Let's Talk About Capacitor Failure | Hacker News Reader

Let's Talk About Capacitor Failure

bytecellar.com

154 pointsblakespot109 comments

If you want your aluminum capacitors to fail less often:

- If you are designing the PCB and not repairing, keep them as far away from heat sources as possible. If you're repairing a PCB and the caps are near heat sources, consider isolating them with thermal material

- Use the highest temperature rating capacitor that you can find, if cost is not a concern.

- Most caps now have a lifetime rating such as (4000 hrs in 85C). This lifetime also has built-in ripple current assumptions, but for a first order approximation, consider using the "longest rated at highest temperature" cap, if cost is not a concern.

- Again, as a very rough guideline, use the highest voltage rated capacitor that you can find for your target rail. This is an approximation, because your second goal should be to generally keep the ESR (equivalent series resistance) low, and as the voltage rating increases, so does ESR. Commercial products tend to choose the voltage rating around 1.2x - 1.5x target voltage. Most industrial products that I've seen target a minimum of 2x.

> use the highest voltage rated capacitor that you can find for your target rail

I think you meant to say lowest (or you have some other sign negation in that section, or I misunderstand something)

No, I meant highest. For example: If your target rail is 12V, use a minimum of 16V rated capacitor, but prefer a 25V rated capacitor over 16V.

This application note has a very detailed explanation of Aluminum capacitor reliability (page 12): http://www.cde.com/resources/catalogs/AEappGUIDE.pdf

Thanks. I misread the increase in ESR as a governing effect, rather than a secondary effect.

Nope, higher. Voltage and heat kills electronics. The better the rating, the longer the reliability.

>consider isolating them with thermal material

This only slows down the transfer of heat, both in and out. I'd be surprised if anything useful would come of this as eventually the cap is going to achieve equilibrium with the environment, or if you wrap it up in a small coat, it will take longer to transfer heat out of the cap than normal and run hot.

I thought OP meant use thermal material while doing the repair, such as when using heat guns or soldering irons.

Think a linear wall of isolation, so the capacitor can still dissipate heat to the air.

A jacket around the capacitor would be a bad idea, especially since capacitors can generate heat internally, and heat is what we don't want!

Their lifetime is also non-linear with respect to ambient temperature. A 10 C reduction[1] in ambient temperature roughly doubles the lifetime.

So shielding the capacitor from heat can be very beneficial, even if it's just a few degrees.

[1]: www.rubycon.co.jp/en/products/alumi/pdf/life.pdf

>eventually the cap is going to achieve equilibrium with the environment

the environment is probably around 25 degrees. It's sitting between the environment and some heat producing element, so it will reach steady state somewhere between 25 degrees and the temp of that element. Making it as far away (thermally) as you can from that element should lower it's steady state temperature.

Adding detail about your ESR sentence: use the highest voltage capacitor if it is a low frequency use (e.g. 100Hz). For high frequency use you also must have a good ESR rating - here's the part from the article about that:

"On the other hand, electrolytic capacitors used in switch-mode power supplies (SMPS) operate at very high frequencies and take considerable amounts of ripple current. Every capacitor has an intrinsic loss, and this is often expressed as an ESR. Power dissipated in a capacitor can be expressed as I^2*R, where I is the ripple current in Amps RMS, and R is the ESR. With higher ripple currents, the loss in the capacitor itself can become significant, and cause self heating of the capacitor. For use in SMPS, special low-ESR capacitor types have become popular in the past few decades to reduce this loss and allow capacitors to take more ripple current without shortening the life of the capacitor. However, many of the cheaper brands of caps, most notably those coming from east asian manufacturers, have historically had problems in this regard. There was a period in the late 90’s to early 2000’s known as the capacitor ‘plague’ where many manufacturers went with cheap low ESR capacitors from these manufacturers, only to find many power supplies were failing prematurely (within a year or two of manufacture)."

I remember many years ago a bad batch of dell small form factor machines that constantly failed to popped capacitors. It would always be near the power supply where the heat was greatest. I guess in that certain area the capacitors were outside of their operating range...

Nevertheless, as a bored 16 year old, I enjoyed ripping apart all the machines and calling the higher level support lines to order replacement motherboards.

As time went on, they took my word on if a machine had failed due to this specific issue. I really enjoyed the responsibility and trust. Since I could fix the machines faster than Dells support turnaround, I just asked them to send the parts.

