I suspect that most big old electrolytics found in older computers mainly die from the seals breaking and moisture getting in, just as capacitors of the 1940s (even micas) are fond of doing. No amount of reforming will fix this problem. -- Will

In article <AANLkTi=D6VV26U2YtTfJAYK3o09CAgcxnQ9gtSWEs9vK at mail.gmail.com>, William Donzelli <wdonzelli at gmail.com> writes: To add some specifics to the claim: it was in reference to the power supply on a Terak workstation. (PDP-11 with a monochrome frame buffer, see pics here: <http://picasaweb.google.com/legalize.slc/Terak> and docs here: <http://bitsavers.org/pdf/terak/>. I believe this system has a simple linear power supply design, but bitsavers doesn't have schematics online and I'm away from my unit to inspect it right now. -- "The Direct3D Graphics Pipeline" -- DirectX 9 draft available for download <http://legalizeadulthood.wordpress.com/the-direct3d-graphics-pipeline/> Legalize Adulthood! <http://legalizeadulthood.wordpress.com>

In article <201010071328.o97DSiIH038757 at billY.EZWIND.NET>, John Foust <jfoust at threedee.com> writes: Well, isn't the answer just what we always do? Replace the caps and try again? -- "The Direct3D Graphics Pipeline" -- DirectX 9 draft available for download <http://legalizeadulthood.wordpress.com/the-direct3d-graphics-pipeline/> Legalize Adulthood! <http://legalizeadulthood.wordpress.com>

William Donzelli wrote: I thought that when the seal failed, the electrolyte dried out? Either way, though, you're correct; once the seal has failed the capacitor needs to be replaced. I worked on the PDP-1 Restoration Project at the Computer History Museum in Mountain View, CA. This equipment was manufactured in 1961. We did about six months of inspection, maintenance, and repair of the PDP-1 and the Type 30G display before we first applied power. Part of this process was to check all of the electrolytic capacitors, and reforming or replace them as necessary. It appeared that all of the capacitors were original. Many of them needed to be reformed, and we used an extremely over-engineered approach to doing that. Only a very small percentage of the electrolytic capacitors had failed such that they could not be reformed. I fully expect that all of those capacitors will fail eventually, but there isn't nearly enough data to predict a mean lifetime. Eric

On Thu, 7 Oct 2010, William Donzelli wrote: Ah, like the Wima Tropydur (called "Backpflaume" here, i.e. prune). Used mainly in consumer electronics during the 50s. Christian

On 10/07/2010 03:37 PM, William Donzelli wrote: We didn't attempt any failure analysis. Our reformation procedure involved computing a current limit at voltage steps of 0.5V each. If at any given step, the leakage current of the capacitor didn't drop below a reasonable level after an extended period (perhaps 12 hours), we considered the capacitor to need replacement. IIRC, most of the ones that failed never dropped to a reasonable leakage current even at a fairly low voltage. Eric

But running the device will dry out and stress the seals faster. Getting old sucks. -- Will

' How could you know? Has you been there before? ;oD

Barry wrote: I expect it will be substantially *less* troubling.

lol! ________________________________ From: Barry L. Kline <blkline at attglobal.net> To: On-Topic and Off-Topic Posts <cctalk at classiccmp.org> Sent: Fri, October 8, 2010 3:20:03 PM Subject: Re: capacitor aging claim On 10/08/2010 03:44 PM, Eric Smith wrote: But it does eliminate any possibility of "aging gracefully."

On 10/07/2010 03:15 PM, Richard wrote: Actually, it is. The oxide breaks down whenever the capacitor isn't powered, but it is an exponential decay effect. Leaving something powered off a short while doesn't let it break down too much, and when you power it up again, the oxide reforms a little bit, back to where it is supposed to be. If you take a 25V electrolytic capacitor, and use it at only 6V for a long time, then apply 25V suddenly, you're likely to damage it. Yes.

If the innards are contaminated, it is possible. An example would be the aforementioned paper Sprague "bumblebees" - more often then not, moisture has penetrated the case and has compromised the dielectric. You can safely use them at low voltages, as guitarists do, but at rated voltage (200 to 600 Volts, depending on the type), they will likely fail. I am not so sure about oil caps - they tend to be pretty bulletproof. Even when the seals leak, they still tend to be good at rated voltage. It could be that the oil just keeps everything else out of the innards. -- Will

Yes, although I think the term "bumblebee" comes from the OEM version with color code bands, much like resistors. It seems like at least two of the six or seven bands are yellow, thus the name. Vitamin Qs (and similar Astron and Gudeman caps) were, and still are, pretty bulletproof. But yes, when they fail they can make a hell of a mess. The oil is not PCB laden, but plain mineral oil. I save every one of these caps I can get my hands on. The oil caps actually were paper, but the oil soaked the paper through. Plain paper capacitors generally had no oil. -- Will

Probably wax. Which, yes, I suppose is kind of like oil. -- Will

Chuck Guzis wrote: It applies to aluminum electrolytics, and possibly to other kinds of electrolytic, but I don't think it would apply to non-electrolytic capacitors. Eric

