Re: MMC calculations - UK Caps

Hi Alan,

At 07:09 PM 7/13/99 -0400, you wrote:
>Terry wrote
>>I have updated my MMC capacitor power and reliability calculations.  They
>>now include the new and important life expectancy formulas.  The new page
>>is at:
>Thanks for this Terry. 
>I worked through it see how hot our 
>new mmc caps might get. 
>I must empasise that we are taking results from
>one manufacturer and apply them to another 
>manufacturers capacitor. 

Since all these caps are made pretty much the same, one manufacturer's
numbers are similar to another's as long as the construction is the same.
I have pulled a lot of caps apart and there is really not any great
difference beyond the label.

>Our caps are:
>0.047uF, 1500DC, 450V AC, 1800 V/uS
>axial caps are 33mm long 15mm dia,
>made by Arcotronics.
>(I had to guess at the thermal
>dissipation - I guessed 70 degrees
>C per watt.) 

I would say they are in the 40 - 50 range.  I have a great data sheet from
EVOX that is good for helping out this guessing at:


>I'm also assuming that the dissapation factor
>tan delta is the same for all polypropylene caps. I also assumed
>that Ksync would equal 1 for a static gap. It is also assumed
>that all the heating occurs in the dialectric, and that the resistive
>heating in the foil is not significant.

All the data sheets I have seen have pretty much the same numbers for tan
delta.  Ksync is 1 for any gap that always fires at peak voltage.  Only
async rotaries would be 0.5.  I never would have considered this but this
happens to be what Chip is running.  Resonant systems that fire a number of
times during a cycle would be somewhere in between.

>These caps are 0.047uF,
>Vcap is the voltage at which the gap fires
>divided by the number of caps in the string.
>ie peak voltage per cap.
>for the UK breaks per second is 100
>(50 Hz mains)
>at 200000Hz
>Vcap     Temp rise
>3000    13.2
>2750    11.1
>2500    9.2
>2250    7.5
>2000    5.9
>1750    4.5
>1500    3.3
>1250    2.3
>1000    1.5
>750     0.8
>500     0.4
>250     0.1
>at 100000kHz    

??? probably didn't mean kilohertz there :-)

>3000    16.5
>2750    13.9
>2500    11.5
>2250    9.3
>2000    7.3
>1750    5.6
>1500    4.1
>1250    2.9
>1000    1.8
>750     1.0
>500     0.5
>250     0.1
>remember Terry was using radial caps from a 
>different manufacturer.
>These figures should show the correct trends
>but could be out by a factor of 2 or more!
>We could get the catalogue from Arcotronics
>and see if they give relevant figures, or we
>could just try running them hard and taking their

The manufacture's numbers have a lot of guessing in them too and they "pad"
the numbers to be safe.  In general, these are all only guides.  The "real"
proof is in the running...  Of course, if the temps come out super high or
the life is like 5 seconds, you can bet there is going to be a problem.  I
saw one manufactures "real" test data on TAN delta, the numbers were only
20% of what the caps were rated at!  They are simply playing real safe in
case someone makes them prove those numbers.  I get the impression they
were taking into account someone spilling pop in the cap winder too :-))  A
number of places have done good testing but they don't quote those numbers
because they don't want anyone to really design something to those numbers.
 They know a lot can go wrong on the production floor and they want to be
sure there are no problems no matter what.

>Terry reckons a temperature rise of
>0-5 degrees as very good,
>5-10 degrees as good
>10-15 degrees as iffy
>15+ bad
>in terms of reliability.
>I went through the life expectancy formula as well and put it
>on a spread sheet.
>The caps that we are using have a AC rating of 450V AC, (X)
>The table shows the expected life time in hours,  (L) for the AC
>voltage across each cap, (Y) = transformer AC voltage divided
>by the number of caps in the string) 
>The formula is L= 1000000 * ( X / Y )^15
>(the life time in hours is the ratio between the rated AC voltage and
>the actual AC voltage raised to the power 15 and then multiplied by
>one million. (looks a bit arbitary to me)

