failure/degradation of metalised film capacitors ... interesting

[jaunty]

when working on repairs most people tend to absolutely overlook metalised film capacitors like the ones shown here (myself included) ... not realizing that they can degrade heavily over time (not sure if specifying has anything to do with it?) ... i suppose checking it's capacitance can be a strong clue to this happening. I thought this was interesting and worth of note ... when you can't find other obvious problems in a circuit ... i'm guessing it can happen to X and Y class caps too

[David Hess]

It is even more of a problem with X and Y capacitors where nobody is likely to notice.

As far as the problem, some capacitors are cheaper than others and they use the cheapest capacitors which are also the lowest quality in stuff like this.

[jaunty]

Yes i'm kind of wondering at the moment if i'm not having the same problem on an X capacitor ...

[Gyro]

It's actually a 'mechanism' rather than failure/degradation in most metalised film capacitors (apart from the really crap cheap ones, that is). The idea is to clear a fault that punctures the dielectric. Self-healing is normally indicated in the datasheet. Breakdown is normally the result of overvoltage or severe current transients. It's a desirable behavior in X-caps - you want them to self-heal on transients rather than fail short, as a foil capacitor would do.

Unfortunately, the downside of self-healing behaviour is that it does result in progressive loss of capacitance as the metalisation burns away around new punctures, although it doesn't happen in low power circuits.

[David Hess]

The problem is that the cheap ones self heal because they progressively deteriorate.  Good quality X and Y class capacitors self heal when damaged and bad quality ones continuously self heal or otherwise degrade over time.

This problem cropped up in audio coupling circuits starting decades ago where low quality metalized film capacitors produced a noisy "scratching" sound.

[Gyro]

This problem cropped up in audio coupling circuits starting decades ago where low quality metalized film capacitors produced a noisy "scratching" sound.

That's why I like former Soviet hermetic PIO. 

[SeanB]

You get it with motor run capacitors as well, though there you find your nominal 50uF capacitor drops down to around 30uF before the compressor motor stops being able to start reliably, and then cycles on the overload cutout till the motor itself fails short circuit and blows the internal wiring apart. Generally I test these capacitors with an insulation tester, and you can hear the capacitor self healing as faint crackling noise coming from the capacitor as the self clearing takes place. Insulation resistance wise they are poorer than new, but still well above around 1M, but when they become noisy you know it is time to change them. New capacitors do not break down at 1kV Dc, with the 440VAC rating they have.

Funny enough I have some old PCB oil filled capacitors, around 50 years old, that, despite being in service for decades, still meet the capacitance value on them, and do not break down at 1kV. Comes from being sealed units, and being metallised paper units with a good oil fill. They are however three times the size of the equivalent palm oil filled metallised foil replacement capacitors.

[floobydust]

There are stacked and wound constructions for film caps.

I thought "seal healing" was marketing bullshit from a very long time ago. My reasoning- once a breakdown and arc starts, if there's lots of follow through current, the cap burns up.
You have metal film and plastic so I'm struggling to see how the heat "clears" the carbon and metal vape etc.
But this 2011 APEC presentation says I'm wrong:

"...Metallized film capacitors utilize a deposited metal (aluminum or zinc) that is only a few hundred angstroms thick. This compares to a thin foil electrode of typically ranging in thickness from 0.20 to 1.0 mil. The metallization is so thin it allows for “clearing” and avoidance of short circuit failure mode."

APEC Film cap pdf

Film cap failures I have seen are due to the metalization open-circuiting on the end electrodes, on stacked types. Apparently stacked types are also "self-fusing" page 8.
I'm still a skeptic that an X-cap is going to quietly "heal".

[Gyro]

I'm still a skeptic that an X-cap is going to quietly "heal".

No, it does work, I've sacrificially tested them. Wind the voltage up slowly from rated peak and they start ticking, keep going and the capacitance keeps going down and the 'proof voltage' goes up as you knock out the defects. If you really push it, they swell a bit.

With a non self-healing film, you get one tick and it becomes a low value resistor.

The exception of course is the old Rifa Metalized paper once it gets moisure ingress... they go into self-heal meltdown, or rather, burn-down!

[T3sl4co1l]

Really interesting is when you find parts specified for CFL starter use.  These are subject to high voltages and currents.  The datasheet shows typical capacitance increase versus surge time beyond rated voltage.

FYI:
What a CFL circuit is is, a BJT half-bridge, self excited (driven by output current through a transformer).  The resonant load is the CFL tube's filaments in series with the capacitor and an inductor.  When cold, the filaments have very low resistance, so the Q is high and the output current and capacitor voltage are large.  Because the filaments are wired to either side of the capacitor, the peak voltage can cause breakdown, starting the tube's discharge that much sooner.  As the filaments heat up and emission begins, the Q drops, and as the plasma discharge takes over, the voltage across the capacitor stabilizes to about 100V and current is limited by the inductor.

