Self healing:

Self healing is a big deal for some kinds of capacitors.  Self healing is an old concept, going back to the earliest days of metallized paper capacitors, around 1910.  The picture below (a simplification ) shows how it works with metallized film capacitors.  The energy from the arc blows the metallization away.  This does not work with film-foil capacitors.  Polypropylene self heals well because it leaves little carbon at the site of the arc.  Epoxy-paper also heals well for the same reason, plus the hardness of the material helps control the size of the expanding arc site which helps raise the gas pressure at the arc site.  This helps extinguish the arc.  All this makes these materials suitable for X and Y film capacitors, as well as any other film capacitors that operate under high voltage.  Self healing occurs in a matter of microseconds and takes only microwatts.  This is not a perfect system, faults can fail to clear under some circumstances. 


Self healing occures in other types of capacitors.  This has been mentioned elsewhere. 
Reliability 2
This page covers a variety of issues that more or less relate to reliability. 

Reliability of electronics: 

Whole libraries have been written on the reliability of electronic systems.  Software can be used to calculate the reliability of systems based on the reliability of each part, one by one.  If you are a hobbyist rather than a reliability engineer, you can make your project reliable by following a few simple rules of thumb. 

Prepare for the obvious failure mechanisms.  Equipment running off the AC line should able to handle strong line surges.  Equipment in general should be protected against reverse-polarity power supple voltage spikes on power up.  Design against static electricity and EMI/RFI interference. 

The 50% rule.  If a part has to withstand 50 volts, use a 100 volt part.  If a part will be carrying 1 A, use a 2 A part at least.  Heat is always the enemy.  The rule-of-thumb is that reliability drops 50% for every 10-15C increase in operating temperature.

Voltage and current: 

Tantalums in general are vulnerable to current spikes.  The nature of the dielectric results in weak spots.that makes them prone to failures.  Both MnO2 and polymer electrolyte types are self-healing, but this is not perfect.  This makes tantalums suitable for output capacitors for three-terminal regulators, the current is limited, but not for input capacitors where the current isn't.  Some tantalums are pretested for surges.  Silver-cased tantalums are very intolerant of reverse voltage.  The silver grows dendrites that will almost instantly destroy the capacitor.

Counterfeit Capacitors

When you are talking fakes, you are talking EVERYTHING.  Just like any other electronic component, capacitors are the target of counterfeiters.  This includes low quality parts mislabeled as to manufacturer, parts labeled as one thing but really a cheaper part inside and parts that are totally nonfunctional, such as tantalum capacitor cases with no actual capacitor inside.  Such parts are typically traceable to state-owned Chinese companies with go-betweens in Taiwan.  The best defense against counterfeit parts is to buy from the factory, or from authorized distributors.  This not a perfect defense.  Some people are sneaking counterfeit parts between the factory and the distributor, but about 95% of counterfeit parts are from unautherized distributors.  Hobbyist tend to be bargain hunters however and buy capacitors and other parts from Ebay at their own risk. 

How can you tell? 

Use a decent LCR meter to measure the value of a capacitor and its dissipation factor or ESR.  The dissipation factor at 100 kHz could be used to tell you the material of a film capacitor for example.  There are lots of meters out there at reasonable prices.  It might also be possible to measure the breakdown voltage of an aluminum capacitor by applying increasing voltage through a 1M resister and measuring the leakage.  I might try this. 
          Is the name of the manufacturer spelled wrong?  That really happens. 
          Are the markings of less than perfect quality?  Does it just look a little shabby in general?

The picture on the right is the most awesome I have seen on this subject but I have no idea who owns it.  If the owner would email me I'll add an attribution.  

More bad capacitors: 

These are pictures of old metallized-paper line-filter capacitors from Rifa (now part of KEMET).  The paper is saturated with epoxy.  When I saw the first damaged capacitor I was quite surprised.  Rifa is an old European electronics company, the kind of people who take electronics safety very seriously, unlike us.  Also, line-filter capacitors are made to be very reliable.  I would not expect to see even one bad one.  Epoxy-paper parts are generally considered to have better self-healing than polypropylene. 

