EEVblog 1486 - The REAL Truth Why Film Capacitors FAIL!

[EEVblog]

Think you know how film capacitors fail and degrade in capacitance over time? - self-healing due to surges, right? WRONG!

Capacitor expert and AVX Fellow Ron Demcko confirms what's really going on after a teardown of some failed and one good polypropylene X class capacitor.

00:00 - Teardown of a some failed film capacitors
00:52 - Self Healing and drop in capacitance
01:44 - Capacitance Measurements
02:15 - Teardown of a new Suntan brand polypropylene X2 film capacitor
03:03 - Different failure modes based on size and winding pressure
05:52 - Unwrapping the film
07:51 - The film inside a NEW film capacitor
09:19 - Teardown of the FAILED uTx brand heater capacitor with half capacitance
10:06 - How Schoopage and the end pin terminations work
12:23 - Teardown continues...
12:45 - WOW! What on earth is this?
14:54 - Elecami Wolf also did a teardown
16:05 - Teardown continues...
17:01 - Separating the film showing both slef-healing and extensive metal film corrosion
18:29 - Another uTx brand failed film capacitor from a Corsair PSU, with 90% loss in capacitance!
Elecami Wolf Youtube Channel: https://www.youtube.com/c/ElecamiWolf
19:55 - Let's call an expert! Ron Demcko from AVX tells us the REAL REASON for the failure!

[PartialDischarge]

When he talks about corona and “noise” he is actually referring to partial discharges, high frequency currents that flow thru the insulation.
This is highly dependant on temperature and voltage and also construction. Right now partial discharges are not a required type test on capacitors but the thing is that some of them may have these even below the max rated AC voltage, so the industry is happy not talking about it. Designers that choose capacitors for the long term on expensive installations (power inverters...) do test these partial discharges before selecting.

Here’s an old but interesting doc on this
https://ntrs.nasa.gov/api/citations/19940028538/downloads/19940028538.pdf

[Huluvu]

Based on the time to failure most likely the PP foil was already suffering from to high moisture during the manufacturing process. Manufacturers very often use only the 85/85 test which is useless when the moisture was already trapped within the foil.
A proper method would require a well done THB (temperature + humidity + bias) test which basically costs more time = money.

[EEVblog]

Based on the time to failure most likely the PP foil was already suffering from to high moisture during the manufacturing process. Manufacturers very often use only the 85/85 test which is useless when the moisture was already trapped within the foil.
A proper method would require a well done THB (temperature + humidity + bias) test which basically costs more time = money.

Yep, I suspect so too.

[floobydust]

Nice followup on these film cap failures - but missing what an EE should do about it. I've had dropper caps last 1 year, others last 20 years.
I'm not a believer in the AVX fellow's moisture hypothesis as the main failure mechanism, sorry. It seemed like mostly speculation that didn't make sense enough to design something that lasts.

Vishay response: "It {failures} is because of Corona effect. Corona effect is electrical discharge (partial breakdown) which results from ionization of air on the surface or between the capacitor plates in AC voltage applications or in rapid changing DC voltages (pulses). However, film capacitor has the feature called self-healing. After one discharge, one self-healing. As a result, loss increases and capacitance decreases.
When there is too many corona effects accumulated, the capacitor will be failed.
Therefore, if we make internal 'series construction', we can make the voltage between capacitor plates half. It will be much lower than corona starting voltage, in order to avoid corona effect.
F1772 is internal series construction. At last, I suggest you also read our attached application note. Please note two tables in it.
We recommend that for continuous across line application and in series with the mains application, internal series construction capacitor should be used."

