Re: MMC practical apps.
Subject: Re: MMC practical apps.
From: Terry Fritz <twftesla-at-uswest-dot-net>
Date: Wed, 07 Jul 1999 22:57:04 -0600
With MMC caps we have a wonderful opportunity to study and test all kinds
of problems with high power capacitors. They are very consistent and lend
themselves well to destructive testing. They also have some manufacturer's
data that is supposed to tell us wonderful things about the parts. I think
I have now seen every failure mode the manufacture's have mentioned "up
close and personal". Tesla coils can easily drive them to every destruct
mode. Of course, or problem is how NOT to destroy them :-)
One could string together a few thousand poly caps and make a truly
indestructible cap but the cost would be very high. So, we push them to
the limit and try and find just how hard we can safely push them. These
limits are beyond the manufacturer's conservative specs but the caps stand
up very well and we benefit from dramatic cost savings in the reduced
number of parts.
At just this moment, internal ionization seems to be a problem. However,
there are a number of other possible problems that do not seem to be a
concern. The lead connections are not failing, the metal is not vaporizing
from high dv/dt, the end connections are sound... Basically they work very
So how does all this reflect back to the old "reliable" rolled cap. If a
big rolled cap is getting hot, and maybe even just warm, this is a definite
warning sign! When a poly cap gets hot on the outside you can bet that it
is getting way too hot on the inside. Heat is definitely the first killer
of caps. Second is pure over voltage puncture. MMCs really do have the
remarkable ability of failing due to over voltage and recovering almost as
good as new. At work today, a power supply failed a cap and we took out a
near by "good" one and looked at the layers. It looked like it was machine
gunned with arc throughs, but the cap still survived! If we could somehow
extend this self healing feature to the rolled cap it would be a major
advancement!! And now we have internal ionization. This is one of those
little things that was barely mentioned in the data sheets that seems to be
very important to us. We like to run our MMCs at the highest voltage
possible to reduce the number of caps and save a ton of money. However,
when about 700 volts is placed across poly dielectrics, the metal edges of
the plates start to corona. At low levels it is just a glow bombarding the
>1 mil layers with electrons. It will fail eventually but hopefully not
too soon for our purposes. At higher levels (EMMC) the corona seems to be
able to melt and cause voids in the dielectric and the first real signs of
serious trouble are showing up. If I used twice as many caps, this problem
may not be a factor. I just don't know that for sure.
There is no doubt that all this MMC testing will result in vast
improvements in all TC caps. We can now almost predict how hot an MMC will
get in a given TC application. That is a stunning advancement in itself!
But we are also learning of RMS currents, peak voltages, ionization, and
other things that we can now really quantify and design for.
So to get to your real questions without rambling :-))
Rolled caps have very thick dielectrics but they also have very high
voltages across those dielectrics. Voids may be able to become real hot
spots causing failure. Also if the foil is not very close to the poly, the
oil my overheat due to its higher losses. You really only want the oil to
fill the voids as a last resort. The foil must be in very close contact
with the poly to insure low loss. The fact that your cap was getting so
warm may be a sign of this type of problem.
Re-pumping the cap really should not remove any more voids. It is really
important to pump the cap down when completely dry and then introduce the
oil. This will insure that the oil will penetrate to every small void. If
a void does form latter, the surrounding oil will hold it in place and it
will be very hard to remove...
If the spider crake is clean and shows little sign of burning, it is
probably do to pure explosive effects. If it is burned, it may be a sign
of serious overheating of the poly at that point. If it has a lot of
carbon buildup that is probably more from the oil burning than the poly.
Hope I said something of use here...
At 05:40 PM 7/7/99 -0400, you wrote:
>Ive been reading on the recent MMC trend and it seems to me that its a
>great compact alternative. The only problem I see for the MMC is its
>cost to coilers with small budgets like myself. I was reading Terry's
>note on the falling capacitance and dielectric air pockets. Since most
>coils incorporate the rolled poly or poly plate caps I was curious if
>this dielectric bubbling could cause a similar problem for them as well?
>I noticed a C reduction in my rolled cap before it died on my 4" coil
>that I posted specs on a little bit ago. The reduction was quite small,
>so small in fact that I never thought about it till now. I realize the
>oil in the cap should fill these pockets of air as they form (if they
>do) but could periodic re-vaccum pumping could help this? Since im sure
>that not all air can be removed in the first pumping of the cap. My cap
>only got hot and died spilling some smoke filled oil into the catch
>bucket. My dissection of it revealed a spider effect on the poly surface
>that looked like cracking of a car windshield when hit by a rock.