Capacitor Quality

[tautech]

I'm almost certain that you are wrong.
The "bad capacitor" problem was a real issue in the industry:
https://en.wikipedia.org/wiki/Capacitor_plague
The major brands were not affected.
And who discovered the problem? And ex-scientists form Rubycon.
http://www.molalla.net/members/leeper/alumin~1.pdf

It is somewhat ironic when you link this thread with the last sentence of the PDF article above:
Teapo suspects it is an attempt by Japanese aluminum capacitor suppliers to gain market share, and that the unfortunate incident involving a few companies will unfairly categorize the entire aluminum capacitor industry in Taiwan.

Coincidentally, your immediate suspicion of Jamicon branded caps in episode #824 (14:30) reflects that same fear. Still from the same article:
Further intelligence revealed that Kamei/Jamicon sources its electrolytes from Sanyo in Japan, and likely was not affected by the faulty electrolyte supply.

This^
And from the Wikipedia article:

Industrial espionage implicated
A major cause of the plague of faulty capacitors was industrial espionage in connection with the theft of an electrolyte formula. A materials scientist working for Rubycon in Japan left the company with the secret electrolyte formula for the ZA and ZL series of Rubycon and began working for a Chinese company. The scientist then developed a copy of this water-based electrolyte. After that, some staff members who defected from the company copied an incomplete version of the formula and began to undersell the pricing of the Japanese manufacturers with this electrolyte to many of the aluminum electrolytic manufacturers in Taiwan. The subsequent electrolyte produced lacked important proprietary ingredients which were essential to the long-term stability of the capacitors and was unstable when packaged in a finished aluminum capacitor. The bad formulation of electrolyte allowed the unimpeded formation of hydroxide and produced hydrogen gas.

And:

The report of Hillman and Helmold proved that the cause of the failed capacitors was a faulty electrolyte mixture used by the Taiwanese manufacturers, which lacked the necessary chemical ingredients to ensure the correct pH of the electrolyte over time, for long-term stability of the electrolytic capacitors. The further conclusion that the electrolyte with its alkaline pH value then had the fatal flaw of continual growth of hydroxide without conversion into the stable oxide, was verified on the surface of the anode foil both photographically and with an EDX-fingerprint analysis of the chemical components.

Yet to be discussed is HOW cap ratings are applied in designs today.

It is my strong opinion that most componentry is used in excess of safe real world values. In that I mean not using a part, any part, at close to any of it's maximum ratings.
We commonly see failure of all sorts of components in all sorts of equipment, many for just a few cents more would have been avoided with a better rated component.

Where do caps fall in this; it is my opinion that they should be given more working tolerance below their maximum ratings, look at designs of yesteryear where it was common to see additional capacitance and voltage headroom used. Not any longer. 
Obviously the reason is cost, the point being IF so called 2nd rate caps were used conservatively (and ALL should), would we have the concerns over cap brands that we do today?

So where do you set the maximum value of any specification for capacitors in new designs as a proportion of Datasheet maximums?
65%, 75%, 85%, 95% ?

Will it have a bearing on expected life? You bet you it will.

Maximum ripple capability is rated at a certain frequencies.

Fixed that for you. 

[funkyant]

A lot of you probably are aware that there's a whole forum dedicated to the pursuit of identifying crappy caps to avoid. Tales of Electrolyte recipe espionage and everything (ha!):

http://www.badcaps.net/index.php?pageid=about_us

[jitter]

I wonder how this discussion of A-brand vs B-brand caps is different from audiophile vs standard caps.

Sure rubbish C-brand caps in a hot environment fail quickly. I have torn down several failed power supplies (plugpacks and built in) which had C-brand caps in them. But they also were cramped designs running way too hot for long life. But I have yet to come across failed B-brand caps. Even my PC PSU bought during the capacitor plague era and filled with the now dreaded Teapo caps still functions OK, and it has seen quite some hours of service.

In my work in industrial electronics I mostly come across A-brand caps, no doubt.
But what are A-brand caps? For some reason, the ones mentioned all the time are Nichicon, Nippon Chemicon, Panasonic and Rubycon. But over here, you'll also come across Vishay/BC (formerly Philips) a lot and are considered A-brands. Would that brand be considered an A-brand in your part of the world?

[funkyant]

In the end, the best people to ask about caps are not design engineers - it's the repair techs.

[EEVblog]

Basically, any professional engineer knows there is some risk with going with that new brand based on some accelerated life cycle testing on a small sample.