So, CFLs being as bargain basement as possible, they use as little capacitor as is necessary, and having caps rated for number of starts is a valuable spec.  Other common fails, open filaments (no load current, it simply doesn't start up at all), dead transistors, dried electrolytics, fused resistors (they don't use fuses in these, at least not the common smaller ones; they use fusible resistors).

Tim

[floobydust]

I'm still a skeptic that an X-cap is going to quietly "heal".

No, it does work, I've sacrificially tested them. Wind the voltage up slowly from rated peak and they start ticking, keep going and the capacitance keeps going down and the 'proof voltage' goes up as you knock out the defects. If you really push it, they swell a bit.

With a non self-healing film, you get one tick and it becomes a low value resistor.

The exception of course is the old Rifa Metalized paper once it gets moisure ingress... they go into self-heal meltdown, or rather, burn-down!

I have seen electrolytic (photo-flash) capacitors do the same - break down and arc inside, making the tick sound. But no big drama.

Those Rifa PME271 caps are still for sale  I wrote a flaming (pun intended) email to Vishay asking why they still sell them, no response. But Kemet has the lineup now.
They are also claimed to be "self-healing" so something happens where the metal vaporizes and doesn't nicely disappear. Perhaps it's the paper.

I think if you have lot of energy behind the HV, a partial discharge should burn up the whole capacitor.

[Gyro]

I think the key reason the PME271 metalized paper caps are still sold is that they have a higher dV/dt withstand rating that plastic film types.

I suppose, to be fair, the ones we typically see burning up tend to be 'older' ones during equipment repair/restoration, that have cracked and suffered moisture ingress. They certainly don't age gracefully though.

I had a 2uF one as the main discharge capacitor in a CD electronic ignition kit years ago. It served me well on three cars before it eventually caught fire and took the PCB with it!

[David Hess]

Paper (cellulose) has some desirable characteristics compared to plastic film for self healing capacitors.  It produces less free carbon when vaporized and does not soften or change structure when heated so it is easier to make a safety capacitor using a paper dielectric.  I have seen some plastic film capacitors in line applications catastrophically "extrude" themselves.

[T3sl4co1l]

Those are fun, I had an application with 400kHz ripple that melted those DC link caps (MKP style, non X2 rated).  The body swells, the bottom encapsulation cracks, and the molten guts extrude out against the board.  Eventually the layers short out and you get a little fireworks to go with.

Polypropylene has a somewhat low temperature limit, but it's better not just because of losses in general, but losses at temperature, compared to polyester (the most common material, in MKT style and most general purpose types).  Polyester, when you run it too high, it doesn't swell, it melts at a local spot and burns a hole through the body.  In other words, loss has a positive tempco so you get hotspotting and runaway.

Tim

[floobydust]

Paper has a rough surface compared to plastic film, so the metallization might have some dendrites or grow whiskers with age?
I don't think the Rifa failures are about moisture. The epoxy case shrinks and cracks too.

It takes me back to very old paper and wax and foil capacitors, the trendy low voltage "leakage testing" supposed to predict emminent failure of these vintage capacitors due to "moisture". As if it's raining inside these capacitors.

I don't think we know enough about what goes on inside to make either construction fail with age. Have to do a TEM or something to really know.

[Gyro]

Fair comment. Yes, it looks as if the epoxy cracks with age, either from shrinkage, or maybe even from internal expansion due to self healing (although it seems to happen with the old unused ones I have in a drawer too).

The cracking presumably exposes the winding and dielectric to the atmosphere. The destructive atmospheric component has been presumed to be moisture, I guess it could just as easily be Oxygen or something else.


Edit: Surely dendrite formation would be cleared by the self-healing mechanism though.

[Seekonk]

As a general rule, X capacitors should not be used in voltage dropping supply applications. They will fail over time.  KEMET even puts warnings to this effect in their data sheets.

[floobydust]

Kemet: "Please note that the popular X2 Series “R46” is NOT suitable for series applications." But no explanation why 

I use X-caps for capacitive voltage droppers because they are rated and tested for mains overvoltage transients. In most circuits, the peak currents are lowered by use of say 100R series resistor so the fuse does not blow if you plug in at Vpeak.

I had to laugh about the Rifa 271's: "Best flammability performance" Yup that's proven lol

T3sl4co1l mentions the dielectric heating film caps can experience in high-freq high power use, but at 50-60Hz for a few uF I think it's very low.

[T3sl4co1l]

T3sl4co1l mentions the dielectric heating film caps can experience in high-freq high power use, but at 50-60Hz for a few uF I think it's very low.

Right, not an issue here.

On the same project, I did wonder if the EMI filter had any problems.  I don't know that it ever got warm, but it did have several amperes of switching ripple into it (out of a 30A mains), so it wasn't nothing.