After about the hundredth machine, I never wanted to seat a processor again. This was back when processors had hundreds of fragile pins too!

It’s likely that those experiences really pushed me towards software development instead of IT. Good times...

Totally agree! When Samsung first released their flat screen LCD TVs, ours kept failing after two years of usage and was out of warranty. Apparently a bunch of Korean TVs (including LG) had a bad batch of caps from their vendors. Getting it fixed would of cost $300. Buying my own cap and fixing with my dad cost $1. It was a fun bonding experience too.

Yep! System board part number 38HRF and equivalents -- if I never see another one of those doggone things, it'll be too soon!

I got a few of those Dells for free and re-capped the motherboards with a soldering iron and some desoldering wick. If I recall correctly, the replacement mobos had a chance of having bad caps, too, so re-capping one yourself was the only way to be sure.

Bonus: I never had to re-seat any processors or swap out memory.

> After about the hundredth machine, I never wanted to seat a processor again. This was back when processors had hundreds of fragile pins too!

They still have hundreds of fragile pins. I just bent a few on an AMD Ryzen 7 that got popped out of its socket due to too weak a socket clamp while removing the fan.

You know the saying - ‘The future is already here, it’s just unevenly distributed.’ Turns out, it applies to the past, too.

Intel went to LGA where we the pins are on the motherboard. In my experience they are much less likely to be damaged there as they get protection from the socket.

I’m not sure why AMD stuck with PGA.

Lower cost, and some people are really attached to it because it's easier to repair pins if they get bent. You can straighten out PGA pins with a razor blade or a mechanical pencil pretty easily, LGA pins are... "challenging". It can be done but it's not easy.

(also, to be blunt, AMD processors undergo such extreme depreciation that they're practically disposable. After two years, AMD's flagship 1800X processor has lost 2/3 of its value, a nice high-end mobo like a C6H is literally more valuable than the flagship processor you had put on it. So it makes sense to have the processor be the one with the easy-to-damage sacrificial part on it. Intel it's the other way around, the processors are expensive and your mobo is probably the cheaper part to replace if needed.)

WOW you aren't kidding. First gen Threadrippers are cheap now.

I've seen the 1950X as low as $450 at Microcenter. TR4 motherboards are quite expensive, but right now the Taichi is around $260 after rebate at Newegg. Note that not all of the motherboards are designed to handle the higher-TDP 2970WX/2990WX, if you think that's an upgrade you'd make then look for one with a beefier VRM. Also, the cheaper ones lack 10 GbE or some other higher-end features.

You can also find the 1700 as low as $130 if you watch around. Needless to say, if you have any batch-processing type tasks that don't need AVX2, that's a hell of a deal too.

Threadripper and Epyc are LGA though. They probably stick with PGA for Ryzen because of cost.

LGA designs place the quality control onus on the motherboard manufacturer and their supply base. When your marketshare has relegated you to being the "budget-conscious" choice like AMD found itself for over a decade, the last thing you want is the budget quality motherboards 3rd parties are supplying to cut corners and potentially tarnish an already shaky brand perception. I've always applauded AMD for bringing excellent low to low-midrange hardware to market even during the worst of times, but of the many Phenom II, Piledriver, and Kaveri builds I put together for people back in the day, only a handful included a midrange or high end motherboard. Most were built with overall price/performance on a strict budget as the #1 goal and the $55-80 motherboards that ruled that market weren't as lovingly designed and carefully packaged as the $90-130 motherboards that dominated the Intel market at the time. You can find plenty of forum posts complaining about budget Intel boards coming with bent pins out of the box and the usual advice was to just buy a nicer motherboard. Now consider that a "nicer motherboard" would blow the budget for many of these AMD systems and that they often used even lower end boards than the cheapest recommended Intel builds. Manufacturing and making sure a female PGA socket survives packaging and shipping is far more error tolerant than manufacturing an LGA, especially if that LGA would only be made in quantities 1/20th the number of Intel LGAs. In short, I believe AMD keeps using a PGA expressly to avoid these issues given their market position.

> I just bent a few on an AMD Ryzen 7 that got popped out of its socket due to too weak a socket clamp while removing the fan.

Twist/shear the fan off instead off pulling straight.

Those were probably the Dell Optiplex GX280's.

Horrible design, internally anyway. Caps near the power supply or CPU heatsink were prone to failure. At the beginning of my IT career I swapped many of these motherboards. Fortunately the power supply and motherboard were easy to remove and replace.

