Why do capacitors leak?

[bigmessowires]

I'm branching off the discussion from a for-sale post about whether original unused motherboards in a vacuum-sealed bag would be less likely to need recapping than used motherboards. And putting the discussion in this forum, since it's the most hardware-oriented.

I don't know exactly what makes electrolytic capacitors leak and fail after a few decades. But in general when you think about computers, cars, bridges, or anything else man-made that breaks down over time, it's due to chemical changes in the materials they're built from. And those changes are usually caused or exacerbated by:

• heat
• oxygen
• UV light exposure
• water (liquid water or just humidity)

An unused motherboard in a vaccum-sealed bag, and stored in a dark place at room temperature, will be quite well protected from all four of those destroyers. That's not a guarantee the capacitors will still be good after 40 years, since the electrolyte may also react with the seal or other components. I don't think you can stop cap rot forever. But if I were a betting man, I'd bet that sealed NOS motherboards would be in substantially better shape than other identical motherboards the same age. Thoughts?

 

[GRudolf94]

They will be in better shape for sure - it also excludes the 2nd biggest destroyed of boards after electrolytes: human hands.

Now, as to what causes degradation for a start, my vote goes to the caustic-by-necessity nature of the electrolyte, plus oxidation of the sealing material in contact with the atmosphere.

 

[Daniël]

It would largely depend on the type of cap, I'd think. I think it's practically a fact that late 80s/early-to-mid 90s SMD caps had seals that were more prone to failure with age than the THT caps that came before it, and even SMDs that came after it. Machines with SMD capacitors that are the same age now, as the machines with the early leaky SMDs were when they started going, don't seem to have it as bad.

Storage doesn't seem to prevent it, but anecdotally, I've had Mac hardware with SMD caps from Japan arrive with absolute pools of electrolytic fluid, which I correlate to the higher humidity and generally harsher storage environments over there (Macs tend to be far more yellow there too, which is what I base that off). Similar age stuff stored in the West is mildly crusty by comparison.

 

[Phipli]

I've had Mac hardware with SMD caps from Japan arrive with absolute pools of electrolytic fluid,

I've generally seen this on machines that were started after being in storage, then left again. I caused it in my own LaserWriter by running it a few times, then leaving it for a few months. Running a machine with pooled electrolyte also heavily accelerates corrosion on power traces and pins. Full on electrolysis removes material. Spot the pins that saw more current...


My theory is that cooler + fairly high humidity storage things last better. My parent's house is in a high humidity area and has no central heating. The significant number of machines stored there, none were particularly bad. All the worst machines we have were bought in recently.

 

[aladds]

The main issue with an unused electrolytic will be how the insulator layer is formed when they're new, and how they're then designed to maintain that over time. You have two metal layers with an electrolyte between them, and the insulator is "formed" on one side of the electrolyte by slowly rising the voltage to cause a reaction (over-simplification but I can find links to explain this in more detail if you like). Over time, this insulating layer does break down, but using the capacitor re-forms it and maintains the ability for it to store charge.

They shouldn't leak if left - but if you power the board up without allowing the voltages to slowly rise over time (re-forming) then they could leak (or worse, break down). In analogue audio gear you can just do this with a variac and slowly get the whole thing up to the working voltage, but with digital electronics things don't generally work well at lower voltages so it's usually best to remove capacitors, re-form them and put them back. For cheap SMD capacitors it's generally not worth it and people will probably say that if you're removing them you might as well replace with new.

Tech Tangents did a good video on this, actually, although that's more aimed at larger, more expensive capacitors in older power supplies, where replacing them would be rather expensive (and wasteful!)

This is separate from any issued with mechanical failure of rubber seals or pressure which causes the can to rupture. That's likely storage/humidity related.

 

[Phipli]

They shouldn't leak if left

Lets be honest, you could say "They shouldn't leak". Using them isn't meant to cause a leak.

 

[Phipli]

This is separate from any issued with mechanical failure of rubber seals or pressure which causes the can to rupture. That's likely storage/humidity related.

Given they were designed for home use, and all of them are degrading in homes (and elsewhere), I wouldn't say it is a storage issue. You can't expect a home user to keep their consumer electronics in a special, climate controlled zone?

 

[François]

One thing that could explain leakage in computers that had seen a lot of use is stress induced by cycles of heating and cooling. When the computer is running, heat will expand the capacitors and when the computer is turned off cooling will shrink them again. Repeat every day for several years. This can cause a lot of stress to the seals, much like bending and straightening a metal paper clip will eventually break it.

