100, including an alcohol bath to clean the parts.

[weblacky]

Raion: I think you misunderstood, I specifically said:" But a refreshed Tezro PSU would be like $450-$500, as the PSU itself (as NOS) would be like $250-$325."

So that pricing would be for a Tezro supply, as stated. I stated no other pricing, I'm unsure where Indy and $400-$500 met?

A rebuild Indy PSU would be like $250, given that I tend to get them for about $50-$70. But they'll likely need professional plating, so that's not cheap. Nearly all the Indy PSUs I've gotten have minor to serious compromised galvanized coatings. Those cannot be left as-is. Indigo2 PSU don't appear to get that as much.

Also I'd still like input on the original question, perform semiconductor overhaul on rebuilds or not? Still working parts = still good?

[robespierre]

I recommend only changing parts that are faulty, and only changing one of them at a time.
If you get too ambitious, the risk is that you end up with something that doesn't work, but you can't point to the component causing the problem.
It's like a regression test: always check if the last thing you did got screwed up.
Of course, there are times when the failure of one component is caused by another, so both must be replaced at once. I'm talking about independent components.

[weblacky]

Yeah, that's really the cusp of my worry. If the parts work, why change them. But I'd be aggressive in the caps, in consumer electronics (not like industrial applications) I don't see why a 20+ year old cap is going to be my friend. But yeah, I understand the concern.

My plan was to be sort of be extra cautious and actually research the OEM caps, find the family model info, then use the manufacturer's chart/table to find the recommended upgrade/replacements to closely match the old specs. As you say, I don't want to knock the circuit off by radically changing the ESR/ESL from designed values or the like. So I'm not going to get "random" high-end caps, I'm going to try to mimic the old (fresh) ones. I'll get the natural evolution of the same cap from the same manufacturer family line (if possible).

As long as they are OSCON, Nichicon, and the the like...I'll do fine tracking down the recommended part update tables from the cap manufacturers and planning to match them, cap for cap.

I've been told this approach is the best way to ensure continued operation of amplification, radio, and power switching electronics that may rely on the caps themselves for the actual oscillation of the circuit. So the "new" caps need to (as close as is reasonable) be similar to the old caps specs (where is counts). Lower ESR isn't alway better in all caps, I now know this. However...I might experiment with higher ripple tolerate caps for the output filter caps after a few successes to see if I can slightly improve the long-term outcome of the the PSUs, but I think that's the extent of my fiddling based on my limited knowledge about these designs.

Thanks for contributing!

[Trippynet]

I think it depends to be honest! When I had issues with my Indigo2 PSU, I sent it to a firm here who re-capped it and sent it back. It continued to behave erratically, so I sent it back to them (their repair came with a 1-year warranty). They then found 4 resistors that weren't matching reference values so changed these and sent it back.

It was better after this, but still not perfect (lasts maybe 90 minutes then trips out). I tried sending it to them again, but they'd had enough so they declared the PSU as a write-off, refunded my money and returned the PSU to me. Disappointing, but on the plus side I have a re-capped supply and it has only cost me a few lots of postage, even if it is still flaky.

So my thoughts are, don't replace components if they aren't broken, but it's highly advisable to be able to stress-test the PSU you repair to make sure everything works. That was the problem with the firm I used - they could re-cap the PSU and test components, but they didn't have a way of actually running the PSU themselves to test it when it was live and under load.

[weblacky]

I think it depends to be honest! When I had issues with my Indigo2 PSU, I sent it to a firm here who re-capped it and sent it back. It continued to behave erratically, so I sent it back to them (their repair came with a 1-year warranty). They then found 4 resistors that weren't matching reference values so changed these and sent it back.

It was better after this, but still not perfect (lasts maybe 90 minutes then trips out). I tried sending it to them again, but they'd had enough so they declared the PSU as a write-off, refunded my money and returned the PSU to me. Disappointing, but on the plus side I have a re-capped supply and it has only cost me a few lots of postage, even if it is still flaky.