I think this was a Pentium 4 machine which also was notorious for running hot.

Saw it with Dell Optiplex GX 270 as well. I supported between 50 and 100 of those at some time and we had to swap 30 motherboards in 6 months.

More info used to be available here: https://www.badcaps.net/, haven't verified recently.

One thing I've heard recently though is that Dell still makes batches of bad hardware and their service is stil (or again) bad. In particular a recent xps batch has had a 50% failure rate while Dell support keep blaming the users IIRC. Source: sysadmin friend of mine.

PS: once you get through, their technicians used to be great.

https://en.wikipedia.org/wiki/Capacitor_plague

A story of industrial espionage, betrayal and an incomplete electrolyte formula.

'Industrial espionage was implicated in the capacitor plague, in connection with the theft of an electrolyte formula. A materials scientist working for Rubycon in Japan left the company, taking the secret water-based electrolyte formula for Rubycon's ZA and ZL series capacitors[citation needed], and began working for a Chinese company. The scientist then developed a copy of this electrolyte[citation needed]. Then, some staff members who defected from the Chinese company copied an incomplete version of the formula[citation needed] and began to market it to many of the aluminium electrolytic manufacturers in Taiwan, undercutting the prices of the Japanese manufacturers.[1][42] This incomplete electrolyte lacked important proprietary ingredients which were essential to the long-term stability of the capacitors[4][23] and was unstable when packaged in a finished aluminum capacitor. This faulty electrolyte allowed the unimpeded formation of hydroxide and produced hydrogen gas.[36]'

Huh, I did not know this was the reason behind it.

I'm nominally in the electronics industry this was a big deal in part because Chinese companies were also feeding counterfeit parts into the supply chain.

This page doesn't talk about it, but what I recall is that the stage was set for this by a move away from a fire retardant (containing bromine) that was linked to developmental disorders. To eliminate the fire retardant, they also moved to a new capacitor chemistry.

That would represent a pretty big spike in demand from manufacturers. New product, possibly new partnerships.

That would be interesting because it would represent a second case of a safety oriented change causing a significant reduction in product quality in as many years. The other being the switch to RoHS lead free solder that caused a great many BGA failures for companies like nVidia.

The failure in implementing lead-free strategies is really interesting, because the sole reason lead was put into electronics solder was that it prevented the tin whisker growth pure tin solder or silver solder experienced.

So some odd-seventy-eighty years later people remove the lead and substitute it with "straight nothin'" and are subsequently surprised they get tin whiskers again?! WTF?

I thought it also made the solder slightly more malleable, so a flex in the board would be less likely to crack the joint. I understand that the nVidia problems were due to using such a large BGA that the thermal expansion of the chip was enough to crack the lead-free soldier after a number of cycles. With the old leaded solder you could get away with it, but the new stuff wasn't properly lifecycle tested before being integrated into the production lines.

Lead is added to solder because it lowers the melting point. It forms a eutectic alloy.

Tin whiskers has nothing to do with it.

>Tin whiskers has nothing to do with it.

This is not accurate. Tin whiskers are a well-known problem resulting from pure tin coatings (although the mechanism is not well understood) and lead has historically been added to mitigate their growth (among other benefits).

From Wikipedia:

"Traditionally, lead was added to slow down whisker growth in tin-based solders." [0]

From NASA:

"No single mitigation technique provides effective protection against whisker formation except the addition of 3% or more of Pb by weight" [1]

"Suggestions for Reducing Risk of Tin Whisker Induced Failures...1. Avoid the use of PURE TIN plated components if possible...Alloys of tin and lead are generally considered to be acceptable where the alloy contains a minimum of 3% lead by weight...Although some experimenters have reported whisker growth from tin-lead alloys, such whiskers have also been reported to be dramatically smaller than those from pure tin plated surfaces and are believed to sufficiently small so as not to pose a significant risk for the geometries of today's microelectronics." [2]

[0] https://en.wikipedia.org/wiki/Whisker_(metallurgy)

[1] https://nepp.nasa.gov/whisker/reference/tech_papers/2011-kos...

[2] https://nepp.nasa.gov/whisker/background/index.htm

> Lead is added to solder because it lowers the melting point. It forms a eutectic alloy.

Pure tin already has a melting point of 230 °C. Yes, Sn63Pb37 (arguably the best solder alloy ever created) has a somewhat lower melting point at around 180 °C but no one uses such low temperatures except when absolutely necessary.

