How to best test Electrolytic capacitors for premature failure reduced life?

[userx2]

Hello all,

We have a problem where switch mode power supply output capacitors have started failing prematurely.
It is a random capacitor batch issue that has arisen fairly recently.

This is a 12V power supply module we use from a manufacturer. It is solder mounted onto our PCB.
The capacitors are inside the module 1000uF/16V and there are 2 in parallel.

These output smoothing caps degrade in that their value moves from 1000uF to about zero in less than one year in the field. The power supply then fails by the voltage dropping as soon a the load is switched on.

I am currently at a loss of what we can do about this except for trying to identify a bad batch of supplies before production.

I am currently thinking of cycling the output load (a 12V pump drawing 530mA) ) on and off but that may not be the best way to test this.

I know these issues with the electrolyte quality have been around for a long time .
Has anyone had any experience with bad quality electrolytic capacitors and how to best test for this?


Best regards
X

 

[D.A.(Tony)Stewart]

Do you have a QA dept to manage bad parts? Or are you the design Eng.
Do you have an AQL plan when these types of failures occur?

About 20 yrs ago an epidemic problem occurred with PC PSU's due to a mis-copied electrolyte formula that resulted in outgassing and rapid degradation.

Choose a stress test that accelerates the failure rate. Temperature, RMS ripple current limit, voltage limit, ESR & C aging rate.
Analyze the design test results for operating parameters vs limits and compute the MTBF acceleration factors. You might need to get better parts and disqualify supplier until a fix is found.
Find functional test parameters to correlate with stress tests such as Vpp ripple.
Perform burn-in tests at elevated temperature at 80% of limits for voltage and temperature on test boards with active step loads.
Design test boards with Ripple detection circuits and LEDs/buzzers for sample testing each reel before use with accelerated stress tests.

How does the quality and price compare to this THT part? Kemet 1mF 39mOhm @ 100kHz Lifetime @ Temp. 3000 Hrs @ 105°C $0.2

 

[FvM]

Checking capacitor ratings (ripple current, temperature, specified lifetime) against actual operation conditions would be the first step.

 

[24vtingle]

What is the module P/N and manfacturer?
Also, are you sure the pump isnt back-pumping high voltage on to the 16V rated caps?

Where are they installed?
How hot is it there.
Is there a chance for cooling air to circulate around the product case?
Do they have plastic or metal case enclosure?
Any fans
Any air vents
How hot does it get in the enclosure?

2000uF at 16V sounds well de-rated for your application.
Are the end caps of the 'lytics bulging outwards or not?

You said SMPS, but are you sure its not a 50Hz mains transformer supply? (these have high ripple at low f which harms lytics.

Put a small sense resistor in the 12v line leading to the pump (say 0.1R) ....and show us the scope shot across it when the pump is on full power.
You may have to do a "no dangling ground tail" scope attachment to get less noise.
Do this with the pump constantly on full power...and with the pump repeatedly being switch on off on off .....say every 30 seconds....and show the sense res waveform during the switch ON bit.

The pump may look like a s/c at the moment of power ON...and so a big surge could be getting drawn from the caps at this point...and if the pump is faulty or something, and turns on and off continuously, then such repeated inrush surges to the pump may damage the lytics.

Put a scope porbe on the 12v rail and then power the 12v.....then plug the 12v into the pump....what happens to the 12v rail at that time...do a sample/hold scope shot.....find a way to trigger it on the switch that you use to turn the pump on if you want.

What sort of pump is it.
Does the rail stay up at 12v when the pump is on full power.

When lytics really fail they often spew out their dielectric all over the pcb below them.
You may have to redesign it with the el caps in "More open air". Is the power source the mains?

Do you have thermocouple data of the lytic case top when the pump is on full power?...preferably set up in an thermal environment like what the customer has.

Maybe the pumps are malfunctioning and drawing really big current spikes......it makes you think that because 0.56A sounds low...and 2mF is a decent bit of capacitance.

 

[24vtingle]

Large numbers of companies have put out electrolytic capacitor free power supply designs into the market.....saying that they would pay for themselves as they wont fail....but none of them ever took off...because in most cases when 'lytics fail its more to do with mistakes that have been done in the way they have been implemented.
There are loads of PSUs going strong well over 10 years with lytic cap banks in them.

