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

Re: MMC voltage rating



Gary and all,

Why capacitors must be de-rated for AC or ringing duty is kind of
tricky. An excellent capacitor dielectric is an excellent insulator. Any
locally charged areas sitting on the free surface of the dielectric
(areas not intimately covered by metal) will be very difficult to
dissipate - these surface charges become "stranded", only slowly
dissipating through the very high leakage resistance of the dielectric
and its surroundings. 

Now suppose you charge a capacitor to an initial voltage +Vo, and then
suddenly reverse its polarity to -Vo. Because of the dielectric's low
conductivity, any locally stranded charges on the dielectric surface
near the capacitor's plate edges cannot move or change instantaneously.
In the worst case, nearby stranded charges, coupled with the capacitor
voltage reversal, can lead to an electric field stress between the plate
and the nearby uncovered dielectric that is equal to the total change in
voltage. In this case, a portion of the dielectric system actually
"sees" a stress equal to the total voltage voltage swing (2*Vo) - hence
the need for derating for AC or pulse duty.

Now if the local voltage stress is sufficient, one or more an areas near
the edge of the plate will break down, causing small electrical
discharges between the plate and the charged area - "partial discharges"
- that partially neutralize these stranded surface charges. However,
theareas of altered surface charge distribution may now become high
stress points during the NEXT voltage reversal (i.e., as in AC
operation). While a "slowly" changing AC voltage permits a degree of
surface charge redistribution to occur, a fast pulse transition
(particulalry with overshoot and ringing), or RF does not. That's why
this type of dielectric stress is frequency dependent. While the energy
liberated by any given partial discharge is quite small, the cumulative
effect of repetitive partial discharges results in eventual dielectric
degradation and failure. While your cap may behave normally, the
dielectric is being slowly destroyed - you don't want partial discharges
to be occuring in your caps if you want long life! 

Hope this helped!

-- Bert --


Tesla List wrote:
> 
> Original Poster: "Lau, Gary" <Gary.Lau-at-compaq-dot-com>
> 
> I'm also puzzled by the frequently suggested advice about multiplying the
> peak voltage by a factor like two to arrive at a "safe" voltage rating.
> While the peak-to-peak voltage seen by a cap may indeed be 2.0 or 1.8 times
> the peak AC voltage, at no point in time does the actual voltage across the
> dielectric exceed the peak voltage.  I don't see how the peak-to-peak
> voltage is at all relevant to the voltage rating.
> 
> Having said that however, I suspect that a cap's AC voltage rating does
> indeed need to be modified from it's DC rating.  The Wima data sheets on my
> web site clearly indicate that the effective dielectric strength diminishes
> with increasing operating frequency.  For PP film, at 200KHz, the effective
> strength is only about 30% of its DC value.  Terry, perhaps this is why your
> DC tests went so well but under AC conditions, you're seeing these air
> pockets developing?  Plus, there is this mysterious "ionization inception
> voltage" thing.
> 
> Regards, Gary Lau
> Waltham, MA USA
> 
>                 Original Poster: Finn Hammer <f-hammer-at-post5.tele.dk>
> 
>                 I have been pondering this for awhile, it gets more and more
> important
>                 to ask:
> 
>                 Since I have, say 20 kV on the piggs secondary, the peak
> voltage would
>                 be 28800 volts, right?
>                 If I set the gaps to fire at this voltage, the cap will be
> charged to
>                 this voltage, before the gaps fire.
> 
>                 But after this, doesn`t the primary swing opposite with
> around 80% of
>                 this voltage?
> 
>                 So, shouldn`t We design the caps around a P-P voltage of
> (PIG P-P Vout *
>                 1.8) in this case, 28800*1.8 = 51840 V.
> 
>                 Just wondring, Finn