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

Re: Cap AC voltage ratings

Hi Reinhard:

I did understand your test method but I fear that conclusions drawn by
that method will not guarantee safe operation.

I'm not sure that if a capacitor survives a one minute test at a
particular voltage, that it is safe to assume it will endure that stress
for any other length of time.  The vendor data that I've seen
described the AC voltage rating as being something that will not allow
ionization to occur.  I suspect that if that did happen, it would not
result in instant failure but rather a slow degradation of the
dielectric.  So, I fear that while a string may survive the initial test,
it may slowly degrade to an eventual dielectric failure.

Further, the message in the Wima data sheet was clear - dielectric
strength is reduced from it's DC-rating as a direct function of operating
frequency.  I fear that the success we've had is just due to the generous
overdesign factors that responsible vendors use.  Plus using these
devices for much shorter durations and total operating hours than
conventional applications require.

As far as whether dielectric breakdown voltage may be assumed to be
tightly controlled - If a manufacturing process has the option of
controlling a parameter to plus/minus X% or minus-0/plus-anything, it's
much easier to choose the latter.  While the vendor can control the
average dielectric thickness quite precisely, thus controlling the
capacitance, if the dielectric is anything less than perfect, i.e.
blemishes, dimples, imputities, differences in dielectric puncture
strength will result, and without affecting the capacitance.  Thus, there
will be a range of dielectric puncture strengths while maintaining a
tightly controlled capacitance value.  The vendor is not going to assume
that his dielectric is perfect, so he's going to over-design the average
thickness to allow for a reasonable range of defects that will not drop
the strength below the unit's rating.

>BTW: Did you get the XLS sheet and my private mail?

Thanks, I did, but unfortunately my German skills don't go much beyond
counting to ten!  I've also been feverishly working to complete my own
MMC.  Last night was it's first test run for a total of about 4 minutes.
So far, so good.  I want to log some more time on it before I declare
success, and then document it on my web site.  After the 4 minute run it
was just slightly warm, but only just slightly, much less than my rolled
poly or Fair Radio Sales caps.

Regards, Gary Lau
Waltham, MA USA


>Original Poster: "Reinhard Walter Buchner" <rw.buchner-at-verbund-dot-net>=20
>
>Hi Gary, all,
>
>Original Poster: Gary Lau  19-Mar-1999 1622 <lau-at-hdecad.ENET.dec-dot-com>
>
>It's encouraging that you were able to AC-operate your PP caps at their
>DC voltage rating.  But I'm not so sure though if I agree with your
>technique to determine their breaking point.
>While a cap's capacitance value is a tightly controlled parameter, it's
>breakdown voltage is not.  {numbers that follow are speculative:}
>If a manufacturer sells a cap marked 1000V, I suspect that if you bought
>a hundred of them and destructively tested them to the point of
>breakdown, you might find values from 1500V to 10,000V.  The 1000V is a
>guaranteed minimum.  There is no maximum.  If the test string that I used
>happened to be built with some of the higher-valued breakdown voltage
>devices, I will get a very optimistic "max" voltage rating.  But then
>when I build the big array of 200 of these, I will have many of the lower
>voltage parts mixed in too, and when these break down, the voltage across
>the remaining parts goes up, then they fail too.
>
>I think we are misunderstanding each other, here. Let me use an
>example of what I mean. Except for the values, this is basicly what
>I did.
>
>Given:
>10kV xformer
>100nF caps rated at 1.4kVdc
>(I am using these numbers to make it easy).
>
>Our 10kV xformer put out a peak voltage of 14kV (roughly). So for peak
>AC=3DDC we would need 10 caps, right? To test the "Tesla properties" of
>these caps, I would build a test string of 30-40 in series. This would
>give me a 30-40kVdc rated cap. Measure the capacitance of each cap
>and the total capacitance of the string. Now add this MMC to your tank
>circuit and fire up the coil (tune for longest spark). Assuming the MMC
>survives, short one cap, re-measure and compare the capacitance
>values of the remaining caps. Also take note of the new total
>capacitance of the string. Fire up the coil again, re-tune for longest
>spark. Repeat this process, keeping close tabs on the single cap
>capacitance. Once you reach peak AC volts =3D DC RATED voltage,
>STOP (in our example this would be a string of 10 caps). Going any
>further is too risky, because (like you said) you might be overloading
>some caps. (if the string breaks down before you reach peak AC=3DDC,
>stop there, of course). Now you know the maximum voltage that your
>MMC can take. For added safety I would add between 30 and 40%
>safety margin. This means for our example (assuming they survive
>down to peak AC=3D rated DC) I would go for a string of 13-14 caps.
>
>I hope this made things a little clearer. You should NEVER go any
>higher than peak AC volts =3D DC volts (doing so is playing Russian
>Roulette with your caps). Keeping close tabs on capacitance is
>important, because any degradating of the dielectric will show up as
>a loss of capacitance. Sorry, if I didn't make this point clear enough.
>
>I'm not sure I can agree with your first statement in the second
>paragraph. If we say that the capacitance is a tightly controlled
>value (agreed), then the breakdown value has to have similar
>tolerances. Why? Well, the capacitance is the result of the dielectric
>K and the spacing between the plates (i.e: dielectric thickness). So,
>if the capacitance is to be uniform on a batch of several thousand,
>the spacing (K is uniform for virgin material) has to be pretty exact.
>This means the breakdown voltage will be similar for all caps (made
>to "x" specs), because the v/mil breakdown rating is pretty constant for
>virgin material. Assuming a certain "x" 1kVdc cap will fail at 3kV
>(numbers are just spectulative), then the weakest might fail at 2.8kVdc
>and the strongest might fail at 3.5kV. I don't think the failure voltage
>has such a huge variation as you thought (1-10kV).