The Capacitor Plague | Hacker News Reader

The Capacitor Plague

deadprogrammer.com

84 pointsjcsalterego47 comments

Electrolytic caps typically have low lifespans compared to other types of caps, especially over larger temp ranges.. even with proper construction.

When I use to do design for circuits that needed a long lifetime I would really question any use of electrolytic caps.

what would you use? tantalums? ive have the joy of recovering old equipment with dried tant caps on the power rails. (23 of 'em!) maybe tants are made better these days...

argh. how does one delete comments? i refreshed by accident :(

There is a delete button. Click the "link" link, and then you should see a delete option.

...but only within two hours of initially posting the comment.

sadly, this system is incompatible with 'noprocrastinate' :)

what would you use? tantalums? ive have the joy of recovering old equipment with dried tant caps on the power rails. (23 of 'em!) maybe tants are made better these days...

In some cases you can use ceramic capacitors instead of electrolytic ones. I've done this exactly because I don't trust electrolytic capacitors to be anything more than big and cheap.

for a PC switching power converters, such as in this article?

sooo...where are you buying these mystical 2200uF ceramic caps? :)

On the positive side, this has led to some improvements, at least in the motherboard market. I've seen both Asus and Gigabyte market motherboards with "High Quality" "100% Japanese" capacitors.

This almost ruined my company... I still have nightmares about it.

Not to dig up the dark past, but could you share more. I'm interested.

I once bought a set of monolight strobes (for photography) made by a respected brand name (Bowens.) Every unit in the set failed shortly after the 1-year warranty due to failure of the electrolytic capacitors in the power supply. I used these lights heavily, so for other users, they might last two or three years. That's still a remarkably short time to failure for this kind of component.

A friend has an Apple iMac G4 that they paid a shed-load of cash for and expected to last well - it nearly made 3 years. The point of failure? Electrolytic caps in the [overheating] graphics card. The card is soldered in on the screen side of the mobo.

I think this is a known issue that Apple will fix, however. I remember reading something related to Apple fixing bad caps in iMacs, anyway...

They did change the caps from what I've read (there was a court case about it in the US I gather) but the real fix on teh iMac was putting ventilation holes in! Who'd have thought. Damned form over function. Also I gather they use slot mounting GPUs now?

Is this a long running problem? I remember reading somewhere that most of the problems with old hi/fi and tv electronics were due to bad capacitors which can easily be replaced.

I assumed that they were carefully designed so that the electrolyte would go off just out of warranty (no I'm not kidding) and force a resale.

Perhaps electronic manufacturers that are big enough could have a state mandated "cost of parts and actual labour costs" repair service.

actual labor costs quickly make repair more expensive than replacement.

I think the point was to make it at-cost (i.e. break-even) for the manufacturer so that they aren't trying to gouge the customer by creating a product that will surely break (forcing them to pay marked up repair prices or just buy a replacement).

Yeah, call out charge for washing machine repair is around £100 (minimum), actual cost is travelling 14 mi (at most where I am)@ 50p (that's the reclaimable amount for mileage) plus the "engineer's" hourly amount for perhaps 2 hours (I'm going with £25 per hour gross). So not £157 but £57 - the company absorbs the downtime costs of callout personnel - this would make a whole heap of difference in whether it is worth repairing ones washing machine.

For about 6-9 months in 2006 We had a series of Dell Optiplex GX 270s that had a remarkable failure rate - close to 60% of our systems failed. We knew that Dell must have been having a tough time, when the FIRST question they asked on a support call was whether we saw any "Bulging Capacitors on the Motherboard" - I mean, not "Have you Tried Rebooting", or, "Have you tried powering down for 90 seconds", but "Do you have bulging capacitors."

I'll say this for Dell though - they fixed every last one of those systems within 4 hours of our call.

I have to wonder whether Large Manufacturing organizations are aware that there is a good chance that going with subpar capacitor manufactures will cost them, but do so anyways because of the amount of money they save.

How did they fix a hardware failure in 4 hours?

Telekinesis? Teleportation?

Is there something I'm not understanding?

Enterprise customers can get hardware replacements same-day.

When I worked for $BIGCORP we'd get hardware failure calls from IBM to the tune of "server just phoned in with a drive failure, that part's not in stock locally, a plane left Atlanta with the part 15 minutes ago."

It was damn slick.

You just call them, and they separately courier-ship parts to you and send a technician onsite.

Go buy a $984 3 Ghz 2 Gbyte Dell Optiplex 780, and, for an additional $216, you get 3 Year 4HR 7x24 Onsite Service.

And yes, I've called them at 11:00 at night. :-)

First it was the Maxtor hard drives that went bad, then it was the capacitors popping. The company I worked for at the time had literally thousands of these things--and nearly every single one had to get a new motherboard and hard drive before the warranty cycle completed.

Also, we had a Dell 'certification' that essentially let us go to a part of the website, and order the replacement parts for any warrantied system. It was really slick. When the site didn't get it done, there was a super-secret US-based support line we could call :)

So, for hardware manufacturers, is there any reliable way to identify good quality caps?

If you know that a cap is suspect, and the article mentions what are probably helpful sources like badcaps.net, then you can accelerate the aging process by running the board in question at 10-20% overvoltage in a temp chamber heated to 100C or whatever your design's edge case is. MBTF values assume nominal voltage and temp, so increasing this has the effect of decreasing the mean time. This is not 100% guaranteed, since you are tweaking parameters of a statistical formula, but you nevertheless stand a good chance of weeding out a particular brand/model cap if you test multiple units and compare the results to those from a control group. A past job did exactly this to qualify part substitutions, and this sort of testing is effectively a requirement for temp-sensitive parts like caps. We were able to successfully weed out bad caps with this testing.

