[Prev][Next][Index][Thread]
Re: MMC potting
Hi Dave,all,
Comments below. The old are in quotes and the new
ones are placed between stars.
Original Poster: "David Dean" <deano-at-corridor-dot-net>
Original Poster: "Reinhard Walter Buchner" <rw.buchner-at-verbund-dot-net>
I wrote:
"1.) You say, you are using two strings. This means each string
will have to take 50% of the primary amperage. Depending on
design (of the coil) this might lead to severe overstressing of
the caps. Do you know how high your peak and rms primary
current is? What power are you running?"
You answered:
"Please correct me if I am wrong, but it seems to me that the amperage
through one string would be exactly the same if there was only one
string as it would be if there were a hundred in parralel. The amperage
through a string would be dependent on the capacitance of the string.
the voltage it was charged to just before it was to be discharged, and
the resistance (make that impedance) of the primary circuit."
*****************
No, not quite (if I understand what you are writing correctly).
Let us imagine a primary current of 600A (which is what I
happen to be running at the time). Let us further suppose we
need 20nF -at-10kV. Now lets take a look at two MMCs:
Case #1)
Each MKP cap has 100nF -at-1kV.
I need 2 strings of 10 caps each
(so I need 20 caps total)
Check: 100/10 = 10nF * 2 strings = 20nF
Using Ohmīs Law, we find, that each string must supply
300A (600A/2 strings). This is the current per string.
However, as each string must deliver 300A, the current
must flow through EACH cap in the string. If you now
back calculate the dv/dt from the amperage (actually from
di/dt), you will easily exceed the rated dv/dt of even high
dv/dt (like 10,000V/ĩsec) rated caps. This is one rating
which should NEVER be exceed, otherwise the caps
wonīt last for long. A very high current pulse can wreck
lots of havoc, like evaporating the metal foil. It could
even cause the cap to explode. It will in any case cause
local hot spots, that will degrade the dielectrica over
time.
Case #2)
Each MKP cap has 50nF -at-1kV.
Using 10 caps in a string (as above) I now need
4 strings (or 40 caps) for my 20nF cap.
Check 50/10 = 5 * 4 strings = 20nF
In this case I have the primary current of 600A running
across 4 strings. So now, each string (and each cap)
only has to supply 150A. This will allow the MMC (esp
if it is an EMMC) to run cooler and it WILL last longer.
*****************
I wrote:
"2.) You write, that you potted them in epoxy resin. Why? I
can see NO advantage in doing this. However, there are a
number of disadvantages doing so:"
You answered:
"Because I have epoxy resin sitting one the shelf. I do not
have any perf board or any pc board material lying around,
Snip
incorperate well into my design. I need the cap mounted
vertically."
************
Okay, what I really meant is why pot them at all? You
could just place the string in a pvc pipe and cap the
ends (like you did) and not use any potting compound
at all.
*************
I wrote:
"a.) Internal heating of the caps canīt be removed. The epoxy
will act as an insulator."
You answered:
"I don't buy that. Epoxy has a fairly high thermal mass. It absorbs a
lot of heat. It conducts it rather slowly. It acts not as an insulator
so much as like a thermal flywheel."
Snip
"just a few short bursts from time to time, I think it will work
O.K. Most of the reports I have read on this list describe
the heating as not noticeable to bairly noticable. If I am
wrong, time will tell, and I will let you know."
***************
You are correct. I should have stated that the thermal transfer is
slowed down quite a bit. This is a real problem when using epoxy
encased PTs. These can be overrated, but not as hard as oil
cooled units. My MMC stays cold on the outside, even after a
10 minute (really clocked) run. However, if you are venturing
into the EMMC design, you will experience heating. Dielectric
heating must be kept as minimal as possible. Heat is one of the
things that really destroys the dielectrica.
*****************
I wrote:
"b.) This setup is unrepairable. One of the MMCīs advantages
is easy repair (in case something blows). By potting them, you
defeat this advantage."