Vishay has specialty film caps designed for "series impedance" {dropper} applications that have a third-plate? to half the E-field I believe.
Dave's failure mechanisms chart is from CERN Metallized Film Capacitor Lifetime Evaluation and Failure Mode Analysis 

[Per Hansson]

I agree floobydust, when I looked at this (see link below) I also made a reflection that is worth mentioning again:
When they are saying that for continuous across the line application you should use this new series they are basically admitting that their old series was crap:
They are after all X1 rated caps so that should be their design purpose!
https://www.badcaps.net/forum/showthread.php?t=80208

[Huluvu]

From my experience (28 years in Power supply development)
Enemy No1 is the moisture which is already trapped inside the PP foil
Second is the lead free soldering process
3rd mechanical problems due to preforming or assembling process

[floobydust]

It looks like the "race to the bottom" for film thickness and component size leads us here- failures due to cap manufacturers pushing the dielectric too hard. To the limiting factor of "moisture"? Instead of burning up, the "self extinguishing" seems to cover it up. At least the crappy Rifa paper PME271's did the right thing - burn up to protest the working conditions.

I don't have access to the cap safety standard IEC 60384 and the PP film cap 60384-14 with humidity, voltage proof, endurance etc. tests - but it's entirely possible the standard is flawed. I also see "pick and choose" what sub-clauses and tests you follow, like in the IEC resistor standard, as other loopholes. So it's very hard to tell what the component is truly about for lifetime and endurance.

Kemet's assessment (has pics) RFI-X2-Capacitors-for-High-Humidity-Enviornments is all about making the part fatter for better resistance to humidity. R46 stating "Not for use in "series with mains" type applications."

What's the difference between "across the line" and "in series with mains"? I don't see any from the capacitor's point of view.

[Seekonk]

I asked James Lewis formerly of KEMET about that series statement for the R46.  Somewhere I have his response. In essence these were born to fail and last long enough to get thru EMI testing. I guess he could be honest after he left. Apparently most X2 capacitors are made to this price point.  Yet, most engineers think these are some kind of high standard for serious use.  They are guaranteed to fail in capacitive dropping applications.  If buying off ebay, check the use before date.

[Kleinstein]

Kemet's assessment (has pics) RFI-X2-Capacitors-for-High-Humidity-Enviornments is all about making the part fatter for better resistance to humidity. R46 stating "Not for use in "series with mains" type applications."

What's the difference between "across the line" and "in series with mains"? I don't see any from the capacitor's point of view.

In case of in series with mains there often is an additional resistor to limit peak currents. This means less stress to the capacitor from peak current, but also less external power if a parial discharge started. The self healing usually wants a short but strong discharge, but not a slow, more contineous glowing.  One possible explaination for the crack / trance like damage could be a weak discharge that does not stop but contineously moves, like following a small trapped gas bubble. One all the way though, much of the area is lost and no longer connected.

[Per Hansson]

Out of curiosity I measured some caps out of some old power supplies I have here.
Not a super exhaustive test of course but the first one on the left in the picture was the most interesting:
A Corsair CX400W power supply that had a X2 cap on both the IEC input PCB and then directly after that on the main PCB.
They will obviously have been exposed to them exact same number of mains transients but only the second one in the chain failed!
Bonus point is that the one that had not failed is the exact same type as the one that did fail Dave's units!

[floobydust]

You're not supposed to check those...

In series with mains, the surge resistor typ. 47-100Ω limits current to much less than across the mains.
Voltage seen, transients included are the same. So why are cap manufacturers blaming the application?

"While the electrical stress on components used in series and in parallel with the mains are comparable, the performance requirements in applications in series to the mains are much stricter." 
in other words- you'll notice a low value failed part a dropper circuit, but an EMI cap fading away low value after a some time- nobody really notices.

Any partial discharge events would have high currents because the capacitance is behind it. For the "self-healing" events I have no idea of the duration or current required to vaporize the electrode and melt the plastic too. But these should not be happening in the first place!

[RLP]

I found a failed film capacitor just a day after seeing this video!! It's a 0.47uF X2 class capacitor from the mains input filter of an APC UPS, and it measured about 4nF - so that's about 99% capacitance loss?! This one is nearly 20 years old - so moisture getting in over time seems quite plausible.