Dyson brought out a lytic free office lamp.....supposed to take over the world but didnt......it was excellent...but nobody buys this "lytic's are a disaster" story...and often with good reason.

 

[crutschow]

Are the capacitors rated with low ESR to handle the high ripple current/frequency of a switching supply?

 

[dick_freebird]

I suggest also, to test individual mass and reject
any out-of-family low, as likely having vented.
Of course this needs very identical prep otherwise,
lead length, residues and what-not. Not doable
on-board. But maybe for IQC and reclaim / repair.

 

[userx2]

Thanks folks.
I cannot answer any of the questions you all asked as I do not know.
The power supply is a sealed potted module supplied to us by a power supply manufacturer. It is not my design.
It has also been good since 2012 already with not many failures.
Until about last year November, when we started getting the first failures under warranty.
Ww carry the cost of sending technicians out to replace the whole circuit board containing this module. This is a bad situation which I am trying to avoid in future.
The manufacturer has done an analysis and came up with the capacitor failure as the cause of failure. They claim it was likely a bad batch of capacitors. They also claim it is the same brand and part umber they have been using since 2014.
They have however made 3000 power supplies fro that batch. Potentially these will all fail prematurely and it will cost us a ton of money.

Therefore, I want to i introduce some batch testing for this to avoid this type of issue.I can only really apply and switch loads to it and raise or lower temperature.
Hence my question about the best way to try and endurance test this


BTW, this is not an uncommon problem since I have encountered it in various power supplies from time to time.

EDIT: I just saw a picture from the mfg of some failed capacitors. I can see that their tops has bulged out somewhat.


Best regards
X

 

[D.A.(Tony)Stewart]

Thanks folks.
I cannot answer any of the questions you all asked as I do not know.
The power supply is a sealed potted module supplied to us by a power supply manufacturer. It is not my design.
It has also been good since 2012 already with not many failures.
Until about last year November, when we started getting the first failures under warranty.
Ww carry the cost of sending technicians out to replace the whole circuit board containing this module. This is a bad situation which I am trying to avoid in future.
The manufacturer has done an analysis and came up with the capacitor failure as the cause of failure. They claim it was likely a bad batch of capacitors. They also claim it is the same brand and part umber they have been using since 2014.
They have however made 3000 power supplies fro that batch. Potentially these will all fail prematurely and it will cost us a ton of money.

Therefore, I want to i introduce some batch testing for this to avoid this type of issue.I can only really apply and switch loads to it and raise or lower temperature.
Hence my question about the best way to try and endurance test this


BTW, this is not an uncommon problem since I have encountered it in various power supplies from time to time.

EDIT: I just saw a picture from the mfg of some failed capacitors. I can see that their tops has bulged out somewhat.


Best regards
X

That reminds me of my experience at Burroughs Peripherals Plant where QA would fly & fix globally, and created a lot of frequent flyer coupons. Although I was Test Engineering Mgr., I understand your concerns. It is better to have planned replacements than emergency orders. The assumption is that it is limited to one batch, depends on your statistical quantity of failures and trust in supplier data. I am surprised you still warranty 2014 goods without a maintenance contract. I would want to fly to witness all the failure teardowns at the PSU OEM site and come up with a field test plan to screen boards before they fail. Caps have improved in ESR quality alot since 2014 but I don't know which part they used.

Potted caps have better environmental electrical and moisture protection, but far worse thermal conduction, which is the accelerant of cap. failures with temp. rise from ESR degrading C. I can tell you that RMS ripple current must be significantly derated from the component spec in order to achieve long life >20 yrs for e-caps. I have been in both Aerospace and Commercial arenas where the cost of quality is tightly managed, so I appreciate your concerns. I assume electrolyte reforming failure from long-term non-operating conditions is not a factor.

Do you have a field maintenance or factory test process to measure ESR for premature warning of failures yet?

I imagine an engineering analysis of the potted caps is needed to assess the present reliability risk. My idea is to apply an active step load test to measure the ESR and C from the dynamic error with some internal design tests on loop gain. From that, the peak and average power can be displayed then multiplied by the thermal resistance to get the Cap junction temp rise.