You answered:
"It is really easier to unpot something from epoxy than the tar in NSTs.
All you need is an oxyacetalyne torch with a small tip, a stiff brush,
a little skill and some patience. You play the flame across the
epoxy and it turns all crumbly kind of like dried out jello.9 If you
heat it too much it changes color, gets darker.) Then brush the
crumbs away, and flash with the flame again. I have used this
method to unpot timer modules and solid stat relays so I could
repair them in the field in emergancies when replacement
parts were not available or would take too long to get."
*****************
True. You CAN get at them, but w/o potting this would be
much easier. Plus, if you get distracted during unpotting, it
could cost you your caps. I just canīt see the advantage
of potting them at all, because it doesnīt help the design
survive, just as putting them in oil doesnīt help them,
either.
*****************
You wrote:
"I have done it many times just to see what is inside the
little magic black boxes."
*****************
Ah yes. The inquistive mind. Everything (and I mean
everything) I buy, find or can get my hands on, will
fall to the screwdriver. I can *almost* tell a good
design just by looking at it from the inside. If you have
an amplifier (for example) that has resistors, etc all
criss-cross wired on the underneath of the PCB, you
can bet it is still in the design "correction" stage. If it
was an expensive piece of equipment, you can also
bet your socks that I will bring it back and get another
(different) unit. Plus, you get an idea (from looking
at the circuit layout) how things work. Sometimes I
will even measure a few voltages at various points
in the circuit and jott them down. That way, I have
an idea, where to look, if something dies inside in
a few months or years (following Murphyīs rule,
usually one day after the guarantee is over) later.
I canīt remember the last time I took something to
a shop to be repaired. P.S: I was really amused,
when I saw Chipīs slogan for the first time ;o)). I
liked it!!
*****************
I wrote:
"c.) The epoxy will do nothing for corona suppression, simply
because you wonīt experience any ;o) (outside) corona. As
for internal corona (partial discharges), the epoxy does
nothing to prevent this."
You answered:
I don't know what you mean by internal corona. If you mean inside the
caps, then there is nothing I could do about that anyway. If you
mean corona at the points where the leads are soldered together, I
would not expect too much there anyway as i twisted the leads of the
caps together and then wrapped the resistor leads around them,
soldered and snipped the exess off so that makes some ball shaped
blobs which would not emit much anyway. As far as the epoxy causing
a problem with corona, I have seen too many high voltage parts that
were potted in epoxy work fine and last a long time to worry about
that.
*****************
The internal corona is what we are fighting most. This is the
*real* factor determining the V/mil a specific material can
take. If you exceed a certain voltage per mil, your dielectrica
will experience internal corona. This looks (only that it is en
miniature) just like the corona coming off a HV needle point.
It literally sprays in every direction. Any air entrapped (and
all caps, even high tech pulse and commercial designs,
have ever so small voids, where air is entrapped and canīt
get out) within a cap will really enchance this corona. This is
called partial discharges. The effect you see is called treeing,
because of the shape of the damage done. Check the
archives for my mail about partial discharges about 3-4 months
back and / or have a look at Gary Lauīs site. He posted a
specimin of his extended foil cap, which Scott Hanson had
analysed (he also sent me a copy of the picture). This is a
VERY beautiful example of PD and treeing. The problem with
entrapped air is the radical change in the electric field lines
(because of the "K" change from the dielectricaīs high K
to the airīs low K value). This will also raise havoc. This and
PD are probably the most causes of death among rolled
poly caps. The next problem we have to deal with, is voltage
creepage ALONG the surface of the dielectricum. I once
saw a 7.5kV xformer arc over a distance of 3" (!!) because
of this effect. The voltage can cross (arc) much longer
paths on the surface (slow build up of electrons) than it can
in pure air. Corona inside (and near) a dielectrica, like PP
or PE, will degrade the material, not to mention the heating
involved. Obviously, the epoxy does nothing to stop this
effect on the inside. Outside corona is not a problem with
the MMC design as you are not overvolting the cap or casing
(and lead spacing) by a large factor (>>8-10x Vrated). The
epoxy doesnīt hurt in this case, but it doesnīt help, either.