A lot of engineering is about lowering the risk and using "best practice". It's like the old "Nobody ever got fired for buying IBM/Fluke/Keysight etc".
In the case of caps, no engineer ever got fired for designing in Panasonic caps if it's within the budget. It's smart design.
If for some reason those Panasonic caps went tits-up because they had a production problem, then no one is going to hound you for the choice of using those Panasonic caps. If you used CapXon, well, good luck defending that choice.

So "best practise" is really about making sure you cover your arse with the same sized pillow as everyone else? Sigh.

 

Professional design engineering is all about tradeoff's. In fact it's almost part of the definition of practical engineering.
No engineer has infinite time and money to optimise every single component in the design, and test/characterise/lifecyle test every component in the design.
So what do you usually do for good design practice? Design specs aside, you chose the best quality parts you possibly can based on industry experience and brand reputation, that is possible within your budget.
In fact part of good experienced designer is knowing about brand reputation etc and applying that accordingly.
If you aren't doing that, and are just blindly following datasheets specs as if all manufactures and their parts are all created equal and have potentially equal quality , then I'd say that you aren't a very knowledgeable design engineer.
Indeed, if you sat in a design review meeting for a design where reliability was paramount, and tried to argue that you should use CapXon caps instead of Panasonic, then after being met with snickering and looks of incredulity, you'd get a very stern "prove it".
No that may not be "fair", it may not even be "right", and the no-name caps might perform just as well as the name brand ones, but would you stake your professional reputation on it?
If a no-name cap company wan't to earn their place in "the list" of proven quality cap manufacturers, then it's up to them to do that.
Welcome to the real world.

[EEVblog]

I sense a new movement happening - Capacitor Justice Warriors (CJW's)
They could be a new plague (pun intended).

I suggest they join forces with the Uni-T Justice Warriors (UTJW's). Strength in numbers 

[iampoor]

Interesting how the supply uses Analog Devices parts, delta fans, Omron Relays, A Caddock precision divider, and then......5 lelon caps?!? Doesnt seem to me like they were really trying to save any money....besides those caps and steel chassis. (Im 10 minutes into the video). Im not sure if that means those caps have been properly speced, or they just have a nutso purchasing manager. God questin for Eric next time.     

[wraper]

Good brand caps are sometimes just good looking porns, not essentially good in products. For instance, a 85C NCC cap rated at 2khrs won't outlive a 105C CapXon cap rated at a same 2khrs.

BS, if they are both operated in not very hot environment, NCC cap most likely will outlive that Crapxon with a huge margin. Some series of crapxons bulge even when just sitting on the shelve. Temperature rating should be taken with a grain of salt, it not indicative of quality at all.

[EEVblog]

Interesting how the supply uses Analog Devices parts, delta fans, Omron Relays, A Caddock precision divider, and then......5 lelon caps?!? Doesnt seem to me like they were really trying to save any money....besides those caps and steel chassis. (Im 10 minutes into the video). Im not sure if that means those caps have been properly speced, or they just have a nutso purchasing manager. God questin for Eric next time. 

It's likely that the designer did not specify a brand. The BOM could simply have "1000uF 10V RB 5mm pitch Electrolytic".
And/or Siglent get a good deal on Lelon caps and simply uses them in a ton of products.
Indeed, the BOM could actually specify a specific Lelon part number, and the designer simply picked that part from the company approved schematic parts list.
Whether or not someone at Siglent actually sat down and thought about and/or tested or investigated Lelon, and/or approved them and/or made any sort of cost/quality tradeoff is possibly, but I suspect more unlikely than not.
Maybe there is some really pissed off engineer at Siglent who's a Panasonic fanboy and speced those in, only to be usurped by the purchasing department and the powers-that-be.
Can be a whole host of reasons and scenarios.

[Zero999]

LOL once I was desperate for a capacitor to use a in a prototype, so I pulled an old Crapston capacitor from an old monitor, tested it and it still read the full capacitance so I used it and it worked fine. Of course if it was for production, then I'd use a decent brand capacitor. I don't see the point on saving a few pence on a capacitor, only for it to go wrong and cost a lot more in future.

Despite their bad reputation here, some of the most reliable capacitors I've seen are solid resin dipped tantalum. At work we have plenty of them on PSU boards which are over 30 years old and have been running nearly 24/7 for all their life, without a single failure.