What is your product load impedance? L, R, C?

I would use an active NPN step load to measure ESR, C from abrupt drop and overshoot as a burn in test.

 

[prairiedog]

Using a Taiwan 12V PSU potted module in a product, what I found is they secretly cheapen the components.
You get the "golden sample" with Nichicons, UCC etc and then after a year it's ChongX, and they can reduce the capacitor's values as well to save more money.
They even changed from synchronous rectification over to an ordinary rectifier diode - I caught this with a thermal imaging cam wondering what the hot spot was inside - diode running 90°C across the pads where a mosfet would sit. All this hidden in potting compound.

So always take apart potted modules and never trust what you are getting. The manufacturer can use whatever components they like, unless you have a contract with them explicitly stating details. You can't test-in quality. Possibly measuring ESR but there is likely a PI filter there.

 

[userx2]

Thanks.

@D.A.(Tony)Stewart
No we luckily do not warranty 2014 stuff. There was however never a problem with that and those old units are still out there working.
The warranty issues have started since 2024 November and are growing. The power supply/capacitors fail only after a few months out there so by the time I first heard about it, 1000s of units had already been shipped since.

I also really want to be sure that the replacement PCBs with power supplies will actually last and not become another warranty callout in another few months time.

@prairiedog
Interesting. One has to be very careful with all this but as you said, there is only so much one can do.
In this case, it seems however that the capacitor manufacturer is to blame for suddenly supplying dud capacitors to the PSU maker.
The PSU maker should really QA test that and get it remedied from their side before shipping the PSUs.
I am not sure how they handle that sort of thing yet.

However, since that cannot be guaranteed, I am still going to try and test something on my side.

Best regards
X

 

[24vtingle]

In short, if you are giving this warranty you need to design your own PSU and spec your own electro caps for it...
You presumably have no idea of the installation circumstances of this power supply at the customers' places.?
So they could be putting it in a really hot non ventilated place?
Is the enclosure metal or plastic?

Please could you give the PSU module part number and manufacturer.?
People here can check if its a "reputable" one such as Meanwell, TDK, Murata , XP Power, etc etc

You of course cannot do accelerated life testing on every one of your production units so the only way is to design your own PSU module.
But first find out the thermal environment where they are being installed.? (or if its just an "unknown" environment)

If you do go on to design your own psu module....still test a few of the el caps from any batch...since they may be an old batch that has not been stored correctly....test a few of them for esr as Tony said. Some esr testers will put the right dc voltage on the cap whilst they do the esr test.

I just saw a picture from the mfg of some failed capacitors. I can see that their tops has bulged out somewhat.

OK but it'd be better if you took the dead psu's apart at your place and look for the bulging tops.......what they show may be the bulge after they super-stress tested it or something?

In summary, with a bought in PSU, you will never know what type of el caps the psu manufacturer put in any particular PSU unit...and even if they put "good quality" lytics in it...you will never know if they are from a "good" batch, which has been properly stored till the point of production into the PSU...as such, you must design your own PSU so you can quality test a few el caps from each batch of el caps you buy in.

Incidentally, this is a PSU for a 12V pump which draws 0.5A?..... does it even need lytics?....well...presumably there is a startup surge and thats why you need 2mF?
Otherwise just design your own PSU and just use a few Film caps on the output ...and design a psu that can supply the increased current at startup if thats needed.
What type of pump is it by the way?

How often during the day does it turn on and off?...if regularly then its problems, as start up current surges may be enormous......is the pump and its driver yours?...if so you can soft start the pump so it doesnt draw such heavy surge current at startup.

As you know its also possible to design a PSU that uses a big lytic cap to handle the startup surge, but then gets switched out after that so that a few film caps can take over and supply the operating ripple current. I imagine that ripple voltage rating isnt too important since its just a 12v supply for a pump.?

If you dont know how often per day the pump goes on/off/on..., then you may need to put in a processor to data log the current draw on the 12v rail...so you can see just how the customer is using the product...ie how many surges and how often........then the returned failed unit can be analysed.

You say you had no problems between 2014 to 2022...but the customer may have started using the pump in a different way since 2022....they may have even changed the pump?

