[Prev][Next][Index][Thread]

Re: Multiple Caps: How and Why



Hi Bob, all,

(Please see note on my original post below, too)

>Original Poster: Bob James <nrgdude-at-pond-dot-net>
>While on the subject of series and parallel caps I
>would like to hear more about series caps of
>differing cap values and the attentuation effects
>(voltage dividing or multiplication effects) and
>associated current effects --as this seems to be
>a subject of very little attention....

We will need to divert this into two categories:

a.) Same caps within a "block", but different caps
in different "blocks" (using more blocks in the total
MMC).

b.) Different caps within a "block" or even a string.

Letīs define a block: A block is a set of series/parallel
caps connected in (uhm, well...) a block; better yet, call
it an array. Example: An arrangement of 14x12 strings
with each cap in this block being 62nF-at-1000Vdc (my
MMC). Okay, now lets start with perspective "a".

If I now add a second MMC block to the first one (aka
in parallel) which consists of 14x6 strings with the same
caps as above. No problem. Actually I am just increasing
the block size here and the amps per cap go down.

Same goes for an added array of 28x6 strings, if each
cap is now 124nF-at- 500Vdc AND the individual caps
have the same dv/dt*C as the ones above. In this case
the dv/dt per cap can actually be half of what the 62nF
units have as I=C*dv/dt.

If I now use completely different caps, say a 5x8 "block",
where the caps also have different dv/dt and/or Vdc (Vac),
we have to consider Ohmīs Law. A multiple "block" MMC
is basically just a divider, just as two paralleled resistors
are. In other words the "current-per-block" will divide up
in the same ratio as their capacitance to one another is.
Once we know this, we can calculate the "current-per-
string-per-block" Aka, how much current each capacitor in
the (separate) blocks sees. You will have to make sure
that dv/dt or di/dt is NOT over their rating. Usually this is
not a problem, esp. if you use caps which are similar
(like using 56nF caps in one block and 47nf caps in another).
However, if you have got a real exotic combo, this might
lead to problems, esp. if you use the EMMC design.

Prospect "b" should NOT be used in (E)MMC design. Using
different combos of caps (dv/dt and V for example) within a
string will lead to problems, since the voltage (& the discharge
current) is distributed very unevenly (It gets worse the further
apart the individual cap values are).  Eq resistors will not be
of any help and might start to smoke as they try to equalize
the differences. Think of the caps as resistors across a voltage
source. If your C/C ratio is 1:10, so will the voltage distribution
and this can easily exceed your capīs ratings.

Bottom Line: Using different caps in different blocks (array of
strings) IS okay, as long as you consider dv/dt and peak
pulse current distribution (keeping the C values close is best).
Using different caps within a certain string should be avoided
at all cost.

Remember, the bigger your total capacitance gets (all primary
caps together), the more current you will be pushing and the
stricter your choice of individual caps gets.

Further considerations would be:
a.) FRes of your coil (higher FRes stresses caps more)
b.) Voltage: lower voltage forces more current per pulse (the
     coils being equal otherwise).
c.) BPS: higher BPS rates stress caps harder


Tesla List wrote:
> Original Poster: "Reinhard Walter Buchner" <rw.buchner-at-verbund-dot-net>
> Resistors:
> --------------
> If you PARALLEL two resistors -at- 100 ohm you will get:
> 1.) a total value of 50 ohms  1/Rt= 1/((1/R1)+(1/R2))
> 2.) a total voltage capability of 1x the single R

There is an error here, which Bill, the Arcstarter, pointed
out to me. (Bill: Iīll take the 12 gauge ;o}). The equation
is correct, except for the fact that it DOES NOT deliver
1/Rt, but rather Rt. Itīs just an algebraic slip, (I already
brought the 1/x to the other side of the equation) but it
might cause some confusion on a beginnerīs path.
Sorry !!

Here is the corrected equation:
Rt = 1/((1/R1)+(1/R2))  (NOT 1/Rt)!!

For general usage:
Rt(a)= 1/((1/R1(x))+......+(1/Rx(n))),

Rt(a) is the paralleled value of all resistors
n: is the number of resistors in the parallel chain.
Rx: is the individual value of each resistor.

Example: 4 resistors 10, 20, 30, 40 ohms all in
parallel.

Rt(a) = 1/((1/10)+(1/20)+(1/30)+(1/40)) = 4.8 ohms

A quick and easy check, to see if the result is at all
correct, is to verify that the resulting Rt MUST be
smaller than the smallest resistor in the chain. In
other words, if a parallel a 1 ohm resistor to a 10
Mohm resistor, Rt MUST be smaller than 1 ohm
(somwhere around 0.99x ohms in this case).


Coiler greets from Germany,
Reinhard