Of course, there was only one thing to do with it - tear it apart! Photos attached (albeit not great ones as the lighting is from above rather than behind the film).

This capacitor does also have that same crack running along the entire length of the film! Otherwise it looks like there's more than 1% of the material left, but it's pretty patchy so a lot of the remaining material will be disconnected from the terminals.

It's interesting that in the first photo there is a section of uniform material at the start of the roll. It looks like this is a short section where the metallisation on one side (effectively one capacitor plate) was present but the metallisation had not yet started on the other side, so that region of the film isn't electrically a capacitor! Once the metallisation on the other side starts, there's a pretty obvious visual difference. The fact that it hasn't visibly degraded in that initial region suggests that the visual changes we're seeing are not directly the result of moisture eating away at the metallisation, but rather due to the self healing resulting from moisture having degraded the materials causing some kind of breakdown, or something like that, i.e. if the cap had never been powered up it would probably look perfect all the way through.

[EEVblog]

Comments from Dr. Tomas Zednicek founder of European Passives Component Institute, and former AVX, regarded as the probably the most knowledgeable guy in the industry:

Quote:
I did look at the whole video – congratulation to your teardown skills 😊. Regarding the phenomenon you observed - I see two/three potential issues on the failed cap:

1] uneven metallization. If you look though the unwind foil of the fail cap vs the reference one it looks that the centre section of the failed one looks lighter compare to the edges, this might indicate that this area is thinner compare to the edges. This does not have to be a manufacturing error but even an intension. You mentioned shoopage reinforcement (also in ref here on my web). The target is not only to make the connection more mechanically robust, but make lower ESR and improve performance on the pulse load. You can image this as an “skin effect” at pulse – high freq spike load – the most exposed area is the one closest to the termination lead – and that has to have a lower ESR – achieved by the edge thicker metallization/shoopage reinfforcement. If you go towards the inner side of the cap - centre of the foil width (if you looked at unwind foil), the resistance increase with the length path and thus the pulse will not expose that area. So actually you do not need uniform metallization thickness across the whole foil width, thinner metallization at the inner side is supporting better self-healing for longer life time and thicker metallization at the edge increase pulse robustness. This is a reason why you see defects in the centre of the foil where self-healing / delamination happens first. This type of failure mode suggest that the cause may not be originated by pulse overloading (in that case you would see a burn spot on the edge - as you expected) – but there is a degradation of the foil under a regular (non-spiked) applied voltage. If the metallization layer is thinner (as a manufacturing fault) or improperly designed you may see such type of failures.

2] moisture degradation. Quality of film caps may be also strongly impacted by quality of the encapsulation resin. Indeed – this class of commercial caps are non-hermetic – but cheap and poor epoxy resins used for encapsulation may not protect so well the capacitor core element and even capture some moisture with time. The metallization may be sensitive to moisture and if there is a higher humidity inside it speeds up its degradation processes. As per the dark spot areas on the metalization of the failed capacitors I could guess that some level of degradation present and I would point moisture on the top of likely the cause. Also presence of “lightning like” spark line at the centre of the foil suggests that there may present a moisture / conduction path causing the defects.   

3] combination of 1] and 2]

As often seen on the market I would guess 2] may be the root cause of the failure in this case, low cost unbranded capacitors often use inferior cheap epoxy resins that are sensitive to moisture and longer term it degrades capacitor performance. You can check this easily by testing of moisture robustness of these capacitors capacitance drop at humidity test (even after 500h 40C/90% you may notice differences and even shorter if you would apply 85C/85% load). The conclusion here may be to use safety capacitors from brand manufacturers with trust if you really care about safety.

[EEVblog]

Nice followup on these film cap failures - but missing what an EE should do about it. I've had dropper caps last 1 year, others last 20 years.
I'm not a believer in the AVX fellow's moisture hypothesis as the main failure mechanism, sorry. It seemed like mostly speculation that didn't make sense enough to design something that lasts.

