How are film capacitor voltage ratings determined?

[EC8010]

The best I could find on a general search was that dielectric withstand voltage was taken to be 3/4 of the voltage that caused arcing/failure, and that working voltage was taken to be 1/3 of dielectric withstand voltage. That sounds rather fluffy and "wet finger in the air" to me, but is this genuinely how it is done? I ask because years ago I suffered failures of motor run capacitors that upon investigation proved to be physically small for their CV product, indicating a high field strength, so I specified a higher working voltage, resulting in a larger capacitor that lasted longer. Ideally, I would like to see something that relates probability of failure to applied voltage. Just how does a capacitor manufacturer decide what voltage rating to print on their film capacitors? Do they just trust what the film manufacturer says about field strength?

[dietert1]

Recently somebody asked me about getting WIMA film caps with ratings of several KV from Germany. First i said: That doesn't exist. But then we found them at Digikey. They appear like the usual film caps yet with very strange voltage specs that indeed include KV - e.g. FKP1 up to 6000 V. When you look into the details, these are meant as pulse caps. The magic word is "self-healing capability". They will loose a little capacitance on each "event".

Regards, Dieter

[Gyro]

Film capacitor covers quite a broad range of parts. In addition to different dielectrics with different characteristics, more important in this context is the type of electrodes. You can have foil electrodes (eg, FKP polypropylene) which are used in RF circuits (low inductance) and metalized electrodes, sprayed onto the dielectric film. There are also two types of metalized film: the thicker type, which is used for high pulse current applications (Epcos do a bunch), and the almost transparent type used (as dietert1 says) for self healing types. On overvoltage or dielectric overstress, FKP and high pulse metalized film caps will fail catastrophically (short circuit). Self healing types will burn back some of the thin electrode film around the puncture in the dielectric film, sometimes even isolating complete sections. This leaves the capacitor still operational but with reduced capacitance, this happens on every overstress event. Whether the capacitor remains 'functional', and for how long depends on how the surrounding circuit copes with the reduced capacitance. If you check the datasheets for the various types, you should see a variation in operating voltage to dielectric withstand voltage. This applies also to capacitor dissipation - the AC current it can carry and the rise time (dV/dt), these are a function of frequency and waveform.

Self healing types are used for class X filter capacitors which must fail safe (not short) in across the mains applications. Motor run capacitors are also of the self healing type. This is preferable to the capacitor shorting and taking out one of the motor windings. They are subject to spikes and surges the same as X caps are. At a certain reduction in capacitance the motor will loose power, eventually to the point of not starting.

Motor run capacitors are actually rated according to MTBF. Taking a quick example from the web, I see a capacitor that is rated for 10k hours at 425V~, the same capacitor is also rated for 475V~ but the MTBF drops to 3k hours. Yes, the capacitor has a dielectric withstand voltage, and is probably quoted for an operating voltage on a distributor's website (probably 450V in this case) but this doesn't appear on the capacitor case, just capacitance, tolerance, and the two voltages, with MTBF figures. Clearly, by increasing the 'voltage' of the cap, you have increased the MTBF, probably greatly - in the above example changing the operating voltage by 50V around 450V leads to a 3:1 ratio in MTBF.

All self healing caps will degrade in value over time, both from mains spikes and inductive back EMF spikes from the motor during switching. The figure to focus on though, rather than dielectric withstand, is what is your acceptable MTBF for the motor operating conditions. You probably want to de-rate further if the motor is frequently cycled versus one that is constantly running.

[EC8010]

Yes, I have observed the loss of capacitance due to self-healing. And I was able to deliberately cause it by applying a large DC to the capacitor from a voltage multiplier from a Variac. In that way, I found that the fitted capacitor would not tolerate much over-voltage before measurably losing capacitance, whereas the new capacitor tolerated 2kV without any loss of capacitance. Once I'd discovered that, I asked maintenance to measure capacitance on all the motors. The results were appalling - hardly any were anywhere near their stated value. Wholesale capacitor replacement stopped the annual summer failures. Incidentally, the motors ran 24/7 and were too hot to touch (even when the capacitor value was correct), and the capacitor was bolted to the motor, heating it, and reducing its life.

I had not previously spotted MTBF specifications on motor run capacitors.

But what I was really hoping for was some thoughts on how a capacitor manufacturer decides on a voltage rating when introducing a part. General purpose capacitors, not specifically motor run (I just used that as an example I had been forced to investigate in detail). I recently bought some metallised polypropylene capacitors that are remarkably small for their CV product. I'm making some test equipment to compare these with other polypropylene capacitors and I'm wondering whether they're just bending the specifications and accepting reduced reliability or whether the rating is genuine. I've already used an electrometer and variable DC supply to measure leakage current of film capacitors/insulators and found that there's a current that increases linearly with applied voltage (therefore pure resistance) plus an avalanche current that rises rapidly with voltage - almost certainly responsible for the sharp change in MTBF noted by Gyro. The trouble with that technique is the risk of damage. New kit being made is an updated version of the Airmec 251 ionisation tester which used an ampliifer plus loudspeaker after the electrometer. Apparently, a hiss could be heard before breakdown. This technique seems very useful.