I've never had the solid state caps go bad on me. I've been using strictly solid-state capacitor motherboards on all my computers for two years now, and I haven't had a single (motherboard) problem with them.

Test them under ridiculous load. 120% of peak voltage, huge ripple, let them get nice and toasty. The bad caps will pop in minutes, a good one can survive for weeks.

The problem is that the prototypes you get from your Chinese manufacturer will be made with top quality Japanese components. They'll switch them for the cheap junk later while you're not looking, about a 10th of the way into your 1 million unit production run.

Interesting post, I never knew about the Capacitor Plague! Will be sure to remember it.

Marginal (although not always illicit; sometimes just marginally spec-ed) capacitors have long been an issue including in consumer electronics.

As an example, the Mac Plus power supply was notorious [well, at least if you were repairing them] for failing. Some people put out instructions for upgrading it by replacing a subset of components including IIRC certain capacitors.

I've also read accounts of people improving the output e.g. of cheaper CD and DVD players, by replacing certain components particularly in the post DAC circuitry, again including capacitors, with better quality/rating components. I've never experienced this first hand, so I can only pass on second hand anecdote.

This is indeed true. Some (all?) amp designs place resistors and capacitors in the signal path. This distorts the sound. For one channel, that would be fine, since the distortion is usually just a phase shift that you won't notice. (Play a song. Then play a song but start it 1ms after you were planning to start it. It will sound fine both times :)

The problems start to show up with stereo, though, where you have two channels to deal with. When you use low-quality components, the phase shift can be significantly different between the two channels. (Components have a, say 10% value tolerance. If one channel is +10% and the other is -10%, you have a problem.)

This difference destroys the phase information between the two stereo channels, resulting in poor "imaging".

When you replace the 10% tolerance components with 1% tolerance components, you lose the phase distortion, presumably increasing sound quality.

(I am not an EE, but I have built a lot of headphone amps. The sad part is how people with $1000 amplifiers often have worse components than an amp built from $30 worth of parts. Ah, the free market...)

Also, I now have a sudden urge to build another headphone amplifier... :)

Glad to be of service. Hmm... and since you seem to know your way around them, do you take orders?

(Another startup born on HN! ;-)

I did for a while. I am not sure I am knowledgeable enough to do a good job anymore, however.

If you are ever in Chicago, we can have beer and I will teach you how to build an amp :)

Far north suburbs. I've been a recluse, lately, but maybe when things pick up for me, I'll see you at one of the HN meetings. (Didn't realize until now that you're in the neighborhood.)

> (I am not an EE, but I have built a lot of headphone amps. The sad part is how people with $1000 amplifiers often have worse components than an amp built from $30 worth of parts. Ah, the free market...)

Don't forget your monster cables. (Or what where they called?)

So basically, this, plus looking at the pics he links from the over clockers forums leads me to this conclusion.

You should buy

     A whole house surge protector
     A really nice UPS supply with a lead acid battery that does line cleaning for your computer equipment.

Doesn't solve the real problem, but at least saves a significant amount of hassle... Why systems survive so well in a proper colo environment over most homes. (Aside from air particles and temperature).

That is a different sort of ripple. Your big electrolytic capacitors are used in the AC/DC conversion where the rectified sine wave is smoothed into a constant voltage.

  |                           |
  |  _   _   _         ===>>  |
  | / \ / \ / \ /             |--------------
  +V---V---V---V-             +---------------

The capacitors store energy during the peaks and return it during the troughs. The ripple is the current in and out of the capacitor which heats them and does "stuff-I-don't-understand" to their chemistry.

The article complains that some capacitors don't last their 4000 rated hours. But even if they do, 4000/24 is 167. All of your electrolytic capacitors are tiny time bombs.

That is some impressive variable-width ASCII art, I must say.

I was going to chastise you for breaking the entire internet by failing to render <PRE> with fixed width fonts, but it looks like the W3C did that in HTML 4.0 when they reduced the fixed width requirement to a "may render in fixed width font".

If you ever see two members of W3C next to each other, please knock their heads together for me.

It looks like lifetime is specified in hours of life at a high temperature - e.g.:

http://www.rubycon.co.jp/en/catalog/e_pdfs/aluminum/e_BXF.pd...

Lifetime is specified at 10000hrs at 105deg C.

As temperature decreases, lifetime appx doubles per 10deg C:

http://www.low-esr.com/endurance.asp

It looks like a cap rated at 4000h @ 105deg C should last around 60 years at 100deg F or so. Anecdotally I have a guitar amp made in the 60s with the original electrolytics still working. Sounds great.

Interestingly enough I have heard in a few places that there is a 'sweet spot' in terms of duty-cycle that extends the life of electrolytic caps... I would that that high-use electronics (or 'always-on' devices) probably miss that spot.

As per my other comment on this thread, you can intentionally decrease the lifetime (aka MTBF) by running the part at 10 or 20% overvoltage, in addition to extreme temp, to create a corner case and stimulate premature failure in reasonable amount of time. 20% overvoltage won't kill the cap, as any part has absolute maximum ratings well beyond what the spec sheet says. Even bad caps have some degree of guard band, else they would never leave the factory.

I have wanted one of these for a really long time: http://www.brickwall.com/

Reply on news.ycombinator.com