Re: Another MMC cap candidate (UK)

Hi Alan, all,

A little quicker (and longer) response this time ;o)).

>Original Poster: Alan Sharp <AlanSharp-at-compuserve-dot-com>
>I've been following the correspondence on multi-mini- caps.
>And I was very impressed by the performance of Alex Crow's
>coil at the UK teslathon.

Very nice pictures on Mikeīs site. I hope we will get around
to organizing a Teslathon, here in Germany, soon. Big
compliment to all those shown. I esp. liked Vivianīs plexiglas
SRSG. Great work Vivian!! You seem to be able to find
LOTS of plexiglas for cheap ;o)).

I noticed you guys were running the MMC in parallel with
some saltwater caps (if I made it out correctly) on the 9.5"
coil. Did you try running the coil soley on MMCs (i.e: w/o
the SW caps attached)? Even though the capacitance would
have been lower, the performance should increase, because
SW caps are very lossy. This will noticeably reduce the
possible spark length.

>What I have summarized is that the caps will not be killed by
>If the 1.414 * AC is these than the DC voltage rating of the cap.
>So for my 8 KV AC transformer I need 8 of 1500V DC caps in series,
>I'll probably use 8 by 8 in series / parallel.
>They will also not be killed by over current is dV/dT is sufficent.

Here is the way I build my MMCīs:

I.) General comments:

a.) Go for poIypropylene only.
b.) Get physically big caps. The bigger surface:
b1.) will allow a better heat disappation (prevents premature cap
b2.) usually is a sign for metal FOIL endplates (very important!).
b3.) From b2, these caps have MUCH higher dv/dt ratings.
b4.) Using metallized plastic ENDPLATES WILL result in cap failure
       (I tried and lost some Philips caps.).
b5.) If you find any cap having a dv/dt less than 1000V/ĩsec, you can
       bet on them not having metal FOIL endplates. DONīT use these.
c.) Leave a space between each cap row. This will allow any created
     heat to be carried away by air convection.
d.) KIS (keep it simple). While I did propagate using interlinking of
     the cap strings a while ago, I no longer believe this is worth the
     trouble, because:
d1.) it complicates the whole thing. Removing an interlinked string (for
       tuning or repairs) envolves lots more work.
d2.) If an interlinked cap in a string blows(opens), you get unequal
       voltage distribution, that might lead to more problems.
d3.) Design your setup, so that it is easy to pull a complete string. I
       mount mine upside down and use a friction fit to hold them in
       place. Pulling a string is a matter of seconds (only two
       soldering junctions).
d4.) Following d1-d3 makes the MMC very universal. This means you
       can use it for more than one coil.
d5.) Cap killers are high BPS. This being said, stay away from BPS
       rates >>480+. Go for a bigger cap instead. Following John
       Freauīs posts have shown me that high BPS with a small cap
       will NOT lead to longer sparks than a coil with low BPS and a
       "correctly" sized cap.
e.) I find equalizing resistors unecessary for equalizing the voltages
     across the caps. They do, however, safely discharge the whole
     MMC within a few seconds. The actual resistance is uncritical.
     This will vary with the number of caps used in a string, as you
     donīt wanīt to get too low on total resistance (wastes power
     and heats up the capīs surroundings).

Specific comments for DW (spark gap driven coils) only:
To make things easy (for various explantions), letīs use
this example cap:

WIMA 100nF
dv/dt: 3500V/ĩsec

Array: 10x10 for 10nF per row and 100nF per MMC
Vdc(rated)= 16.5kV
Vac (rated)= 8500V
Imax per string: 350A
Imax per MMC: 3500A

1.) dv/dt rating:
 a.) In order to stay universal, try not to exceed dv/dt too
       much. While (as Terry has shown) you CAN exceed
       dv/dt, you loose versatility doing so. What I mean is,
       if you build your cap on the limit (there are some $$
       reasons making this legimate), for e.g. 120 BPS, then
       you wonīt be able to run it at 240 bps, etc.

b.) Calculate the maximum primary current.
c.) Calculate the allowed current per cap (from C* dv/dt)
d.) As the current flows through each cap (in a string), the
     maximum allowed current is the same for one or 1000
     caps in a string.
e.) From b, now calculate the number of strings needed.
f.)  Find the necessary (wanted) total capacitance
g.) From b-f back-calculate the needed single capacitance
h.) This sounds a lot more complicated than it is. You will need
     to fiddle with the numbers, but there are various possible
i.) In DW coils, the peak current is not being fed continiously.
    This is one of the reasons why we can run them "on the edge"
j.) Richard Hull once said that MMCīs are only good for coils in the
    <2kVA range. I disagree on this. The larger the input power, the
    larger the coil will be (following simple reasoning). The larger the
    coil, the lower the FRes is. As T = 1/F, the lower Fres is, the
    easier it is on the caps, because the cap has "more time", so
    the dv/dt rating becomes less critical. As we are using more than
    one string, we will divide the current up between the strings. In my
    present case, I am running ~600A primary current, but I am using
    13 strings, so each string only sees 46A. I have run these caps with
    up to 65A per string and have had no failures.
k.) Looking at T (from j), the steepest rise occurs during the 1st
     quarter, so this period is of true interest.