The more common design case where this would be a concern, is on a PFC input stage, where 100% of the PFC converter's ripple goes into the EMI filter.  You'd just model the filter and PFC together to see how that works out.  Usually it's a few amperes maximum, into a relatively large (1-2uF?) cap, that probably can't be a bargain-basement model, and that's pretty much it.

Tim

[xavier60]

Kemet: "Please note that the popular X2 Series “R46” is NOT suitable for series applications." But no explanation why 

I use X-caps for capacitive voltage droppers because they are rated and tested for mains overvoltage transients. In most circuits, the peak currents are lowered by use of say 100R series resistor so the fuse does not blow if you plug in at Vpeak.

I had to laugh about the Rifa 271's: "Best flammability performance" Yup that's proven lol

T3sl4co1l mentions the dielectric heating film caps can experience in high-freq high power use, but at 50-60Hz for a few uF I think it's very low.

A common fault in older Yamaha receivers is the dropper capacitor that powers the control circuit for the standby power supply failing, typically dropping to 1/10 of the rated capacitance. And in this case there is a 2.2K resistor in series with the capacitor which would be expected to mostly absorb spikes.
 Myself and many others have used various brands of mains rated capacitors, even rated to 275vac, to have the same failure occur years later. Really weird.
 I'm thinking about using ceramic X or Y class capacitors if there is ever a next time.
https://www.avforums.com/threads/yamaha-rx-v459-power-problem.2047964/

[Seekonk]

Kemet: "Please note that the popular X2 Series “R46” is NOT suitable for series applications." But no explanation why 

I use X-caps for capacitive voltage droppers because they are rated and tested for mains overvoltage transients. In most circuits, the peak currents are lowered by use of say 100R series resistor so the fuse does not blow if you plug in at Vpeak.

I had to laugh about the Rifa 271's: "Best flammability performance" Yup that's proven lol

T3sl4co1l mentions the dielectric heating film caps can experience in high-freq high power use, but at 50-60Hz for a few uF I think it's very low.

I have a box of 2,000 of them and I use them for whatever. With so many spares, what do I care. But, I asked Jim Lewis (formerly of KEMET and featured video of eevblog) about this and he said that X2 parts are very cheaply made because manufacturers only want them to last thru EMI testing.  Many would be ineffective in just a few years because of moisture getting in. Evidently that which makes them safe also makes them fail, extremely thin metalization.  He says they corrode quickly.  He said they were coming up with a line of X2 caps that could be used for voltage dropping.  Evidently because so many people think it is such a great idea.  here was one poster here that was replacing quite a few X2 caps used as voltage droppers. Even his replacements failed in a few years.

[floobydust]

Vishay App Note AC Film Capacitors in Connection with the Mains attempts to discuss parallel verses series connected X-caps.

It just added further confusion, as Vishay says for continuous 24/7 across-the-mains use for years, use particular X2 or X1 caps, "high stability grade" which is one model the F1772, which has three versions, one of which F1772S X2 is suitable for series (dropper) use. 


For the failure mode of capacitance value dropping with age, it seems to be mainly the metallization on the electrodes vanishing.  "electrode corona demetallization" at thin dielectric sections, and also from corrosion.
It is aggravated by humidity:
"... moisture has several effects:  first, it decreases the electrical strength of the gas in the case of oil-free capacitors, leading to corona demetallization of the electrode, and secondly it corrodes the electrode. When moisture is present in the dielectric film, the loss factor is increased, because of the presence of water dipoles, and the insulation resistance is reduced, leading to current leakage and generation of heat."

Metallized Film Capacitor Lifetime Evaluation and Failure Mode Analysis paper.
SEM Lab capacitor failure analysis service has neat TEM pics.

[serg-el]

[dietert1]

It's the fault of the circuit designer, who imagined the mains input were a sinewave and it wasn't. Nowadays it comes with a lot of noise that you hardly see on the scope because it is only some volts. Each and every capacitor connected to mains needs a protection device to avoid excessive currents at high frequencies. That resistor or NTC or PTC or inductor can be small like 2 or 3 Ohms, but it has to be there. Without, the capacitor will continuously take high current spikes in the KHz to MHz region and it is not made for that.
Also when designing filter circuits everybody should ask themselves whether there shouldn't be some resistors to absorb energy. High Q parts are potentially dangerous.

Regards, Dieter

PS: Another, much larger current spike happens when you turn on the circuit at the wrong time. Then you may get a transient of several hundred volts into the capacitor. So - depending on the circuit - this results in current spikes of several hundred amperes. This is why all those chinese LED lamps don't last. And they even ruin the switches.

[xavier60]

In Reply #19 I mentioned that the capacitors still fail with a 2.2K resistor in series.