[iampoor]

Interesting how the supply uses Analog Devices parts, delta fans, Omron Relays, A Caddock precision divider, and then......5 lelon caps?!? Doesnt seem to me like they were really trying to save any money....besides those caps and steel chassis. (Im 10 minutes into the video). Im not sure if that means those caps have been properly speced, or they just have a nutso purchasing manager. God questin for Eric next time. 

It's likely that the designer did not specify a brand. The BOM could simply have "1000uF 10V RB 5mm pitch Electrolytic".
And/or Siglent get a good deal on Lelon caps and simply uses them in a ton of products.
Indeed, the BOM could actually specify a specific Lelon part number, and the designer simply picked that part from the company approved schematic parts list.
Whether or not someone at Siglent actually sat down and thought about and/or tested or investigated Lelon, and/or approved them and/or made any sort of cost/quality tradeoff is possibly, but I suspect more unlikely than not.
Maybe there is some really pissed off engineer at Siglent who's a Panasonic fanboy and speced those in, only to be usurped by the purchasing department and the powers-that-be.
Can be a whole host of reasons and scenarios.

Very true. Just seems like every other part on the board with a brand name is a reputable vendor! Even the fuse holder looks great! It will remain a mystery. 

[SeanB]

It's likely that the designer did not specify a brand. The BOM could simply have "1000uF 10V RB 5mm pitch Electrolytic".
And/or Siglent get a good deal on Lelon caps and simply uses them in a ton of products.
Indeed, the BOM could actually specify a specific Lelon part number, and the designer simply picked that part from the company approved schematic parts list.
Whether or not someone at Siglent actually sat down and thought about and/or tested or investigated Lelon, and/or approved them and/or made any sort of cost/quality tradeoff is possibly, but I suspect more unlikely than not.
Maybe there is some really pissed off engineer at Siglent who's a Panasonic fanboy and speced those in, only to be usurped by the purchasing department and the powers-that-be.
Can be a whole host of reasons and scenarios.

I would guess they have a list of preferred parts, so use them in design. Could be that they got them in bulk for a good price, and find them reliable enough, or they have had good previous experience of them, or they just had a load left over from a previous production run and are rolling them out as fit for purpose here.

[EEVblog]

So how would one go about accelerated testing caps like this?
Off the top of my head I'd say:
- Chose at least two name brand, and two crap brands.
- Same temp range, rated hours @ similar ESR, size, etc.
- At least 5 samples of each type.
- Only buy from authorized distributors to avoid fakes (maybe not possible for the crap brands)
- Buy each one from 5 different suppliers (Digikey, Mouser, Farnall, RS etc) to hopefully avoid and batch related quality issues.
- Test the ESR and other parameters of each cap before testing. Might as well test everything I guess, cap, leakage etc
- Put them all on the same board with individual and identical bridge rectifiers and loads. No parallels!
- Pick a ripple current to stress the parts somewhat. Normally speced at 100KHz, but testing with normal 100Hz full wave should be fine. Although if you wanted to get fancy pantsy, 100KHz SMPS for each one.
- Pick a voltage close to the max working voltage
- Put in thermal chamber at say 85C
- Run for 1000 hours.
- Test all parameter of every cap every 100 hours, after letting settle to room temp.
- Also try and see any difference in internal heating with FLIR.

[EEVblog]

LOL once I was desperate for a capacitor to use a in a prototype, so I pulled an old Crapston capacitor from an old monitor, tested it and it still read the full capacitance so I used it and it worked fine. Of course if it was for production, then I'd use a decent brand capacitor. I don't see the point on saving a few pence on a capacitor, only for it to go wrong and cost a lot more in future.

And therein lies the dilemma of every design engineer, and probably a good part of the reason why certain brands are "respected", and passed down as such through the generations. Engineers like having known go-to parts/brands, it's part of our nature.
Of course the CJW's will insist that every capacitor/part should be judged on it's merits etc. And yeah, ok, fine, but try and justify to the engineering manager why you want to spend 100's hours testing CapXon caps, when it's well accepted that Panasonic is a respected, safer and easier choice.

An almost identical argument would be for lithium-ion batteries. Like capacitors, it's all about the quality control and design of chemicals and other manufacturing processes.
Panasonic is a no-brainer decision for a quality reputation in batteries. All other things being equal, and cost aside, a design engineer and the "higher-up's" would all be just fine with picking Panasonic batteries without further justification. Suggest WunHunLow and you might have a fight on your hands to get them approved.

[VK5RC]

Isn't the problem the lack of independently verified data about component MTTF. To do it properly is seriously hard, lots of samples, over different production runs, different models, temps, sizes etc. I suspect the firms have the data but keep it very close to their chest, like car warranty claim costs.
 So much of what you hear is hearsay, if I have an important project I spend a little more though!

