Cap life testing - Re: MMC cap bank - ion inception testing

["Tesla list" <tesla-at-pupman-dot-com>]

Original poster: "Terry Fritz" <teslalist-at-qwest-dot-net>

Hi Ken,

At 08:35 PM 7/6/2003 +0000, you wrote:
>>>...........
>>I was thinking of something like:
>>http://hot-streamer-dot-com/temp/CapLifeTest.gif
>>I think that main fuse should be 7.5 amps not 0.75...
>>lots of fuses...
>
>This setup does not tell you when a cap has reached end of life by losing 
>too much capacitance. While losing say 11% capacitance may not sound too 
>bad, you've lost conductive metallization in the cap, and it's ESR has 
>gone up. Also, we know that voltage is highest across the smalles value 
>caps when used in series. You end up with a downward spiral effect when 
>one cap starts to wear out.

For the ionization problem, the capacitance does not decrease. We are not 
punching holes in the dielectric and blowing away metal which causes 
capacitance decrease in caps that are over voltaged a lot.  Here, we are 
just chewing on the edges of the dielectric.  They start off with say 5500 
volts of stand off.  When the damage gets so bad that they fail at 2000 
volts, "I" would say the cap is pretty "well done".  I was thinking that 
the first failure would cause a current pulse in the fuse that would blow 
it indicating the failure.  A neon bulb with like a 10Meg resistor across 
the fuse could make a nice easy lamp indication.  Also, the fuse should 
"save" the cap so it can be taken apart to check out the damage first hand.

Normally, we don't overvoltage caps to the 5500 volts needed to start 
punching holes in the dielectric.  Our MMCs will probably never see a 
voltage punch though.  But if we do have a problem like a short in an array 
or a streamer hit.  Our MMCs will laugh it off ;-))  Josh's filter will all 
the popped caps was pretty cool!  Far cooler was the fact it still worked!!

The dissipation factor may go up some over time, but most of the current is 
in the solid plate area where the ionization is far less.  We might not see 
much dissipation increase from dielectric damage just at the foil edges.

A zero voltage crossing solid state relay may fix the line transient 
problem right away.  It would prevent full voltage from going across 
discharged caps during turn-on or a momentary loss of AC which might blow 
the fragile fuse prematurely.  There is the HV transformer there too, but I 
think it would still work.


I was thinking about how long of life we want...  Assuming a museum coil is 
run for 1 minute out of every 10 minutes, 12 hours a day, 365 days a 
year...  That is 438 hours/year for a very heavy use.  For the home hobby 
coiler, we are probably looking at maybe 5 hours/year.

We probably would like our hobby coil to last 50 years at which time "we" 
won't care anymore ;-))  So that is 250 hours total run time.

For the museum, assuming the coil is real popular... The capacitors 
probably have to last till a big overhaul at which time the gap motor, 
relays, switches, etc. need to be replaced.  I am thinking every 5 years 
our museum coil will need a pretty big OEM overhaul which could include a 
new primary cap bank...  Maybe DC or another pro coil maker could tell us 
if that is reasonable to assume.  So we could say 2500 hours.

So we have a "wish" for 250 to 2500 hours of life.  Not too bad and pretty 
easy to test.  Basically 10 to 100 days.  Of course, we could make an MMC 
to run at far lower cap voltage too vastly extending life.  But then it 
starts to get so expensive you might as well buy commercial caps unless you 
need an odd value, ultra reliability, or like to hit the caps with 
streamers ;-)).  If you made a cap with only 500 peak volts per cap, the 
life would be...  just about forever...  It would be interesting to compare 
the price of a commercial cap and an ultra over-designed MMC which could 
easily outlive it.*

BTW - Don't let Mark and I steal all the fun here.  If you Ken (or anyone 
else) wants to do this experiment too, feel free!!  When it really gets fun 
is when we all get different numbers :o))

If CD does not already know these numbers, just watch them use "our" 
numbers themselves too ;o)))  Maybe we can get some free caps out of it 
;-))  Then watch them send us other caps to be tested too!!  Vendors love 
this kind of free testing ;-))  (as long as the results are good...)

Cheers,

         Terry

*A Maxwell 37667 30uF 35kV could be done with 70 caps per string in 14 rows 
to give only 500 volts peak per cap.  A $3000 cap!!  But at 200Arms and 
6000 amps peak rated and a "self healing" voltage of say 385kV...  You just 
are not going to wear that cap down ;-))  It would last forever especially 
considering a number of caps coil fail without really affecting the whole 
array.  You could hit the array with a shot gun blast and the remaining 
caps would easily still run it all...  Hit that Maxwell with 35.001kV 
Pooff!!  It's done....  Of course, the BIG MMC array would be far far 
larger than the 37667...

But, the MMC could also be a 15 x 3 array...... ;-))  $135, breakdown at 
82.5kV (self healing)...

BTW - Maxwell, now general atomics, has a pretty cool web site at:

http://www.hvpower-dot-com/capacitors.htm