 

[joshc]

I don't have anything scientific to say on this subject, because I'm not that clever. All I can offer is what I've experienced first-hand with the various Macs that I've worked on. I think I've re-capped something like 60+ machines in the past couple of years.

There are so many factors that contribute to capacitor leakge, from a chemistry point of view, far beyond what I am able to understand, it's definitely much more complicated than "this was dry stored, so there won't be any cap leakage" or "this computer wasn't used much, so there shouldn't be any leakage". It's just not that simple.

Here's some photos of a capacitor I removed from a IIsi NuBus adapter, which was sealed in box, never used since being manufactured 30 years ago, and yet they were so leaky...





I've seen many variations of this, though. I've had sealed/boxed stuff that hadn't leaked at all, I've had computers where the caps definitely should've been worse than they were, but for whatever reason, they were still OK.

Perhaps different ones leak in different ways? Sometimes the leakage is very dry, sometimes wet. I would assume wet means the leakage was recent, whereas dry means it's been left for a long time since it leaked.

Why do Classic, Classic II and LC, LC II boards have some of the worst cap leakage I've ever seen? Not just the ones I've re-capped, which is about a dozen, but other poeple's I've seen as well... did Apple use a different supplier for those machines?

See what I mean... really really bad leakage, this was a loft-stored machine which probably contributed to the dry/crustiness of the caps...

There isn't much time difference between the LC II and the LC 475, a few years, but 475s, albeit definitely needing a re-cap now, often look OK on initial inspection. This doesn't mean they haven't leaked, but why the extreme difference in how visibly the leak? My guess is a combination of environmental / storage and usage conditions, and perhaps also slight differences in the materials used to manufacture the capacitor as well?

 

[Phipli]

I don't have anything scientific to say on this subject, because I'm not that clever. All I can offer is what I've experienced first-hand with the various Macs that I've worked on. I think I've re-capped something like 60+ machines in the past couple of years.

There are so many factors that contribute to capacitor leakge, from a chemistry point of view, far beyond what I am able to understand, it's definitely much more complicated than "this was dry stored, so there won't be any cap leakage" or "this computer wasn't used much, so there shouldn't be any leakage". It's just not that simple.

Here's some photos of a capacitor I removed from a IIsi NuBus adapter, which was sealed in box, never used since being manufactured 30 years ago, and yet they were so leaky...

View attachment 61624

View attachment 61625

I've seen many variations of this, though. I've had sealed/boxed stuff that hadn't leaked at all, I've had computers where the caps definitely should've been worse than they were, but for whatever reason, they were still OK.

Perhaps different ones leak in different ways? Sometimes the leakage is very dry, sometimes wet. I would assume wet means the leakage was recent, whereas dry means it's been left for a long time since it leaked.

Why do Classic, Classic II and LC, LC II boards have some of the worst cap leakage I've ever seen? Not just the ones I've re-capped, which is about a dozen, but other poeple's I've seen as well... did Apple use a different supplier for those machines?

See what I mean... really really bad leakage, this was a loft-stored machine which probably contributed to the dry/crustiness of the caps...

There isn't much time difference between the LC II and the LC 475, a few years, but 475s, albeit definitely needing a re-cap now, often look OK on initial inspection. This doesn't mean they haven't leaked, but why the extreme difference in how visibly the leak? My guess is a combination of environmental / storage and usage conditions, and perhaps also slight differences in the materials used to manufacture the capacitor as well?

View attachment 61626

I think the caps got better with time, the 68k macs, especially the ones from ~90-91 leaked faster than the 93 ones are.

But yes, when I talk about something's being better for the caps not leaking... I mean slowed things down, not stopped. And with 30+ year old machines, you are unlikely to know how it has been stored the whole time. Even my dad's collection, there isn't a single 90s Mac that has spent its whole life in his possession, including ones he bought new (he sold them to fund the next, and only bought them back a few years later).

I recapped a 475 for slomacuser. It was the tidiest 68k board I have ever seen and didn't "need" recapping... Except... At some point soon it will, and it was the easiest and cleanest recap I've ever done, so completely worth it, because now it will /never/ leak because it has solid state caps. Completely and utterly worth it.

