[weblacky]

Good question, unfortunately several these units were stored in non-climate controlled areas. We’re talking tractor trailers and attics. Two of the units were in personal possession so they are in much better shape. But this is textbook penetration on a wide scale of the galvanic coating.

Not a dent or a scratch but just a uniform issue. Likely dust or some form of particulate that became damp and then stuck to the surface over a long period of time trapping that moisture against the plating, penetrating it.

A few could be passable on their own but I’d say half are absolutely unacceptable. Every rivet has a small tinge of brown and most of the metal has the white oxidation and brown foxing-like appearance. No holes or anything.

I did very much consider those types of sprays. And looked up several brands. For a spot treatment I would absolutely agree with that suggestion but some of these are just too far gone for spot treatment.

You’d also have to prep the surface and for a small area that’s not a big problem for a large area it might be cheaper to use some sort of acid or chemical in order to prep the metal which is why I was thinking of just engaging a local company to just electroplate them after doing the stripping themselves.

Because I do live in a major city I do have local access to plating services. So I don’t know how much we talking about here in terms of cost. However I’d imagine I get a better rate if I did three or four or five at a time versus doing one or two.

Local companies obviously have the ability to just chemically stripped them and plate them in no time. I even thought it might be interesting to go with a zinc chromate finish instead of a normal nickel finish in order to differentiate the supplies as being rebuilt.

But the concerns on those are the labels. I’d hate to not have replacement labels printed up to place on them and that’s really what stops me from doing it right now.

[robespierre]

One thing that goes bad on some older equipment is foam padding. Certain polymer foams can break down (accelerated by air pollution) leaving corrosive products. Foam is sometimes inside computers in the form of double-stick tape and that can rot as well. Certain adhesives can also turn corrosive (the glue that holds down capacitors in power supplies being a particular culprit).

If you ever see a Personal Iris, it has a black tape strip to isolate the I/O panel and that always starts corroding flange on the case it's glued to.

[Raion]

Weblacky, I almost promise a bead blasting cabinet (under $200 from HFT) and parkerizing solution is the best way here. Yes, you're dealing with a tub of hot phosphate, but it's no more dangerous than lead solder. You can reuse the solutions. They will do dozens at a time. We used it when i was working for a gunsmith.

But yeah, heat cycles cause moisture to condense and such. Stuff stored in the desert usually doesn't oxidize as well, but eventually everything rusts if it's not stored climate controlled - that's why I tell people to use a CLIMATE CONTROLLED UNIT OR GARAGE if they need it.

[jenna64bit]

I would fall on the side of fewer replacements over more. Mostly voltage regulator chips, caps, etc. Anything known to fail due to design or age. Would be nice to have a more "trustable" PSU for my O2 & Octane.

Edit: Also definitely pro-derusting and painting. Even if it's just regular rustoleum after brushing down with steel wool.

[Raion]

I'm gonna convince you guys about parkerizing one way or the other.

It's an easy process.

[Irinikus]

I personally think that it would be a good idea to try a complete refresh of all of the replaceable components within the power supply units, as a failure of a single component in one of these things seems to lead to a cascade of component failures within the unit!

I therefore think that it would be wise to generate complete component lists for the various power supply units, thus making it easier to order complete sets of replacement components!

In this way we can hopefully avoid charred circuit boards! (As this becomes a problem when it happens!)

[weblacky]

I personally think that it would be a good idea to try a complete refresh of all of the replaceable components within the power supply units, as a failure of a single component in one of these things seems to lead to a cascade of component failures within the unit!

I therefore think that it would be wise to generate complete component lists for the various power supply units, thus making it easier to order complete sets of replacement components!

In this way we can hopefully avoid charred circuit boards! (As this becomes a problem when it happens!)

I thought about this but it's not really feasible for a couple reasons (unless money's no object for me).

1. Several in-demand PSUs (think Indigo2) have a HUGE amount of variants (Revisions and models).  An accurate bill of materials isn't feasible for those, as I'm sure I'll come across revisions I'd never seen and then it's auditing all over again.  For those models a pure template isn't happening.  However it's very doable for models like Indy and O2 where they are known suppliers and versions for a simpler tracking of PSU versions. Like 1-2 versions.

2. If I had the cash, I did really think about taking a PSU down to its board and having a company use a robotic probe to try to reconstruct the original circuit (part inventory at positions, while probing position permutations to find all connections).  However that takes a lot of money and I don't have a legal way to even sell clones of the PSUs (because I don't own the design).  So fixing existing units (and given the small audience) I think makes sense.