Also, if you dont know the surge current at start up of the pump etc etc, and how frequently it turns on and off........then change the board so the psu is not on there...and then you can just swap out the psu (only) when it fails...and leave the board in there.

I would say use a psu with Aluminium Poly caps instead of wet lytics...but wet electrolytics are the masters at being able to handle high surge currents and there may/may not/ be surge currents in your application.

if you like and can get hold of a pump, i can design you a SMPS for it in no time....i would make it high current at startup so it didnt need lytics for start up...but this would need doing in conjunction with having the pump. The prime power source is mains 120VAC? or something else?

If the 12v psu has a high Z low C input, then the pump may make the startup very painful for the output caps as it may just draw high surge for a long time until the output caps eventually get charged up...ideally the pump "waits" till the output caps get up to 12v before it starts?

If the PSU has a tight output current limit, then at startup, the pump could draw surge and hit the output current limit repeatedly.....the pump could then stutter on and off many times before the pump got started, dragging the 12v rail well down each time....and this nightmare startup scenario would play hell with the lytic caps as they woudl be going through multiple startup surge currents for an unknown period.....maybe the PSU manufacturer has tightened the output current limit on this PSU without telling people....and the above is now playing out?....After all, why shouldnt they just tighten the output current limit?, presumably nothing in its datasheet says its suitable for pump useage with high startup surge currents?

 

[crutschow]

I want to i introduce some batch testing for this to avoid this type of issue.I can only really apply and switch loads to it and raise or lower temperature.

A stress-test at rated load current in an environmental oven at maximum operating temperature would help weed out infant mortalities.
This test is normally run continuously for at least a day, and often several.

 

[D.A.(Tony)Stewart]

You can verify a 1000 hr burnin test on a batch and issue a stop ship to purchasing if it fails and get the suppliers attention, and corrective actions.
This needs to be done on a sampling plan and reject failures with an RMA for credit or replacement. Your contract is the P.O. for the part and its spec.

We used to buy lots of HDD's in the 5.25" era which I qualified with a DVT, but you always need to be sharp and reverse engineer their product, so I tried to know their design as well as their QA dept.
From Hitachi/NPL , Maxtor and others , we would get change their customer notice on design and production changes with an NDA. I suggest you qualify another source like Lambda, maybe PowerOne, circa 1999, although they goofed when they transferred to MX with creepage failures when the secondary was grounded and only tested Hipot with secondary floating. So I (with Purchasing) gave them a stop-ship order and a list of failures which they corrected.

 

[userx2]

I am quite sure that the usage of the unit has not changed since we started making it.
The power supply is acutally a dual output in 12V and 5V. They regenerate the 5V rail with a switching regulator internally from the 12V.

We can't easily just get or use another power supply as the form factor/footprint is set.
Also, to use or design another power supply is easily 1 year of a compliance nightmare and also costs 5 figures.
And, there is then NO guarantee that it will have similar issues for some batches.

They are currently still adamant that it was a once off bad batch of capacitors.

A once off can however become a twice off in the future...

 

[24vtingle]

Also, to use or design another power supply is easily 1 year of a compliance nightmare

Depends if you are shipping overseas or not. Also to an extent on what sector.
If you do "EMC due diligence" with good layout and filtering you can happily ship products
without having the EMC compliance officially tested. (Self accreditation)
Specially if its for internal shipping within US....like any country they dont want to screw over their own internal companys as they reap tax from them.

At 6W, you could do a nice home brew with heavily damped FET drive so it was very quiet from an EMC perspective.

And, there is then NO guarantee that it will have similar issues for some batches.

You are referring to doing your own PSU here? (i think you meant "wont" not "will" in your above?)....well if you did that (i know you dont seem to want to) then you could check some of the caps of a batch yourself before they get used in production...with a module you have no way of checking the caps and the same thing can happen to you all over again.

If you dont want to design your own PSU then why not buy a module which only needs you to add the input and output caps...then you can check the caps but still dont have to design the smps power stage.