Being able to make the alloy eutectic is yet-another advantage, but the main concern was clearly tin whiskers, because these were breaking devices in the field.

I find that interesting since Bromine is so often used as a sanitizing agent in indoor pools and hot tubs. I see that it causes central nervous system issues and I'm left wondering why Bromine in a steamy hot tub is somehow acceptable while its use in a tiny capacitor is somehow harmful.

There's a site to go with it too:

http://www.badcaps.net/

It's more of a general repair forums with lots of useful advice now, but there are still plenty of stories about how replacing a few dollars worth of caps saved some $$$ equipment from the e-waste.

As others has mentioned, an ESR is the way go for capacitor diagnosis. Usually you can find the capacitor by determining the symptoms. For instance, if the power indicator turns on then off, it's probably near the power supply and filter circuit. That being said, it can get super dangerous repairing old vintage machines since there can be a huge ass capacitor upstream you do not notice. Always discharge your capacitors before you work on them!!

The ESR meter is like magic!

I have fixed a ton of discarded electronics by replacing bad capacitors I found with my ESR meter. Several years back I bought a whole lot of broken test equipment dirt cheap and fixed about 75% of it and resold for a nice profit. Almost everything was fixed with new capacitors.

A few years back my oven died suddenly. I popped open the cover and went through all the electrolytics with my ESR meter and found a bad one. I had exactly that value in my parts bin, replaced it and the oven worked like new. The total cost of the repair was less than $1.

Any recommended ESR meters?

The blue ESR meter is pretty famous [1].

[1] https://anatekinstruments.com/products/fully-assembled-anate...

Rarely ESR stays in the acceptable range while the actual capacity is widely out of spec.

While the limited lifespan of electrolytics is well known, I'm very curious as to the lifespan of the humble yet ubiquitous 0.1uF ceramic disc capacitor. There must be billions of these little buggers in digital electronics, and I've never seen one go bad from age.

This site claims "For X7R and X5R the loss is calculated at -2.5% per decade hour and for Y5V it is -7% per decade hour." https://www.johansondielectrics.com/ceramic-capacitor-aging-... but that seems very pessimistic.

Interestingly, they also claim the ageing of ceramic caps can be reset by baking them at 150C for a couple of hours. Many electronics board could survive this, which seems to imply that a digital board which used exclusively ceramic caps could in theory last ~forever (> 1 human lifetime).

You probably see more decline in capacitance just by running an MLCC close to its voltage rating than you do for a decade.

Many of the datasheets will show you the derating curve over voltage. Run a 6.3V X5R cap much over 3.3V, and you'll see the effective capacitance decline precipitously - 50% or more.

I can tell you that most vintage radio and audio guys generally don't bother replacing ceramic capacitors even on 70-80+ year old equipment, so my guess would be 'a long time'.

Most of the time the failure mode is excessive leakage. Which often doesn't cause problems. And ceramic and tantalum capacitors tend to heal if running at more than a few volts. I found a cap on battery backed up memory array that was sucking about 50uA. Putting 5V across it 'fixed it'

Just be sure to always, and unconditionally, replace those paper/wax ones, is what I learned.

Just a couple months ago I found a 27" LG LCD TV from 2005 near a dumpster. Cuirously enought the faling part was some 6K 5nF ceramic capacitors in the internal power supply board. All were kind of burnt, some a little, some desintegrated.

Truth be told there were videos in youtube on how to fix the exact same model.

Now the TV is in my parents kitchen.

Is it true that japanese capacitors are the best?

Example ( https://www.hardwaresecrets.com/how-to-identify-japanese-ele... ):

> Japanese capacitors are notoriously known by their above-the-average quality (good electrolyte and good sealing),

Last year I had to buy a new power supply for my PC and some producers mentioned explicitly that they were using japanese capacitors and I just passively accepted it in the same style as "swiss watch"/"french champagne"/"italian pasta"/"german car" (now excluding exhaust system, hehe) ,therefore intrinsecally referring to good quality, but I always wondered "why?".

And if it's true that the "best" capacitors are made in japan, is there any special reason (historical, social, because of source materials, etc...) for that or is it just semi-random (e.g. many japanese that were picky discovered that producing high-quality capacitors was just their perfect meaning of life and therefore covered a previously ignored slice of the market)?

Thx :)

Good quality is directly correlated to good QC (quality control). Manufacturing engineers spend a lot of time to define the process. There are different methods such as six sigma (and the associated continue process improvement) to control and detect the rejects. However, to get there requires a lot of upfront cost and manpower which is why it costs more for better quality parts.