(My) conclusions:
dv/dt should be kept within limits, in order to stay universal.
You donīt need to be super religious about it, tho. A 1.1x
overrating surely wonīt lead to premature cap death. Looking
at our example cap, we see it will be very hard to actually
exceed the rated dv/dt of this cap. (Okay, Greg Leyh might
be able to with his ALF coil).

2.) Vdc rating:
a.) Try to stay within the voltage rating. Due to the high quality,
     (Terry, once again) you can exceed this limit, too,
b.) but doing so, will once again be a trade-off in universatility.
c.) Most manufacturers test their caps at 2-3 x Vdc. They must
     survive this for 2sec (IEEE standard).
d.) Following a and d will give you a good measure of safety
     against strikes, kickbacks, etc.

(My) conclusion:
Keeping Vdc within limits gives you great amounts of safety,
in case anything goes wrong. However, there is absolutely no
need to go for a rating equal to the one we are using for rolled
poly caps. The rolled poly caps have the BIG disadvantage of
being self rolled and using non uniform, non virgin material. The
MMC does not have these disadvantages. Letīs take another
look at our example cap. It is rated for 16.5kVdc. Terry has
shown, these will fail at 5kV PER CAP, so our example cap
will fail at 50+kV. This (with the proper use of safety gap) rating
should be high enough for at least a 12kV, if not even for a
15kV xformer. It is exactly this DC failure rating (5kV for a
1650V rated cap), which lets me say it is useless to go for
really high voltage DC caps. Some of the GTL members
do not agree, for whatever reasons, and they went ahead
and ordered the 6kVdc units. (which is why I sort of lost
out, as I too ordered caps) If you give Richie and my 4x
Fmains SRSG idea a try, the "failure time" is only 5ms.
This means, if you can keep the voltage within limits for the
5ms, the SRSG needs to fire the next time, there will be no
danger for the MMC. Vdc is should be rated so that you
*should* never exceed it in normal usage (1.41*Vac) and
2-4x this DC rating should match the possiblilty of a strike
to the primary (so that the MMC will survive). Greatly raising
the DC voltage rating (i.e: using caps that have very high
DC voltage ratings; WIMA 6kV) is a waste of money, because
the DC rating really only tells you the voltage peak (single
pulse case) the caps will survive. There is no use in designing
your caps, so that they will survive 150kV, if your xformer
lets go at 20kV. (I hope you see what I mean to say).

3.Vac rating:
a.) Vac rating (as far as I understand this slightly
     mysterious rating) has to do with:
b.) internal heating.
c.) partial discharges.

(Side comment: DC to AC derating factors on WIMA
caps seems to be very reasonable up to and including
the 2kV units. The 2kV units are rated for 700Vac. What
I ABSOLUTELY do not understand is why the 4kV and
6kV units are also only rated at 700Vac. Even WIMA
could not answer this question directly (they used vague
words). For a 6kV cap, the 700Vac is a derating factor
of 1:8.5, where as the 1650/650V is a much more
reasonable factor of 1:2.5. This is another reason why I
donīt have much faith in the 6kV units.)

b1.) Internal heating: As we are using one of the lowest
       loss materials (PP) and our run times are limited to
       <30 minutes, this is of no real concern. Unless we
       really push them hard, they should not get too hot.
       PP should not be run above 105°. However, this is
       internal temperature. As we do not know how fast
       the heat convects to the outside, I would say the
       outside should stay below 40°C. 40° is slightly
       uncomfortable to the touch.
b2.) Keeping the rows seperated by 1/2-1" will allow plenty
       of ambiant air to circulate and cool the MMC. I donīt
       think a fan is necessary. If a fan (must be) used, I would
       go to the trouble to filter the air. You donīt want dirt, etc
       to build up on the caps. This will reduce heat transfer and
       might lead to flash overs.

b3.) Following b1 and b2, a small HF coil, powered by a tube or
       FET, will need other considerations. As I have not built one
       of these, I canīt comment on them, but I would think an MMC
       for such a coil will need different construction thoughts to keep
       the temperature down low.

b4.) Higher temperatures will soften the dielectricum and degrade
       itīs electrical properties. This will cause a change in
       capacitance and punch-through resistance.