Moisture ingress makes the corona problem worse, it's part of the chart.
Also see Dr Zednicek's response above.

[mawyatt]

Moisture is a real problem we experienced back in ~1970. We had a series of hybrids fail during temperature cycling. At first it was thought due to a failed ceramic cap that had shorted, these were all replaced and the problem reappeared, same short circuit on Vcc line. After removing said capacitor and replacing short was gone, then we tested the removed cap and they were fine.

Took awhile to figure this out, but eventually the problem was traced to the 2 part epoxy encapsulate. The epoxy had already absorbed moisture enough to cause this to migrate to the hybrid surface over the temperature cycling. The hybrids were powered and the condensed moisture mixed with flux residue and assembly elements and wicked under the mentioned SMD capacitor. With a supply voltage available thin metallic dendrites formed under the SMD cap and eventually shorted out the Vcc line. When the cap was removed with a soldering iron, the dendrite would be absorbed back into the liquid solder and disappear from sight

Anyway, the moisture problem with epoxy sealants can also be due to moisture already captured within the epoxy before being applied, and not just after the fact of use and allowing moisture contamination.

BTW this was not a "cheap" brand epoxy, it was from Dow Corning!!

Best,   

[floobydust]

I feel duped by a modern component that purports to be fabulous - "Self-healing properties", "High moisture resistance", "Over voltage stress withstanding" when in reality they are failing en masse low value.
I understand without formal analysis looking at a failure photo is just speculation. CERN, Vishay, research papers show perhaps 4-5 modalities but every one is attributed to a different mechanism, there's no agreement on what is going on - and it doesn't actually matter.

Blaming the environment, instead of the component design and inadequate IEC 60384 standard, is the issue I have.
Of course the 'too thin in the first place' dielectric would be sensitive to moisture. Why ingress is not happening at the ends is another debate, as well as the Zeus lightning bolt fractal down the middle.
 
I've repaired a few refrigerator control boards where the dropper cap failed low value. Despite MOV, conformal coating, and the low absolute humidity. The buck stops with dropper applications causing appliance failures and associated class-action lawsuits. This might be the reason some capacitor manufacturers are saying a peep, offering better parts.
A simple question:  What replacement part to use? It's still a gong show. EE's have to read between the lines and spec "harsh environment X2" or "high stability grade" and it looks best to contact the factory.
Kyocera AVX has two X2 offerings, no mention of "series with mains" suitability or technical papers on the issue.

In a better world, the IEC 60384 committee would get a heads up to improve the existing requirements/close the endurance loophole or create a new class of cap fit for purpose in dropper applications. Hard to trust the people in the industry, they all seem to know these parts are sacrificial lambs in EMI filtering.

[EEVblog]

I feel duped by a modern component that purports to be fabulous - "Self-healing properties", "High moisture resistance", "Over voltage stress withstanding" when in reality they are failing en masse low value.

Seems to be the case!

[rsjsouza]

After having experienced many X/Y safety capacitors failures in the past, I always suspected that this is a CYA operation based on a design by committee. They got into a room and came up with a compromise solution that mildly caters to every manufacturer's pulls and worked with their lobbyists to put that into regulation. Each manufacturer comes with their "secret sauce" and usually have both a "minimally compliant" product line and the "Cadillac model", obviously sprinkled with the marketing buzzwords of "compact, space-saver" for the former and "automotive, harsh environment" for the latter. 

The "self healing" effectiveness (as well as its intrinsic durability) are direct consequences of the small form factor that modern equipment requires - push the limits of material and construction and get a compromise solution. I have used many uncompromised solutions that give the "middle finger" to any form factor constraints and are durable as a rock in the harshest of the environments - one example is the oil filled start/run capacitor used everywhere in the wildest conditions for years (decades?) like clockwork.

BTW, this was a great video overall!