Thx (didn't know about "six sigma")

Do you think that maybe the japanese mentality is more "fit" to achieve such precise/high-quality processes and QC than other people/nations/etc... in the area of capacitors?

EDIT:

> Lean management and Six Sigma are two concepts which share similar methodologies and tools. Both programs are Japanese-influenced,... ( https://en.wikipedia.org/wiki/Six_Sigma )

Puah, really, never heard about this stuff - but isn't japanese mgmt structure famous for being complicated (or is maybe just the "formality" of the mgmt being very complicated for europeans?)?

That page of course references Deming (https://en.wikipedia.org/wiki/W._Edwards_Deming) who is really the root of Japan's post-war success in high quality. There's even an annual award named after him to recognize quality management: https://en.wikipedia.org/wiki/Deming_Prize

Where something is made definitely effects the quality. The cultural effect has a play in the equation but it's also the available resources. An assembly line tends to be similar to a clock. Any piece that breaks in the chain can cause the quality to suffer. Therefore, you need workers who can handle the system and also be "trainable" to grow a product.

As others has mentioned, a lot these techniques stem from Demings. Most agree that he had a huge role from transforming the quality of product manufactured in Japan (synonymous to China today) to the top tier quality they can produce now.

People study for years to be able to be good at root cause analysis and assembly process management. To get a good assembly line processing requires on honing in the details. For instance, the location of the workers relative to each other, where they hand off the parts in the line, and also the location and type of tools can easily effect your reject rate.

It also common that marketing material that says "Japanese caps" means capacitors made specifically by Nippon-ChemiCon or Nichicon which are brands known for selling good quality high end capacitors.

Part 2 re: Tantalum capacitors always being leaded parts isn't accurate, there are lots of SMT tantalum capacitors. Due to it's status as a conflict material, though, and it's unique extra-fiery failure mode when you exceed it's ratings, I tend to stick with other low-ESR chemistries like aluminum polymer electrolytic, ceramic or niobium oxide (much more stable - 95% less likely to catch fire and non-conflict, similar performance).

The commenters point is that even typical SMT tantalums have leads:

- A/B/C/.. type cases have tiny leads going outwards

- more modern cases have metal sheet like leads bent down and inwards.

The thing is that tantalums capacity density is hard to beat, specially if you are cost-restricted (those polymers are pricey).

There’s also a lot of research from NASA about the safety of tantalums, and they actually think they’re fairly safe provided you don’t expose them to voltage ripple (so don’t place them on SMPS)

Totally possible I misread it, agreed about the density and ripple, though due to failure mode I’ve always derated them 50% in designs where I’ll derate niobium or (admittedly expensive) polymer 20% to mitigate the risk of thermal runaway.

> There’s also a lot of research from NASA about the safety of tantalums, and they actually think they’re fairly safe provided you don’t expose them to voltage ripple (so don’t place them on SMPS)

Besides even the shortest over-voltage incidents, Tantalums also don't like high current pulses. Low-ESR power sources with quick ramp ups are a no-no.

Yes, I read a lot about tantalums and their failure modes some years ago because we wanted to set some kind of use policy for our own products.

After a lot of research and testing we ended up forbidding the use of tantalums in our designs and taking the cost hit whenever density was an issue.

It’s a drastic measure but I think it was the right call, our products were a perfect mixture of all the things that tantalums don’t support well.

I have seen a few electrolytics fail with a bulged top in switching power supplies and valve amplifiers, and once on a graphics card.

If you want to see how your old cap is doing

1) learn the safe way to discharge them 2) use an ESR meter to test them in circuit. As to the correct ESR, compare a new cap to the one in circuit.

>learn the safe way to discharge them

According to one electronics professor (prefaced this with "do not do this at home") - all you have to do is short the leads with a screwdriver.

Or really use a resistor and some alligator clips. Of course a screwdriver is easier to handle without touching the leads than those tiny alligator clips...

Similar experience in film/tv industrial lighting ballasts and heads. Had old techs evangelize that kind of thing. A lot of wacky people in that field.

I wouldn’t trust a screwdriver to be safe enough in many of those cases. Some of the caps in those things are massive and hold a charge for ages.

I almost killed myself on one thinking there’s no way there would be any charge left after years being stored in an old uninsulated tractor trailer outside. Boy was I wrong.