If you give me your spec (ie whats the input voltage to the SMPS) , ill send you a design with some notes on it so you can consider it for "home brew".
Presumably your Vin is mains 120VAC? , or 2 phase line-line?
So i would do you a flyback. I can break the back of a flyback design in a few mins...ill send you the calc doc.
Transformer takes about 45 mins as need to fiddle with wire lists and core parameter datasheets etc.

Or is the 12v/5v smps switching off another rail like 24V somewhere?
But really i would appreciate to know the current draw of the pump at startup for best design.
I presume the 5V is just <100mA?
Nothing else is on the 12v?

Also, to use or design another power supply is easily 1 year of a compliance nightmare and also costs 5 figures.

OK but how much is the cost of loads of failure returns?
Even if you do Accelerated life testing, its a pain and you cant do it on every production unit...and you may not quite test it hard enough over the short time and you may then still get failures.

 

[userx2]

@24vtingle
Thanks for all that and for offering to help with a design. I won't take you up on that for now.

Our operation is more complex than one may think. Even if I designed a PSU, it would still be manufactured overseas. we do not manufacture any electronics ourselves anymore. We use contract manufacturers exclusively.

The compliance is a nightmare because we sell into a number of different countries. And also the product safety certification is mandatory for our products,
I won't even mentioned EMC as that is not too hard.

Best regards
X

 

[24vtingle]

Even if I designed a PSU, it would still be manufactured overseas. we do not manufacture any electronics ourselves anymore. We use contract manufacturers exclusively.

OK thanks for explaining, i hate to say it, but i once worked at a co that got all manufacturing done overseas......then a mass load of units got made with a "disaster" fault, and the product recall associated with it, put that co out of business. (There was another design issue which added to that to put them out totally but the manufac issue was half of the "kill off".)
The "disaster" fault was an internal mains cable which the ssemblers were putting a ferrite around it...they had to kind of pull the mains wire tight so as to enable insertion of the ferrite, and when they so did, the mains cable snagged a sharp internal corner of the chassis, and smashed through the insulation , making bare mains line wire touch the metal chassis...sometimes it touched and the unit failed outright when On'd...sometimes the bit of vibration from transport etc made the mains wire touch the chassis and it failed with cust.......either way, the reg bodies got onto it and insisted on total product recall.

How the assemblers missed all those strands of bare mains wiring sticking out of that cable i dont know...it was either deliberate or they had been assembling in the dark, or small kids were doing it. I remember seeing the first batch of products which were brought in for checking.....never seen so many pallets of goods in my life.....they were all opened up, and 80% of them showed the fault.

 

[scopeprobe]

Hello all,

We have a problem where switch mode power supply output capacitors have started failing prematurely.
It is a random capacitor batch issue that has arisen fairly recently.

This is a 12V power supply module we use from a manufacturer. It is solder mounted onto our PCB.
The capacitors are inside the module 1000uF/16V and there are 2 in parallel.

These output smoothing caps degrade in that their value moves from 1000uF to about zero in less than one year in the field. The power supply then fails by the voltage dropping as soon a the load is switched on.

I am currently at a loss of what we can do about this except for trying to identify a bad batch of supplies before production.

I am currently thinking of cycling the output load (a 12V pump drawing 530mA) ) on and off but that may not be the best way to test this.

I know these issues with the electrolyte quality have been around for a long time .
Has anyone had any experience with bad quality electrolytic capacitors and how to best test for this?


Best regards
X

It maybe worth looking at the suitability of the parts selected for its application, it sounds like they arn't upto the job which could be due to their quality but could be because of things like their ESR or insufficient ripple rating of them for your application. Also worth checking would be the tracibility data of the parts used for the build if this is availble, the market is awash with counterfeit parts which wouldn't necessarily show itsself until in operation in volume

 

[userx2]

It maybe worth looking at the suitability of the parts selected for its application, it sounds like they aren't up to the job which could be due to their quality but could be because of things like their ESR or insufficient ripple rating of them for your application. Also worth checking would be the traceability data of the parts used for the build if this is available, the market is awash with counterfeit parts which wouldn't necessarily show itself until in operation in volume

Hi yes,

I don't know if these are counterfeit as the manufacturer of the PSU has all the traceability data such as batch codes for the capacitors etc. the capacitors themselves however have no batch markings. the brand of the capacitors is CH.