A search of the internet found more pictures.  They were taken by a variety of people, some just own a piece of old audio equipment, some are hobbyists who restore antique electronics such as audio gear or old computers.  Some are of equipment that might be as much as 40 years old, and none seem to be very recent.   The one series number that showed up was  PME271M (X2, 275VAC).  These are of a epoxy/metallized-paper construction.  They are a current series and very common, but they don't have the quite same appearance as current manufacture.  In some cases it was impossible to tell.  I found no blown filter capacitors from other companies like the countless Chinese manufacturers (that surprised me).  I also found none of Rifa's polypropylene film line-filter capacitors.  It's a good thing I did my search now before Rifa thinks to invoke its right to be forgotten. 

Why is this happening?  Who knows?  One theory is that the epoxy is too hard.  There are reports that temperature cycling will cause failure (several hundred cycles).  It's also reported that the parts develop tiny cracks before failure, which would tend to confirm this theory.  In any case, although the capacitors got hot, cracked, turned brown, smoked and smelled bad, they didn't actually catch fire.  That is a strength of epoxy-paper capacitors. 

More recently I found a picture of a failed epoxy-paper capacitor of another brand.  Could it be that this is not a good technology for long-term survival of line-filter capacitors? 



Whiskers are very fine filaments found growing from metal surfaces, most often plated.  Whiskers were found in electronic
equipment at least as early as 1946.  They are capable of causing short circuits and other problems. Whiskers don't just cause
failures in their own equipment, if disturbed they can float through the air and damage nearby equipment.  Whiskers are mostly
associated with pure tin plating, but have been seen in silver, zinc, cadmium and even gold.  Whiskering seems to be caused by
mechanical stresses caused by the plating process, but is also found in hot dipped zinc.  This is interesting because computer
rooms tend to have a lot of zinc- and cadmium-plated hardware.  The whiskers can dislodge and float into the computer
equipment.  Whiskers are very fine, typically from <0.5 um to 10 um in diameter.  Lengths can run from <1 mm to 10 mm (over
a third an inch).  Shapes vary; some are bent, some are straight and some are other shapes.

Why does it happen?

Whiskering is not well understood.  There is no General Theory of Whiskertivity.  For example, it's not possible to reliably
predict when whiskering might start.  It could take days to decades. 

There are, however, some clear risk factors.  Higher humidity seems worse than low humidity.  Atmospheric contamination can
be important, silver whiskers are associated with sulfur compounds and the formation of silver sulfide.  Contamination of plating
baths may be important in the case of gold.  Contamination may play a role in many whiskers.  Heat cycling with various levels
of humidity encourages whiskering and this is commonly used to test metals and plating.  The fundamental cause is believed to
be "compressive stress" in the plating.   

Where are whiskers found?

Everywhere...component leads and the end-terminations on SMD ceramic capacitors, lead frames, wirewrap pins, connectors
and sockets, relays and circuit breakers. That's just the beginning.  Whiskers have been implicated in the failure of everything
from satellites to pacemakers.  

What are the failure modes? 

The common failure is a short circuit, either temporary or permanent.  If there is enough current available to vaporize part of the
whisker, then the short will clear itself.  If not, the short becomes permanent.  In a vacuum, metal plasma can allow for arcs in
the range of hundreds of amperes if available.  This can cause a catastrophic failure that can destroy a relay or circuit breaker. 
At GHz frequencies, whiskers can act as antennas or affect waveguide performance.  Whisker problems are potentially as bad
now as in past years, even though manufacturers are working to mitigate the problem.  Voltages and currents in modern
equipment are going lower and lower making it less likely a short will be cleared.  Spacing between parts is also dropping. 

How can you see whiskers? 

You don't need an electron microscope (but it helps).  A stereo zoom microscope, 5X-100X, with adjustable illumination
(NOT a ring light) will usually work.  You need to be able to move the light at any angle.  Whiskers are very fine and difficult to
see and you need to get the light at the correct angle to the whisker. 

How do you stop it? 

In the 1940s this was easy, add a few percent of lead to the tin plating, as little as 3% (some people say <1% will do).  Other
additives have not worked so well.  Lead frames, for example, were commonly plated with up to 10% lead.  For tin and zinc
this worked for many decades.  Most people eventually understood not to use pure tin plating, even though the reason was not
always remembered.  Thanks to RoHS this is usually no longer possible.  The adoption of RoHS upset things and the whisker
problem has to be relearned. 