We were demoing some old ballasts—essentially crumpling them before recycling to prevent raiders from seizing the equipment (it happened).

Well one coincidentally-aligned swing of the sledge caved the steel chassis in just enough to short a monster cap. The thing exploded on contact. Thankfully those old steel ballasts were tanks and nothing happened. I was sure glad I was insulated at that moment.

Kids, am I right? ;P

Just want to second what others are saying: don’t toy with those things, and don’t make uninformed assumptions!

It’s also good to remember soakage in this kind of applications.

It’s irrelevant in low voltage electronics, but once you get over 1 kV and in the order of tens of mF, you can get some ugly scares.

Ugly scares, indeed.

We did regular high-pot and soak testing. Some of those boxes would give us quite the light show when they went bad. And I mean without a lamp hooked up.

>don’t toy with those things

on a much smaller scale - in about 3rd grade at one point we had that game that took off like epidemic in our school and fortunately subsided quickly too - shocking each other (today you'd call it "tazering":), in an open fight kind of like 2 scorpions or sneaking upon, with the capacitors, wall socket charged (220v in USSR) and if i remember correctly of 10-200mkF (like an 1in thick cylinder or a block up to half Rubik cube size). The shock was profound to say the least and according to some literature seems to be crossing into accidentally deadly territory in unfortunate circumstances. Fortunately no injuries/deaths happened. Happy childhood in USSR :) - many things from our childhood one just cant do today anywhere.

Doesn't the m and k cancel each other out, leaving just F? You must mean µkF == mF?

you're right, should have been "uF" (10e-6), thanks. My mistake - i transliterated Russian label "мкФ" ("mkF") which is abbreviation from the Russian word "mikrofarad".

One thing that makes that capacitor prank significantly safer is that the terminals are so close together so the actual current path through the victim is very unlikely to send any significant current through their heart. Still possible if they touched one terminal to their hand and the other terminal landed on their body or if the victim got one terminal and the prankster accidentally touched the other.

yes. Never thought about it. Frequently there were wires attached so that the wires would stick out of your hand like 2-8in toward your opponent. Still usually both ends would touch about the same locality on the body.

Big caps even if discharged are dangerous unless you keep them shorted (wrap some wire around the terminals).

Thanks to dielectric absorption, a cap that has maintained a charge for a decent amount of time cannot be fully discharged easily, as it will recover 1-15% of its total voltage to zap you later.

I've had EE professors do more dangerous things, but I concur. Use a resistor! A big cap with a good charge comes into contact with a screwdriver and well, arc flash injuries (and death) are a real thing.

https://www.youtube.com/watch?v=6hpE5LYj-CY

A good old-fashioned incandescent lightbulb with alligator clips is lovely. It's self-indicating if there's still work being done. :)

I used to have a film canister with a big resistor in it, and crocodile clip leads for this job. Like I said though, make sure you are confident you know what you are doing.

Some tricks, fit the clips entirely with one hand, and put your other hand in your pocket. You are trying to avoid a shock across your heart...

Electrolytic capacitors, even high quality ones, will eventually dry out and fail. The older the capacitor, the more likely this is to happen....sometimes they will last longer even with lots of use but..........

> The older the capacitor, the more likely this is to happen

The funny thing is that due to this capacitor plague of the early 2000's, capacitors which are even older than this have a better survival rate. The plot of failure likelihood over time would be something like a line with a weird bulge in the middle.

Only slightly relevant to the topic, there was a capacitor failure in Sydney in 2004 that stopped trains on 6 lines during the morning commute causing 100,000 passengers to be stranded and unable to get to work. Like all complex failures, there was more than one contributing factor, but if the capacitor didn't fail the trains wouldn't have stopped. https://spectrum.ieee.org/riskfactor/computing/it/two-capaci...

I recommend reading the massively informative comments section.

I have seen an (anecdotal) large amount of capacitors failing in consumer televisions. I recently replaced one in a friend's TV. These capacitors are supposed to have rated lifetimes of roughly the lifetime of the TV, so why are they failing in normal use? Is there another Dell situation, where a big batch had slight issues?

The electrolytic cap market supplies a wide variety of quality levels and is permeated by counterfeit products. Ensuring that you actually use genuine components that will perform to spec is very difficult for anything built to a price. The "Dell situation" didn't start or end with Dell.