I'm so glad I heard about caps in time to save my dad's collection. We'd have been devastated. He has carefully stored them sans-battery on especially bought shelves since the early 2000s, because he felt they were too interesting to throw out when they hit rock bottom prices.

I'm curious to see how long modern electrolytics last - have they got better? Or are they now worse due to cost reduction and targeting specific design lives?

 

[Phipli]

One thing we haven't discussed is the early 2000s "capacitor plague" a completely different thing where they suffered from early failure.

With this, they seem to have dried up and lost their capacitance, but without impacting the boards (as if they off-gassed and weren't corrosive). I've had this with ATi Radeon cards, where a card was working, but with a shimmer in the analogue output. It had done this since I owned it in about 2009. Given it was a 2001 board, I got curious if it had suffered from capacitor plague, removed a cap, found it had 0uF capacitance... and that every electrolytic on the board did! I replaced them and the shimmering was gone.

Amazing the card worked at all really!

 

[GRudolf94]

"Caps are a scam by Big Passive"
- eschatologist.net while joking

"So you're telling me a thing that is essentially a break in the circuit if you look at it conventionally, filled with corrosive fishwater is supposed to make a difference in the circuit? Nah eschaton is right, caps are a scam, don't recap, just take them out"

Smh engineers. Just make the boards without caps. It saves everyone's time and money!

 

[Phipli]

"Caps are a scam by Big Passive"
- eschatologist.net while joking

"So you're telling me a thing that is essentially a break in the circuit if you look at it conventionally, filled with corrosive fishwater is supposed to make a difference in the circuit? Nah eschaton is right, caps are a scam, don't recap, just take them out"

Smh engineers. Just make the boards without caps. It saves everyone's time and money!

RIP audio.

And Duos

 

[joshc]

Well, a Classic II board (I have not tested this with others) actually powers on and seems to operate OK without the electrolytic capacitors... (that's with all the other factory solids still there). Albeit with no sound, but yeah. Probably other bits missing that are important though...

 

[GRudolf94]

Duos need One Cap. Else they wreck the MOSFET that is the 5V buck's switching element with reverse spikes, it shorts out and the whole machine gets 24V

 

[bigmessowires]

Fantastic discussion here.

So you're telling me a thing that is essentially a break in the circuit if you look at it conventionally, filled with corrosive fishwater is supposed to make a difference in the circuit?

This was exactly my reaction when I first started learning basic electronics. "But the circuit's not even connected!"

If a motherboard's been recapped with solid-state capacitors, would you expect it to last more-or-less forever, or are there other sources of long-term failure? I might imagine PCB or IC packaging cracks from repeated heating and cooling, or IC failures due to electromigration, but those will only happen if the computer is powered on.

Do you recap your motherboards as "preventative maintenance", or wait until they begin showing problems?

 

[joshc]

Do you recap your motherboards as "preventative maintenance", or wait until they begin showing problems?

I plan to pre-emptively re-cap my G4 logicboards, I have two identical AGP ones and there's about 10 or so caps on there. For the low amount of effort it'll take to replace them, I figure it's worth it.

 

[Phipli]

would you expect it to last more-or-less forever, or are there other sources of long-term failure?

If you use tants you need to update them a little, the safety margin on tants is 100%... which is crazy, why not just label them at half the value?!

Transistors do eventually wear out, but we're not putting that many hours on retro machines. Hum. Clocks is probably the next most frequent failure.

 

[Phipli]

Do you recap your motherboards as "preventative maintenance", or wait until they begin showing problems?

Available time means I've been doing them based on manufacture year and likelihood of leaking... I would do more preemptive work... But the backlog is big. Got to do all my CD drives urgently, and check my NuBus and PDS cards.

 

[GRudolf94]

If a motherboard's been recapped with solid-state capacitors, would you expect it to last more-or-less forever, or are there other sources of long-term failure?

Do you recap your motherboards as "preventative maintenance", or wait until they begin showing problems?

I recap them on an as-needed basis (i.e. stuff I know is prone to failing, or has already failed). Solid state caps also fail spontaneously. MLCCs fail short quite often, tantalums fail short if operated with insufficient derating, or subjected to spikes/reverse polarity. As for thermal cycling failures they're a non-issue as far as anyone knows. Electromigration is only a concern on things being run /hard/ - not the case for any Mac ever. Get an edge-case op-amp you're feeding the beans to, however, and you'll see a failure in a day period. (see TI OPA858 datasheet - they even list it there)