3. Finally, to be 100% fair, even new parts fail.  Yeah, we might all feel better if 100% of all parts were brand new but who is going to pay for a hand-assembled from scratch clone PSU.  Seriously.  No one is right now.  I get the remanufactured argument, I think it's a risk we all need to understand and be able to take.  But while there are some parts that easily fall into the "going to fail" category, others (like resistors) are my bigger worry.  I think the parts I'm thinking of that are bound to fail, given the time they've been installed, is small when compared to how many parts are in some of these!



But I'm surprised how all over the board some of the responses have been.  Which mirror my own logical struggle.  Redo everything vs touch as little as possible, and everything in between.  Yeah, it all has to work, I get that.  But just like a used car - a lot of parts do work and have been working this whole time (have good history).  And really what's the difference between a new part and a used one - really it's manufacturer warranty/wear/and expected lifespan.

Given the cheapness, I'm not against doing like all X and Y Class caps, and NTCs, along with electrolytics.  In terms of semiconductors, I'm still on the fence.  For those it's not always a good thing to put in a faster or stronger FET.  I can detect semiconductor leakage with my equipment so I'm not too worried about that.  Also maybe optocouplers are cheap?

But I think the real truth is, without evidence of improper design (improper derating) a properly derated semiconductor has no rated lifespan.  A bad cap can just as easily short out an old FET as a brand new unit.

I think what I need to consider is where that line is and how to describe these units in a fair manner (like, rebuilt-used).  Terminology that is honest but descriptive enough to set an expectation.  After all, why would you trust a "used" PSU over a rebuilt one?  The reason is clear...you think that the rebuilt one needed to be rebuilt due to failure.  So really you're getting a "fixed" PSU, which isn't the same as a known working for 20+ years original PSU.

And it seems it basically comes down to trust.  If you don't have trust in the repair, you'd rather have an untouched and working 20+ year-old unit then one that failed and has been fixed and refreshed.

I think that's the crux of the issue I have trouble with.  How can you (I guess with warranty) establish a trust with no prior history?  I guess we'll have to see as things progress.  

Perhaps this is a situation where it's no use worrying about the maybes, just handle the cases as they come.

[jenna64bit]

Agreed, we all seem to have somewhat different standards. Maybe figuring out what an overall acceptable price threshold is and then speccing to that is easier than asking specifics? EG, I would happily pay around 250-300.

[Trippynet]

I'd love to just have a working PSU for my Indigo2 again. I've hardly used it in the last year due to the PSU issues and it's quite sad really.

Maybe a few different options? Ie, a price for capacitor re-cap, higher price for replacing other components etc? That way people can choose how much to spend and how major a job they want.

Ironically, I'd be intrigued to have a repair which replaces everything but the capacitors (I've had this done, and the PSU is still flaky).

[weblacky]

I'd love to just have a working PSU for my Indigo2 again. I've hardly used it in the last year due to the PSU issues and it's quite sad really.

Maybe a few different options? Ie, a price for capacitor re-cap, higher price for replacing other components etc? That way people can choose how much to spend and how major a job they want.

Ironically, I'd be intrigued to have a repair which replaces everything but the capacitors (I've had this done, and the PSU is still flaky).

 
Yeah, I've thought about cases like yours.  I've just received my DC load tester (starter dual channel, I'll likely need another one in addition for real testing...vs smoke testing) and I'm hoping to get into a basic bench run, on known PSUs.   What really bothers me is accounting for bench runs vs system runs.  One thing I'm really curious about your's is this issue.  I know you said the 90 minutes sudden turn off thing...I wonder if a bench loaded run would do the same thing?

I'm also  curious about bench runs regarding cooling,  I could places fans on them for extra cooling but that might hide issues once in the SGI cases.  What I will be forced to do is have "mule" system that I can risk explosion on to just run the PSUs for like 1 week with controlled CPU loading schedules to be 100% confident for myself.

I think with PSUs like yours the key will be: A. knowing the symptoms from previous owner (not finding out she's borked in some new way) and B. having a working EXACT duplicate (Model + Revision) (or a probe record of a duplicate) to compare with.  Most of my equipment is based around comparison techniques.  So I think that's going to be the only way to solve your mystery. 

If I can fix the 2 broken Indigo2 PSU units I have, plus refresh all 4 units I own that should work, if you haven't been resolved by then, and I have a duplicate supply, I might do a trade plus a little $ for your unit and see about solving that mystery.