Well, this doesn't specifically mention the "exponential decay effect", but will this help?: http://electrochem.cwru.edu/encycl/art-c04-electr-cap.htm "[...] the capacitor "fill" electrolyte continues the healing work of the original forming electrolyte, repairing and thickening the dielectric locally as required. This healing process is driven by the capacitor's dc leakage current, which is drawn whenever a dc voltage is applied to the capacitor, that is, whenever it is in operation. In fact, electrolytic capacitors often last longer when they are in continuous, mild use that when they are only charged up briefly every year or decade." Feel free to read the entire article if you like, and/or check out some of the links towards the bottom. (I found the Cornell Dubilier article in "Further reading" particularly helpful to me when I needed a good general overview, but alas the link is broken. It's still on the CDE site, though, at www.cde.com/catalogs/AEappGUIDE.pdf .) -O.-

The specific context was a device that's been powered off a relatively short time. The oxide has only deteriorated a small amount, and not to less than the working voltage. Powering it up, at the working voltage, will do some reformation, and will no damage the capacitor. I never said that this procedure was applicable to capacitors that had been powered off for a long time; in fact, I think I specifically said otherwise. I described a different procedure for capacitors that haven't been powered for a long time, which involved limiting the current into the capacitor. Limiting the current has the effect of not allowing the voltage to increase beyond what the capacitor can handle (assuming that you chose the current limit appropriately). This is actually much better than ramping the voltage at a constant rate, because the capacitor make take longer for reformation during some points in the process than during others. We saw that with some capacitors in the PDP-1. Despite that, it is still the case that reformation is what is occuring in both scenarios, which is what I replied "actually, it is" to in the quote above. It is the same electro-chemical process; you seemed to be claiming that it was not. Eric

On 8 Oct 2010 at 18:43, Alexey Toptygin wrote: A Sprague brand for hermetically-sealed metal-jacketed oil-in-paper capacitors. Considered to be magic by the guitar folks; but then, so are polystyrene capacitors. My exposure to them was mostly on military equipment. The jacket is steel, so if one explodes, you do get some nasty shrapnel. It happened to me with a rather large one and might have cost me parts of my physiognomy if it had occurred just a few seconds earlier. What was left embedded in the acoustic tile above my bench was an ugly corkscrew of steel. --Chuck

On 8 Oct 2010 at 12:53, Peter C. Wallace wrote: Noted. But transformers and capacitors containing PCB were commonly known as containing "oil". I had never heard of PCB until the 1980s. So "paper in oil" is at least correct in the colloquial sense of the word. --Chuck

On Sat, 9 Oct 2010, William Donzelli wrote: Maybe its a vintage thing. I had some WW2 vintage 10 UF @ 600V VitaminQ capacitors that definately contained PCBs Peter Wallace Mesa Electronics

On Sat, 9 Oct 2010, William Donzelli wrote: I remember draining bad capacitors when I was quite young (I abused the 600V capacitors by using them for strobe lights charged to 2 KV) And am pretty sure that the Vitamin Q ones were PCB (light green slightly sweet spelling fluid) But is has been many years so I could be wrong. Peter Wallace

Richard wrote: Yes, replacing capacitors is generally easy than reforming them. The main reasons to reform them instead are: 1) If you're trying to preserve an historical artifact 2) If you can't easily or inexpensively get a replacement with comparable specifications.

allison wrote: We weren't too worried about transistors in the logic, as they weren't likely to cause a great deal of damage if they failed. And we weren't worried about transistors in the power supplies, because there weren't any. Most of the PDP-1 power supplies just use ferroresonant transformers with diodes for rectification and big electrolytic capacitors for filtering. We checked the capacitors out of circuit, and we checked the power supplies (with dummy loads) before we ever powered up the PDP-1 with them. We did find one diode in one power supply that had failed shorted. The deflection and HV power supplies for the Type 30G display did contain active circuitry including transistors, and as with the logic power supplies, we tested them before powering up the 30G. We did find some serious issues with these power supplies. The power supply for the photomultiplier tube used by the fiber-optic light pen had also gone bad. The fault was in the transformer, inside a potted assembly. We have temporarily replaced that power supply with a new commercial power supply module, but we might rebuild the original in the future. The most serious fault we found in the machine was that the power control for the Type 30G display was wired *completely* wrong, such that turning it on put line voltage across a closed vane switch intended to sense when the fans were up to speed. This resulted in very high current through the vane switch, well beyond its rating, causing it to weld itself closed, and trip the circuit breaker. It appeared to us that someone had tried to kludge the power control to fix some problem, and failed utterly. We believe that the demos of the PDP-1 system at the Computer Museum in Boston must have been done without the use of the power control, with the power supplies just plugged directly into an outlet strip or the like, since the power control could not possibly have worked. We rewired the power control to match the engineering drawing and replaced the vane switch, and then the power control worked correctly. We know that the deflection power supply for the Type 30 display was missing entirely in Boston; in photos of the demo there it is clearly missing and a bench power supply can be seen. The proper deflection supply for the Type 30 was a product of a third party, rather than being made by DEC. Fortunately CHM had another Type 30 with that power supply, though it was non-functional until we reverse-engineered and repaired it. Throughout the restoration process, and to this day, the team has taken great care to avoid situations where substantial damage could occur to the artifact. Eric