(My) conclusions for internal heating:
This is of no real concern in the AC (de)rating for CW coils. We
should never experience ANY sort of real heating within the caps.
Caps that get hot, are telling us something is wrong and you had
better find out what the problem is.

c1.) Partial discharges. This is a problem, because it is the slow
     death of a cap. This doesnīt become obvious on the first

Side comment on WIMAīs 6kV vs 2kV: Partial discharges also
can not explain why the DC->AC derating factor is so rotten
for the 6kV units.

c2.) Partial discharges occur within the cap in voids, air bubbles
       and entrapped dirt particles (see my mail on partial
       discharges in the archives). Fortunately for us, the
       manufacturer does everything, he can, to prevent this from
       happening. Commercial cap material never sees human
       fingers. The high tech envolved (thin layers, virgin material,
       high pressure rollers, vacuum degassing chambers, etc,
       etc) lets the cap survive the hell we put it through.

c3.) One thing to remember is that the manufacturer (de)rates
       his caps for 24/7/365 usage. Our coils do not see this
       kind of operating enviroment. Further more, caps are rated
       for 10,000-100,000 hours of operation. The MTBF (mean
       time between failures) is rated in Gh!!. To make things more
       clear, lets use an example. For our needs, 2000h are more
       than enough. Furthermore, letīs say we run our coil for 30
       minutes per day, every day. For 2000h, this means we can
       run our coil for 2000h/0,5h = 4000 days. 4000/365 = 10.95
       years!! This is more than enough, I think ;o).

c4.) In order to keep partial discharges within (what I claim to be)
       safe values, I would suggest staying below the factor 2.5. I.e.:
       your 1500Vdc caps (450V ac) should be run at or below 1125
       Vac. The lifetime does not go down linearly. I havenīt made a
       graph, but from gut feel, I think with a 2.5x Vac overrating, you
       will be able to reach a 1000 hr. life span with ease, running the
       coil as described above. Hardly anyone runs their coil EVERY
       day of the year. If we further say, we only run every other day,
       we will still get a lifetime of ~11 years.

(My) conclusion partial discharges vs. AC rating:
This value should not be totally overrated in order to get a reasonable
lifetime out of the MMC. A factor of ~2-2.5 seems "about right". We
shouldnīt forget this is a statisical value and not ALL caps will go at
once. As the MMC is built in a modular form, replacement of single
(even if it is a few) caps is not too difficult or expensive. If you are
(generalizing here) short on cash or just want to experiment around
and take a shortend lifetime into account, you can exceed this rating up
to maybe a factor of 3-5 (i.e: your caps 1300-2250Vac each).

End conlusions to MMC construction:
Considering the cost of a rolled poly AND the fact that it
becomes a doorstop, once it has failed, I think the MMC
has good chances to become THE standard in coiling. I
can see no problems with current capabilities (section 1),
peak (DC rating) voltage capability (section 2), as long
as it is sensibly matched, and no real problems in lifetime
vs AC rated vs. AC useable voltage (section 3).

>Polypropylene Axial Caps
>dv/dt = 1800V/uS
>PF at 10KHz < 7*10,000
>0.047uF Capacitance tolerance   ą20%
>Rated voltage   1500V d.c. working (450V ac.)
>5-45 -at- 46p, 50-95 -at- 41p, 100-495 -at- 35p  +VAT

While the long mail above might not directly answer all your
questions, it pretty much sums up the knowledge and
experience, I have gathered with MMCs. Other info, which I
have posted over the past weeks, (on actual construction,
spark length vs. input power, calculations, etc) can be found
in the archives. If I had reposted all this, the mail would have
been rejected because of the enormous length........ ;o[!].

Your dv/dt sounds good, the 47nF is a sensible size (I think
50-100nF per single cap is a good range to shoot for) and
the cost is not a real prohibitation either. Going for an order
with several coilers, will reduce the price to 35p (test a few
units before you buy the bulk packs ;o)). Judging from dv/dt
and your description, these seem to be caps with metal
FOIL endplates. Axial caps will take a bit more space to
mount (in comparison to radial caps) and might be difficult to
friction fit them into an enclosure (for ease of servicing), but
this wouldnīt stop me from using them. Of course, if you can
get radial ones (retangular form) for 2-5p more, I would
rather use these.

>I'll probably buy 10 to start with and try them in parallel with my
>existing Cap.

Donīt be surprised if your spark length increases (for similar
capacitance), when you start feeding your coils on MMCs
alone. This is exactly what happend to me.

Coiler greets from germany,