Re: Lifetime of CD942C20P15K is ~75 hours

["Tesla list" <tesla@xxxxxxxxxx>]

Original poster: Terry Fritz <vardin@xxxxxxxxxxxxxxxxxxxxxxx>

Hi Jon,

At 10:21 PM 10/2/2005, you wrote:
> That's good to know, Terry, thanks for the performing that experiment. One
> thing it got me wondering about concerns the failure mode of ionization at
> the plate edges. Specifically, would there be half the amount of ionization
> at half of the voltage?

We know(?) that below 630VAC there "should not be" ionization, so the 
lifetime jumps to the usual say 200,000 hour MTBF numbers...  As that 
voltage is increased just a little, I would guess that the ionization 
might occur on the portion of the sine wave above 630 * SQRT(2) 
volts.  So there is certainly a "threshold" effect at 630VAC or 
891Vpeak.  Then it is a function of the time that the sine wave is 
above 891V or the ionization threshold.  But there must be a power 
function in there too since ionization at say 1000V is a whole lot 
less than say 10,000 volts.  But I have no idea it is proportion to 
V, V^2, or some other function (temperature??, ionization "start up" 
time??)...  Maybe an RMS function...  I have never seen any data on 
capacitors operated in such regions...  I wonder if there is any 
detectable "power" increase in the cap's dissipation as ionization 
begins and with various high voltages?  As far as I can tell, 
manufacturer's data on such issues is "zero"...


> What I'm getting at here is the possibility of using twice as many
> 942C10P33K caps per string (at $1.68 each) instead of the 942C20P15K units
> (at $2.45 each).
>
> Looking at 12 of our "old faithful" 942C20P15K in series yields 12.5 nF; 11
> of these series cominations in parallel gives 137.5 nF at a cost of $323.00.
>
> Looking at 24 of the 942C10P33K in series yields 13.75 nF; 10 of these
> series combinations in parallel would give 137.5 nF at a cost of $403.00.
>
> It would end up costing 25% more than the 942C20P15K caps, and there would
> be twice as many of them to deal with.  If there would be a significantly
> longer MTBF using twice as many of them per string, perhaps it might justify
> the added expense and hassle of using twice as many.

If say 50% more money would give us 5000% more life, then it could 
get real interesting ;-))  But we don't have the data or the life 
graphs to plug into the spreadsheets and equations.  It is all too 
goofy to just guess at...

Right now, I have to take down the equipment to play with the 
OL-DRSSTC stuff...  But I should return to this to see if there is a 
great advantage to be gained if we pull back on the peak voltage just a bit...

I do note that CD pulls back their rated AC voltage to "500VAC" 
instead of the usual "630VAC" for poly caps...  I get the feeling 
that it is all very unknown or top secret...  From the "cap 
engineers" I have talked to, I would guess they don't have a clue...

We run them over voltage to save big cash, but one could fall back to 
500VAC per cap where lifetime is all important and money is 
not...  There are thousands of applications that have proven that model out...

Cheers,

        Terry



> Jon
>
> > Original poster: Terry Fritz <vardin@xxxxxxxxxxxxxxxxxxxxxxx>
> >
> > Hi All,
> >
> > I finally did the great CD942C20P15K MMC capacitor life test.
> >
> > I ran three of them at 1414 VAC (60Hz) which is 2000Vpeak.
> >
> > http://hot-streamer.com/temp/GreatCapTest-002.jpg
> >
> > http://hot-streamer.com/temp/GreatCapTest-004.jpg
> >
> > http://hot-streamer.com/temp/GreatCapTest-001.gif
> >
> > The goal was to see how long the dielectric would hold up to
> > ionization at the edges of the plates which is the limiter. [snip]