[SeanB]

Might also want to build a 100kHz power oscillator ( capable of around 5A of current) and put them all in series across the output, and keep them all at the same constant AC current and run at 90C for 85C units, or 120C for 105C units. That way you have the most dissipation in internal ESR, and the higher temperature will provide reduced test time, plus will still be usable as a typical in case temperature.  Just chose those with similar max ripple current and run in a string of mixed types. As ESR will be low you probably could get away with a 12V supply rail as power source, enabling you to use COTS parts for the output stage, along with using a single supply.

[Muttley Snickers]

The particular meter mentioned is supplied with a three year warranty so I would assume from this that the manufacturer has designed for and expects the incorporated componentry to be trouble free for at least that period, notwithstanding components could still fail or go out of tolerance within that period and cause erroneous results which may not be immediately recognised by the user until some strange event occurs.

I dont know whether a device that is out of original specification is fully covered under a warranty agreement, but if the device was so far out of whack after a couple of years then I would hope that most manufacturers would come to the party and provide a repair but this may not be the case in all circumstances.

[firewalker]

So how would one go about accelerated testing caps like this?
Off the top of my head I'd say:
- Chose at least two name brand, and two crap brands.
- Same temp range, rated hours @ similar ESR, size, etc.
- At least 5 samples of each type.
- Only buy from authorized distributors to avoid fakes (maybe not possible for the crap brands)
- Buy each one from 5 different suppliers (Digikey, Mouser, Farnall, RS etc) to hopefully avoid and batch related quality issues.
- Test the ESR and other parameters of each cap before testing. Might as well test everything I guess, cap, leakage etc
- Put them all on the same board with individual and identical bridge rectifiers and loads. No parallels!
- Pick a ripple current to stress the parts somewhat. Normally speced at 100KHz, but testing with normal 100Hz full wave should be fine. Although if you wanted to get fancy pantsy, 100KHz SMPS for each one.
- Pick a voltage close to the max working voltage
- Put in thermal chamber at say 85C
- Run for 1000 hours.
- Test all parameter of every cap every 100 hours, after letting settle to room temp.
- Also try and see any difference in internal heating with FLIR.

There should be on-off session. E.g. once per day. Cut the power, allow everything to cool and restart. To simulate turn on-off events. Daily thermal expansion and contraction could play a role to the life of the component.

Alexander.

[iampoor]

So how would one go about accelerated testing caps like this?
Off the top of my head I'd say:
- Chose at least two name brand, and two crap brands.
- Same temp range, rated hours @ similar ESR, size, etc.
- At least 5 samples of each type.
- Only buy from authorized distributors to avoid fakes (maybe not possible for the crap brands)
- Buy each one from 5 different suppliers (Digikey, Mouser, Farnall, RS etc) to hopefully avoid and batch related quality issues.
- Test the ESR and other parameters of each cap before testing. Might as well test everything I guess, cap, leakage etc
- Put them all on the same board with individual and identical bridge rectifiers and loads. No parallels!
- Pick a ripple current to stress the parts somewhat. Normally speced at 100KHz, but testing with normal 100Hz full wave should be fine. Although if you wanted to get fancy pantsy, 100KHz SMPS for each one.
- Pick a voltage close to the max working voltage
- Put in thermal chamber at say 85C
- Run for 1000 hours.
- Test all parameter of every cap every 100 hours, after letting settle to room temp.
- Also try and see any difference in internal heating with FLIR.

Getting a few fakes would be interesting too. 

[EEVblog]

But if we remove the "reliability is paramount" constraint and accept your subsequent claim "and the no-name caps might perform just as well as the name brand ones" then choosing for no other reason (and you originally offered no other reason)  than protecting ones professional reputation, reduces "best practise" to mean nothing more than CYA. Doesn't it?

No, and I'm not going to try and explain why again.

I also think that there are people in all walks of life who like to delude themselves that they alone don't engage in a good bit of arse covering. Engineers included. They'll heap scorn on managers, on marketing, on customers and suppliers, on anyone and everyone else in order to maintain the delusion. A reality check does no one any harm.

As a professional design engineer for 20+ years I covered my arse with monotonous regularity. Anyone who doesn't is a fool who'll get taken advantage of.

In the real world a big part of best practise is "arse-covering".