The capacitor plague was never fully resolved. There are still a lot of attractively cheap but highly unreliable electrolytic caps on the market. Electrolytic caps represent a substantial proportion of the BoM of most commodity products, so there will always be the temptation to take a chance on off-brand caps. Even crappy electrolytic caps tend to outlive the warranty period and most customers are completely unaware of the issue, so there's very little incentive to spend the extra money on quality caps. In more discerning markets (enthusiast PC components, high-end hifi), Japanese electrolytic caps are a meaningful selling point.

https://en.wikipedia.org/wiki/Capacitor_plague

For that matter a lot of higher end PC components will advertise "no electrolytic caps" or "all solid state caps" as a feature.

Samsung had a batch of these too; I suspect it's a combination of iffy suppliers and designing things exactly to the spec. Thermal design feels like it's often glossed over, especially with cables going everywhere and interrupting airflow.

On the other hand, if it's common knowledge that everyone overbuilds for tolerances on capacitors, it's easier to not feel bad about putting 95C electrolytic fluid in a can marked 105C if you "know" that it's common to bump up the rating by some factor when ordering parts.

I just had my several year old Samsung 65" UHD TV fail from a bad power supply. I strongly encourage anyone able to use a screwdriver to look at the web before tossing a TV or committing to an expensive repair. My case appears to be typical. A couple dozen screws release the back. There is a power supply board somewhat larger than a piece of notebook paper, and two smaller boards which are the LCD drivers and the god awful 'smart' junk they made me buy to get a display. The power supply board is not made by Samsung, you can get a replacement for about $60US. Swapping it is just 6 screws and two easy release connectors.

I went in planning to replace capacitors, but none of mine had visible signs of damage and my board was showing thermal discoloring around other components, so I went with the new board.

Safety note: Turn the thing off, unplug it, and let it sit for hours before you go in. Probably longer than necessary, but I'm sure you have something else to do. These power supplies make >200V outputs for the LED backlights, on the "cold" side of the board isolated from the mains… so the cold side is packing more voltage than the hot side. Remember this if you are tempted to probe a live board.

WRT to the safety note, it's not as bad as CRT TVs where you had to worry about discharging voltages of 25 kV or so, but point taken.

The power supply on my TV died after about 5 years, but since it was an external brick it was trivial to cut the connector off and toss it on an old Dell laptop power supply (the TV wanted 20v, the Dell supply gave 19.5v, but it seems to work fine). I see now that official replacement power supplies are also available on Amazon for $16 if you don't want to hack together a replacement.

Update that no one will read: I went to recycle the old power supply board today and shocked myself twice, then I "petted" it with a strip of aluminum foil and got a really good CRACK out of one of the capacitor banks. So that's more than a week after it was unplugged. I thought those electrolytics would leak faster.

It was in fact a Samsung TV

>I have seen an (anecdotal) large amount of capacitors failing in consumer televisions. I recently replaced one in a friend's TV. These capacitors are supposed to have rated lifetimes of roughly the lifetime of the TV, so why are they failing in normal use?

The point IMHO revolves around "What is the expected lifetime of a TV (nowadays)?"

More anecdata, I own a "large" 32" Sony Trinitron that is incredibly heavy and that works just fine (touch wood) since 2002 or so.

I also had a smaller Mivar that lasted more than 30 years (if I recall correctly 32 years 1984-2016).

More recently (like 2013 or 2014) I procured for a friend's project a number (20) of (admittedly el-cheapo) Hi-Sense 32" LCD's, 3 or 4 failed in about 25 months (or one month beyond the 24 months covered by warranty) and 2 failed within warranty period (but the importer/assistance center closed before that anyway), of these most were capacitors related issues (one was simply a cold solder joint). But - besides the repaired ones - the remaining 13 or 14 still work just fine.

My understanding was about a decade ago, they were designed for 5ish years of average use

My 50-inch Pioneer PDP-5060 plasma (720p) lasted from late 2005 to 2014. I felt that was a short lifespan. I replaced it with one of the last plasmas sold, a 51-inch Samsung, and after four years a vertical line appeared. Now there are two. It is failing. That seems short to me. (OP)

My parents still have a 32" Panasonic LCD from the early 2000s as their main TV. The buttons on the remote are worn out, but the TV still works fine.

It has HDMI (720p or 1080i) so they have a Roku connected to it so they can watch streaming services. It has previously been supplied by DVD, VHS and terrestrial TV, all of which are now defunt :D

I've been wondering whether it's worth opening it up, as the inside must be full of dust.

