[Prev][Next][Index][Thread]
MMC Cap Papers and MMC Life Prediction

To: teslaatpupmandotcom

Subject: MMC Cap Papers and MMC Life Prediction

From: Terry Fritz <twfteslaatuswestdotnet>

Date: Sat, 10 Jul 1999 18:28:16 0600

Approved: twfteslaatuswestdotnet

DeliveredTo: fixupteslaatpupmandotcomatfixme
Hi All,
I have two papers about poly caps on my web site. The first one "LT POWER
CAPACITORS" by S.N. Bhouraskar, tells of many details of poly caps. It is
good reading past the first part.
www.peakpeakdotcom/~terryf/tesla/misc/LT_POW_CAP.html
The second is a report from AeroVox "The Development and Performance of
High Reliability High Energy Density Pulsed Discharged Capacitors". This
paper tell a lot about self healing and voltage stress on caps.
www.peakpeakdotcom/~terryf/tesla/misc/polyrel1.jpg
through... (six pages)
www.peakpeakdotcom/~terryf/tesla/misc/polyrel6.jpg
Of special interest is the "Voltage Stress and Life" section on the fourth
page. This rule states that the life reduction of a capacitor is
proportional to the over voltage ratio raised to the 15th power. So in my
EMMC cap, (15000/6300)^15 = 450000. (15000 VAC transformer and 630 VAC
caps). So my EMMC should last about 1/450000 of its "normal" life.
Assuming the caps would last a million hours at their rating, my EMMC
should last about 2 hours. That is probably fairly close to being correct.
What is really neat, is suppose I wanted it to last 100 times that long.
How many more caps (and much more money) would I have to spend? The answer
is x^15=100 or x=1.36 So if I go to 14 caps per string for a total of 98
caps, my EMMC should last 200 (actually 347) hours instead of just 2. That
would double the number of caps so I would pay $140 instead of $70 but it
would last 100 times as long! 2 hours is a long time for me but 200 hours
is an eternity! The neat thing is if I went by the ratings of the caps and
used 288 caps and spent $400 dollars, I would never know the difference
between the 200 hours and 1 million hours of life but I would notice being
out $270!
Of course, all these numbers are just estimates but the principles are
there. This may be a good way to estimate how many caps we need to get
over the ionization problem without busting the bank doing it. There is no
doubt that we don't need 1,000,000 hours of life out of our Tesla caps and
we may be able to save a significant amount of money by giving up 990,000
of those hours we'll never need anyway.
Finn was concerned about his cap that has 20 1600 volt caps in a string.
Since he can run around 20000 VAC (and sometimes more ;)) and assuming his
caps are 630 VAC types we get (20000/12600)^15 = 1023 So his caps may last
around 1000 (1 million /1024) hours. He has noticed some capacitance
reduction in his caps and I think his museum setup gets a lot of run time.
So perhaps the rule is working here too! The 9 caps of mine he put in that
system would have only lasted 22 seconds by this rule which is just about
what they did! The 10 cap string would have died soon after at 107
seconds...
Since this rule applies to all caps, turning up the voltage into a cap
that is not designed for it can dramatically shorten it's life! One neat
thing about MMCs is that they can be easily configured for high voltage
primary systems. Since the length of the output arcs is proportional to
the input voltage, there is always a desire to use as much primary voltage
as possible.
My EEMMC cap died at ~12000 volts. By the rule, it would last 0.25
seconds at that voltage (4 caps at 630 VAC). Close enough...
BTW  In all my looking around, AEROVOX seems to be the most informed cap
manufacturer of them all. They have much more detailed data and appear to
have done much more testing than all the others combined. Of special
interest is that they specifically list power dissipation in their data
sheets which is very important to us. The other manufacturers seem to be
rather unsure of such things... I can recommend their papers that they
send you very fast at:
www.aerovoxdotcom/forms/form_lit.htm
main site at:
www.aerovoxdotcom/prodindx.htm
Cheers,
Terry