Of course it's a part of it, and a very legitimate part of it. In many cases an actual legal requirement and regular part of professional practice.
Doesn't mean there is no quantitative justification for it though.

That's the end for me, I'm done discussing this aspect of it.

[EEVblog]

There should be on-off session. E.g. once per day. Cut the power, allow everything to cool and restart. To simulate turn on-off events. Daily thermal expansion and contraction could play a role to the life of the component.

Can't hurt.
Or maybe it can, and that's the point 

[diyaudio]

WIMA makes good caps.... 

[wraper]

So how would one go about accelerated testing caps like this?
Off the top of my head I'd say:
- Chose at least two name brand, and two crap brands.
- Same temp range, rated hours @ similar ESR, size, etc.
- At least 5 samples of each type.
- Only buy from authorized distributors to avoid fakes (maybe not possible for the crap brands)
- Buy each one from 5 different suppliers (Digikey, Mouser, Farnall, RS etc) to hopefully avoid and batch related quality issues.
- Test the ESR and other parameters of each cap before testing. Might as well test everything I guess, cap, leakage etc
- Put them all on the same board with individual and identical bridge rectifiers and loads. No parallels!
- Pick a ripple current to stress the parts somewhat. Normally speced at 100KHz, but testing with normal 100Hz full wave should be fine. Although if you wanted to get fancy pantsy, 100KHz SMPS for each one.
- Pick a voltage close to the max working voltage
- Put in thermal chamber at say 85C
- Run for 1000 hours.
- Test all parameter of every cap every 100 hours, after letting settle to room temp.
- Also try and see any difference in internal heating with FLIR.

There should be on-off session. E.g. once per day. Cut the power, allow everything to cool and restart. To simulate turn on-off events. Daily thermal expansion and contraction could play a role to the life of the component.

Alexander.

Exactly, but it's not sufficient anyway. Many crap capacitors performed accelerated tests just fine. But when they reached the real world, turned out they fail prematurely. For example Nippon Chemi-con KZG series or Nichicon HM, HN series if talking about good brands. I have read a while ago that high voltage and temperature used in accelerated tests prevented some negative effects which caused capacitors to fail in real use conditions.

[retrolefty]

In the end, the best people to ask about caps are not design engineers - it's the repair techs.

 Ding, ding, ding. We have a winner.

 Now if there was only a decent broad based database of caps by brand repair activity to guide us out of this problem with 'reputation Vs unknown brand problem'. So far there is a large subjective element stated so far by many.

[Seekonk]

Our company had this legacy product.  The product was the responsibility of engineer X.
A customer that operated the product 24/7 on 50Hzsaid  the product would fail after about 2 1/2 years.
It wasn't that expensive and most customers would have just keep a half dozen on the shelf and
plugged in a new one.  He persisted trying to find out the problem. Engineer X thought he was a moron.
When the product was tested at 60Hz it worked fine

Engineer X went on vacation for two weeks and when the customer called again I was told to resolve the
problem.  Testing it indeed showed a problem.  The dropout time of the relay was shoryer than 50Hz.  The
supply cap was open and the shrink label was a totally different color than a new capacitor.  This baby
was hot!  We used a good quality cap and looking at the operating temperature,  these failures were happening
right on the money time wise.  That was reassuring that the calculations actually worked.

We had an ISO9000 quality system and I was a trained auditor.  When I first started working there I had to
get a product number assigned in order to get parts.  We offered a lot of custom variants on standard boards.
I would be working at my desk on the product when production would bring me a box of 25 units that didn't
work.  They had managed to build these, test them, and make labels before I had even designed the unit. 
Production did not need an engineer to sign off before construction.  With that happening four times I had
gotten a reputation of designing things that didn't work.  After that I just stole parts from inventory.
I tried to change the procedure but they fought me tooth and nail. Left three years later with nothing
resolved.

I was responsible for failure reports for products that were returned.  There were no reports done on
new products off the line that didn't work.  One day I was back in repair.  There were 35 units with the
boards taken out.  There was a resistor in series with a pot.  Test procedure said if it couldn't be
adjusted, change it to this value.  That was likely to happen on one of every 200 units.  A transistor
was discontinued and changed to one similar.  For ten years they had been pulling apart 50% of the units
coming off the A resistor change reduced failures to about zero.  line.  Production refused to track new
units that failed.

Management just wanted to buy a quality system, not live it.  When Sales slump the owner would create a new
mission statement.  You need buy in from everyone.  A worker at a capacitor plant is likely to throw his
old apple core from lunch into the vat of electrolyte.