- The lifetime rating of the capacitor is at nominal ripple current and frequency at a given temperature. Exceeding current or temperature or dissipation (due to higher frequency loading) generally reduces life time by a power law. Conversely, lower stresses generally improves life time manyfold.

- Cheap electrolytics will barely meet their spec and will surely not meet their life time.

- No reason for a consumer good to last longer than the average hours used during the warranty period.

High-quality, generously dimensioned electrolytics (that aren't plagued by particular defects) will last many decades.

The weirdest capacitor failure i've experienced was after a single speck of dust ended up inside the control panel of one the CNC machines at work. It landed directly on a capacitor that was part of a circuit for the backup emergency stop system. The system refused to power on because it believed the emergency stop was tripped.

We ended up bypassing the circuit itself directly in electrical cabinet until we got an electrician in who found the problem capacitor. Luckily the actual e-stop still worked, this was a backup e-stop, I guess, I'm still not entirely sure. We ended up having to replace the entire CNC controller for the machine.

I didn't see it in the article, but one way I'm used to finding bad electrolytic caps that don't "look bad" is to smell them. I imagine there's not a 100% correlation, but the ones I'm used to smell like fish when they leaked.

No replacement for real testing/troubleshooting, but it was a good quick/dirty test when I used to repair microwave transmitters for a living. Also got a lot of curious looks while sniffing suspect boards :)

I'm sure many factors go into the life span of a capacitor. The fix for old VW dashboard clocks (at least in our '81 Vanagon): replace the two capacitors. Now, our VW is 37 years old, but it seems to me a clock should be a pretty low-stress environment. Or maybe I'm wrong, I dunno. Why are 40 year old computers still running on original caps, but a 12VDC clock fails in about the same timespan?

Auto power environments are actually very rough -- 30% (or more!) swings in V+ a nominal 12V system, noise injected by the points opening/closing and the 30,000V coil generating a spark, the old mechanical-switch voltage regulators causing a giant surge, etc. Honestly, it's kind of amazing anything lasts in that world.

I go through an in-dash CD player every few years.

(source: restored loads of aircooled VWs over the years.)

I go through an in-dash CD player every few years.

Time to splurge on an aux input for your phone?

Don't forget temperatures that can esily go from -20 to 140f, condensation, dust, vibration, chemicals.

Capacitors fail either open or short.

Failing open will pretty much always not damage any other circuitry.

Failing short will usually not damage other circuitry, since usually a fuse will blow or the power supply refuse to start.

For the above reasons, I wouldn't preemptively change capacitors. From an effort point of view, changing them in the unlikely event of a failure is far less than changing them all 'just because'.

This is one of the major reasons, where so many decent electronics are being consigned to the scrap-heap due to the most innocuous reasons and there are avenues available to bring them back to life.

https://www.avforums.com/threads/yamaha-rxv1700-power-proble...

Maybe. Personally, I try to replace caps, when possible, but oftent times I find that real failure is corrosion of the board, I can do nothing about.

I thought this was well known, but maybe just a rumor. But my understanding is that at one point someone who worked for a capacitor company stole the formula for making electrolytic capacitors and started their own company. Unfortunately they didn't quite get it right, and they had what was called "capacitor plague."

Only time I've seen capacitor failure impact a system is on an old (Win XP) motherboard I got recently, most of the aluminium caps near the RAM slots are leaking, and only 1 of 4 slots work.

Oh and there was also the 33uf capacitor I put in backwards across 12VDC. Venting failed, and now there's a dent in my ceiling.

> I can’t say for sure that I have ever seen an example of a capacitor failure utterly destroying a system.

I suspect the author has never encountered the imbalance failure mode of matched-pair capacitors used in high-power weapon systems.

Sometimes, a capacitor's failure also cause invisible damage to other ICs, it's pretty annoying since some ICs can't just use volt-meter to check if it is still functional.

One of the things I hate most about electrolytic caps is the failure mode that drips corrosive liquid over your PCB. Not only does it fail, it makes sure to take the rest of the board with it.

This is a treasure hoard. So much accumulated experience on offer, just for the time it takes to read it.

Tempest 3000 eh? I build Zarjaz Rotary controllers for that thing if you’re interested. ;)

Funny story actually...

Sounds like you saw my blog post elsewhere on the site about Tempest 3000, so you surely saw my rotary. I can't recall...do I have a Zarjaz? Maybe an Opal. Hmmmm. Nice game. Rotary or nothing!

Reply on news.ycombinator.com