So my thoughts are, don't replace components if they aren't broken, but it's highly advisable to be able to stress-test the PSU you repair to make sure everything works. That was the problem with the firm I used - they could re-cap the PSU and test components, but they didn't have a way of actually running the PSU themselves to test it when it was live and under load.

Yes, I remember that.  That’s a concern as always.  I’m planning to ensure I have a spare junk system to do like days of on/idle with bouts of activity (like maybe network-based OS loads) to try to test a PSU in the intended system for several days (like 3?) straight. 

But your scenario is the reason I don’t want to do “fix” jobs, because to be really sure, I need to put days of time into a PSU (including testing routines).  That’s not a schedule I can keep.  It’s easier to buy them, fix them, then do extended tests to feel confident about it.

Of course there is the issue of something being burnt out or wrong with an SGI, I don’t know if the older SGIs PSUs have good overcurrent protection or not.  Another worry of mine, the new owner burns out the PSU with a bad mainboard and demands their money back.

I’ve been trying to assemble odd test gear that I hope will give me an edge in finding issues.  I’ve recently (of last week) finished my basic Indy PSU (to mainboard and to PSU) external interface cables for my curve tracer.

   

   


I’ve haven’t sat down and programmed a test routine yet, (that takes planning) so that still needs to be done to collect readings from all external connections.  Then I need to plan out internal connections to test/probe using another harness.  I’m not an EE so I only kind of understand the high-level stuff (hopefully).  I’m more afraid of the large analog logic regions on the older boards.  Not everything is ground-facing for easy probing.  Nearly all the old PSUs have a separate daughter board that comes off the main board that carries the “smarts”.  I think power-off testing will do the trick, but as to what and where to hook, I’m unsure at this time.

My hope would be a to get to a point where all ICs will just use test clips (have quite a few of those now) and test in-place and all resistor networks will have to be manually probed between neighboring IC or ground using common IC lead/ground and hand probes.  I just need to find a way to hit all the big unique spots.

For the resistor networks, I may have to literally desolder each leg and do a massive individual test, unknown at this time.

There have been cheap overseas dc load testers appearing on eBay lately that eevblog claims are pretty good.  They’re about ~$300 a piece for dual channel, 200w per channel, units.  I’ve just put in an order for one ET5420 so at least I’ll have that for 5v and 3.3v loading while doing basic start testing.

I don’t have good info on bench testing PSUs, like ramp up current and sudden changes in current draw.  But these units can do basic stuff like that without a computer program. I also don't own a really good oscilloscope, just a scope meter and a small uScope meter.  Don't need one right now to justify the very high cost for the speed I'd need to actually use it for ripple.  So we're sort of doing a lay-persons way right now, checking for AC V ripple being under 1% of DC V using a fluke multimeter.

I want to be sure, I know I’ll never be 100%, which worries me.  But yeah, testing will be longer than soldering.

This dives into my pricing question because if it takes me 35-50 mins for soldering once you know what you’re doing.  But the clean-up, alterations, and testing take longer.

One big thing with Indy PSUs is the logic daughter card is held on and to the case with glue (not fasteners).  You have to break this glue to get it out…so I’ll have to redesign the interior to hold the PCB unless I want glue everything (which I don’t).  I’m thinking of using some plastic PCB guides and attach a long one to the interior top of the case (with countersunk heads).  So the new plastic rail stabilizes the daughter board once assembled and keeps it rigid.

Also as I mentioned in my above post, corrosion, most of these PSUs have corrosion and require new coatings, or galvanized zinc spraying, or plating.  Given I’m not a photoshop expert and making vinyl labels isn’t my thing, I’m worried about producing the PSU labels if the cases needs that kind of work.

So yeah…lots of stuff…not just replacing 5 caps.

[Raion]

Raion: I think you misunderstood, I specifically said:" But a refreshed Tezro PSU would be like $450-$500, as the PSU itself (as NOS) would be like $250-$325."

So that pricing would be for a Tezro supply, as stated.  I stated no other pricing, I'm unsure where Indy and $400-$500 met?