I took all the data sheets I have read, and drew life data from the caps we
use at work, and churned them all together.  The life numbers will be less
than in many applications but Tesla service is very hard on the caps in
that they go through almost 100% high power voltage reversals.  It is "my"
best guess.  I tried a number of different situations and the numbers
seemed reasonable to "me".  Of course, if anyone has any better ideas feel
free...  Also, remember that most of the over voltage damage is being done
simply by the 50/60Hz AC signal.  I also guessed in some of the damage from
the high frequency voltage too.  It would probably be fairly easy to set up
some test strings running with just high voltage AC and come up with more
accurate constants...  If you found anything neat, you could write a big
paper and all the cap companies would come running...  I should mention
that the (Vo/V1)^15 function is very widely accepted and has been proven in
a number of tests.  It is just a matter of using the right numbers for the
other constants.

>actual  Life time
>Vac Cap in hours
>450.00  1000000.00
>500.00  205891.13
>550.00  49288.70
>600.00  13363.46
>650.00  4022.42
>700.00  1323.48
>750.00  470.18
>800.00  178.58
>850.00  71.93
>900.00  30.52
>950.00  13.56
>1000.00 6.28
>1050.00 3.02
>1100.00 1.50
>1150.00 0.77
>1200.00 0.41
>1250.00 0.22
>1300.00 0.12
>1350.00 0.07
>1400.00 0.04
>1450.00 0.02
>1500.00 0.01
>This is probably very conservative - I did
>subject a string of 4 of these caps to 8000 VAC in
>parellel with the existing cap.
>and they did survive a minutes run time!

That really does surprise me!  Of course, the ionization mechanism may not
follow the function at the extremes.  But the function does predict they
would die in microseconds...  If you knew how long the string did last (a
test) perhaps you could refine the equation.  If you could get a few data
points, you could really have a good idea of what the function is.  It is
interesting, we run some caps in applications at work that should die too.
I initially got worried about that but they seem to run fine after many
years with no failures...  There are probably even more and perhaps unknown
factors that add into all this.  However, it gets to a point where you just
have to take all the equations for what they are worth and just hope your
cap lives...  Temperature is even more complex because a fairly large
number of failure mechanisms can come into play.

>Do not be fooled by the numbers like 205891.13 the
>formula is a best guess to start with and further skewed
>when we apply it to our caps.

Of course this is all a guess.  However, if one does the calculations and
the numbers come out really bad, I would bet the cap will die real quick.
It gets hard if your numbers or sort of "on the edge"  that is when a
little error can make a big difference.  If the calculations come up with
really good numbers, the cap will probably last forever.  My latest life
estimates for temperature and Dv/Dt are far more arm waving than this one
but the numbers still seem to come out reasonable to me...

I am glad you find some use in all this.  It really is interesting to read
through all the data and try and get inside the heads of the people that
wrote it.  I think they know a lot more than they say but they really want
to be very conservative in any numbers they publish.  They don't want to
get blamed for messing somebody's project up so they figure it is better to
far underrate than be wrong.  One measure of how tough these caps are is
how few MMCs we coilers have blown so far.  I do not know of anyone who has
blown any that were reasonably designed.  That is far far better than our
other cap designs have faired in their first attempts at using them in
Tesla coils.  I was a little concerned that all the "gurus" didn't take the
ionization problem real seriously.  They said if you don't cook them to
death and don't overvoltage and arc them to death, they will last forever.
As time goes on, I think they may be right.  My EMMC cap may show a little
drift and the internal parts may be slightly marred, but standing back and
forgetting all the fancy numbers, it looks like it will last far longer
than two hours to me.  Only time and experience will tell us how they will
really perform.  However, there seems no doubt now that they far far out
perform anything else.

BTW - Here's a sobering thought.  I measured the thickness of the poly
dielectric in one of my caps and figure how much total poly thickness my
EMMC cap had to stand off the 21000 volts of my coil.  It's 4 mils...  Of
course, if it does arc through, that is simply considered as "normal" and
does no harm...  However, as far as I can tell, it does not arc though...  

I'll make another guess, judging from the six month history of MMCs, six
months from now, we will consider nothing else for Tesla caps...  Time to
sell the mineral oil stock :-))



>have fun,
>Alan Sharp