*****************
Snip
I wrote:
"e.) Voltage versatility. Your setup also defeats the voltage
versatility of the MMC design (as you canīt change the
number of caps in an expoxied string)."
You answered:
"I just can not think of any reason why I would want to do that. I sure
don't want to make it any shorter, I'll just let someone else find out
how many Es can be strung in front of MMC. And as far as making
it longer, it is a question of economics"
Snip
*****************
Well, one reason could be, if you change xformer voltages or if you
want to run the MMC on several coils, etc.
*****************
Snip
You wrote:
"So now we get to the particular advantage of my EMMC potted
in a pipe. The cap will be a structural member to support the primary
as the primary will have to be elevated above the case to keep too
much energy from being absorbed by the case. That will reduce the
amount of primary "plumbing" required which will reduce the stray
inductance. And if it were to work out well, then I could build more
and bigger caps and run it equadrive and have the caps be the
only structural members holding up the primary, with no plumbing
between the spark gap and the primary because the caps would be
the plumbing, and if I made the primary a double flat pancake using
mutual inductance as the (fine) tuning method, there would be even
less stray inductance, resistance, I^2R loss etc. ! And I don' have
any busses to introduce losses in this cap. The leads from the end
caps are twisted together and soldered into the slots in the heads
of the machine screws which are the terminals."
****************************
That (the coil, etc mount) sounds like a real neat idea. I like it
(esp. from the looks), but I wouldnīt worry about I^2R losses,
etc too much. The (by far) lossiest part you have in (any) TC
setup is the spark gap. Even if you add all other losses
together, the spark gap will still outweigh this by a large factor.
What I was really nitpicking about, was the fact you potted them
at all. I just donīt think it is necessary or desireable. BTW: The
equadrive is a strange animal. You will need to keep the nF
values in both caps (in each primary leg) pretty much equal and
you need twice the capacitance as the caps are effectively in
series. It IS an interesting approach, but I canīt see any kind
of advantage. I like using the KISS (keep it supid and simple)
approach for most things
****************************
You wrote:
"As I understand it, you are not running any resistors across your caps.
I would be very interested to know if you have found the same thing
I did when I got shocked by the middle of the string when it did not
have resistors on it. As I said, I found some pretty high voltages on
some of the caps, and I found that the polarity of the voltages was
sort of random. Some one way, some the other. I wonder if this
might be happening while the caps are running in a coil. If so, then
the caps might be seeing higher voltage stress than they would
if the resistor were there. Just a thought. Anyway, I will always
use resistors on any caps that I build in the future no mater
what design I use if the caps have more than two plates."
****************************
I run a 12 x 13 (13 rows of 12 caps each = 67nF total = 16nF
for 15kV coilers) on my 7500V NST. (This is NOT an equadrive
system. If I ever go back to the equadrive system, I will most
certainly use Eq-Rs) I have run my coil quite a few times for
various lengths of time. I then carefully disconnect one lead
and measure the residual voltage across the whole cap.
Depending on how quick I am (in disconnecting the cap), the
highest voltage I have every measured was around 250V.
Usually tho, it isnīt that high. Many times I have only measured
50-60V and I found this voltage to be quite evenly distributed
across the caps. There ARE variations, but not enough to worry
about. I have never seen a variation of more than 15V. I HAVE
seen a random change in polarity (just like you). I WILL be
using Eq-Rīs (like I wrote to Mike the other day) in my FINAL
setup, but not because of residual or uneven voltage distribution,
but rather beacuse the caps can regain a charge during storage.
However, my final design isnīt finished yet. I will probably be
running around 200nF (50nF for 15kV coilers) capacitance,
6kVA in modified neons and a 200bps SRSG with the 12-13J
input energy. Primary current will be in the 1.5kA (Oops) range.
****************************
Coiler greets from germany,
Reinhard