A rebuild Indy PSU would be like $250, given that I tend to get them for about $50-$70. But they'll likely need professional plating, so that's not cheap. Nearly all the Indy PSUs I've gotten have minor to serious compromised galvanized coatings.  Those cannot be left as-is.  Indigo2 PSU don't appear to get that as much.

Also I'd still like input on the original question, perform semiconductor overhaul on rebuilds or not?  Still working parts  = still good?

I apologize if I misunderstood. 

"I may accept trade-ins as money off another PSU (like $75-$150 off, based on model).  But a refreshed Tezro PSU would be like $450-$500, as the PSU itself (as NOS) would be like $250-$325) and that is just an old PSU that’s not had it’s bad passives replaced.  So take the entire unit and fresh/fix into account."

I mistook this as you quoting how much it'd be to recondition a PSU. Retail pricing, that's probably /okay/ and someone would totally pay for it. I would probably not. 

As for "professional plating" I'd probably waive that and just use parkerizing. I can pick up a beadblasting cabinet for cheap and use the same process I'd use for a gun to parkerize it, making a coating that while not OEM, would be far more resistant to rust and blemishing/corrosion. 

If we agreed on a price though, and you were going to recondition my PSUs: I'd want you to replace my MOSFETs, Caps etc. on my Indy. The O2 just seems to have caps as the achilles heel, so that's that. Case-by-case.

But again, I think for a small system without a super complex PSU, pricing needs to be reasonable for someone to take it up, and it has to have a benefit. Not to criticize your other post on the light bar repair, but many people seem to be far more interested in using LEDs for the well-known reasons of such. I think the argument between OE spec and modified spec depends on the case, as we've seen here. I.e. I'd never buy a DEC Multia for over a hundred that had been used and abused for decades without a fan installed because they're well known for cooking themselves.

[weblacky]

I understand this is bootique and costs money, I've already invested a large amount of time and money into these tools and parts. Just like a car. Some people can come to terms with driving a car that's looks like a beater. While I don't mind driving old cars, somewhat prefer it actually, my car cannot look like a beater, it needs to look stock and new or well kept. Same with my SGIs. I can't have large paint chips or discoloration or cracks. It doesn't look like the designer intended, it just looks like I abused it.

I like very slight modifications to my possessions but mostly they need to be hidden. I don't do all outward customizing, because unless I have some great idea, the designer likely has a better sense of aesthetics than I do. My mod would likely look ugly, also unless it enhances functionality, I keep old things old. If it was designed that way, I keep it that way.

I understand people's mindset of being "re-trained" for LED conversions, but I take the same approach to my car. LEDs inside are fine, I'd never place LEDs in my outside headlamp assemblies...because they aren't designed for it! The reflectors are designed for that near-360 degree transmitted light and the whole thing would look off and create dark spaces in my light field, because LEDs are directional and often aren't placed correctly on the PCB to mimic the transmission profile of a bulb. It was designed for an incandescent and the heat that puts out, so unless I can get a new headlight assembly that's designed for LEDs, I'm not interested.

I feel the SGIs are the same. Unless you saved a system from the trash and had to bondo it, I don't want to walk in a see a safety-yellow Octane with black highlights/racing stripes and a purple light bar. I'd rather walk into a museum, every item is like-new or restored as best as you can to give an authentic view of what life was like then.

What was life like on an SGI in the 90's, come see it for real...not "yo dog, I hear you like SGIs, so....". That's the reason I'm big into repair, I believe preserving the intent of the item is almost as important as the item. Certainly if the original part was gone and I was left with no choice for the functionality...I'd bite the bullet and mod it. But not by choice.

Why wouldn't someone want their vintage item to work just like it did out of box?

[robespierre]

What causes the damage to the steel on Indy PSUs? Is it just leakage from electrolytic caps dripping onto the case from above?
I generally use "cold galvanizing spray" (zinc paint) on steel items with corrosion. Did a furnace fan last year and it worked well. It is cosmetically not the same as actual galvanizing but protects the metal about as well.