There are a number of approaches to prevent whiskers

If the cause is "compressive stress", then anything that can prevent or relieve this stress should prevent whiskering.  Opinions still
vary...a lot.  Matte finishes help or don't help, board coatings will or won't stop the problem, intermetallics (the plating metal
mixing with the underlying metals, usually copper) are or are not a factor.  A nickel layer under the tin (to separate it from the
copper) will or won't stop whiskering.  Silver has also been considered.  Adding bismuth to tin might or might not be an answer. 
Some companies already specify a small amount of added bismuth.  Some people have proposed a reduced amount of lead in a
lead/tin plating.  This probably won’t happen. 

Stop using pure tin, zinc and cadmium.  Satellite builders have moved from pure tin plating to nickel where possible.  NASA
and the military still specify tin/lead plating and solder. 

On circuit boards, use a reasonably hard board coating, such as a urethane or acrylic, of at least 2-3 mils.  Soft coatings
(silicone) are known not to work well.  Parylene is sturdy but goes on very thin and tests are contradictory.  Some Parylenes
may work better than others.  An advantage of Parylene is that it is very uniform, something not true of other coatings.   NASA,
and probably others, has had problems with contractors delivering pure tin plating when tin/lead was specified.  Perhaps all
assemblies should be checked with X-ray fluorescence equipment.  This is sometimes done by the Europeans to test for lead

Many people have changed from bright tin to a matte tin finish over nickel.  This approach may work but it has not convinced
everyone.  They also relieve mechanical stresses with heat soaking for several hours at about 150C.  This is now commonly
done with lead frames and SMD components.  Probably both are required.  There is evidence that a soldering temperature that
reaches the melting point of tin (232C) is very beneficial for some parts.  There is also evidence that a thicker plating is better
than a thin one, to 10 um. 

So, best practices seem to boil down to: 
Matte tin over nickel.  Matte tin has larger grains with lower internal stress. 
Heat soak for several hours at 150C after platting.
Plating thickness of about 10 um.
Addition of about 2-3% bismuth to the tin.  Much more than 5%, would affect soldering. 

More information can be found below.  The Panashchenko paper is an especially good introduction to whiskers with many
good pictures: 



Capacitors can be damaged by radiation, but this probably not a problem unless you are sending a space probe to the outer planets.  In that case you probably want glass capacitors.  They have the ultimate resistance to radiation.  In general, glass, ceramic,  mica, and other "mineral" materials are very resistant to radiation while plastics and other organic materials  have much lesser resistance.  See NASA SP-3025 for more information.

On the right is a table of the approximate relative radiation resistance of various dielectrics in order, best to worse.  The relative order of the polymers is approximate at best. 


Humidity Resistance: 

High humidity can have a significant affect on film capacitors.  Increasing the ambient humidity from 50% to 90%, for example, can cause a capacitance change from as little as 0.5% to as much as 4% depending on the dielectric. The change is temporary unless actual corrosion results.  The time it takes to make the change (hours to weeks) depends on how well the capacitor is packaged.  Boxed capacitors take much longer to reach equilibrium than dipped.   Tantalum and ceramic capacitors have also been reported to have failure problems with high humidity, but I have not seen information on drift.  

The chart on the right shows the rough correlation between moisture absorption and change in capacitance.  The actual numbers were collected from a variety of sources and probably were not derived with the same methods. They should not be taken entirely seriously. 

How about ceramics?  The reason humidity affects plastic dielectrics so much is because water is a very polar molecule with a dielectric constant of about 78, while plastics are typically around 2-8.  C0G ceramics are typically 50 on up.  Add 0.1% water to you get a change of maybe 0.1% at most.  That assumes that ceramic dielectrics can even absorb that much water (no one seems to know).  So yes, ceramic capacitor could probably drift with humidity, but not very much.  Still, I haven't seen any numbers.  NASA has done extensive whisker research.  Nice drawing.  Search on “tin whiskers”.  NEW  Has a drawing of six brands of popular aluminums of their top vents and rubber seals.  Very interesting but not sure how reliable this would be to identify counterfeits. "The World Leader in Supplier and Counterfeit Part Risk Mitigation Solutions."  Bad Sansui capacitors.