Re: EMMC Value Drift
Tesla List wrote:
> Original Poster: Terry Fritz <twftesla-at-uswest-dot-net>
> Hi All,
> I seem to be noticing a very slight capacitance value drift in my
> cap. It has 10 caps per string and has 5 strings of 1600 Volt 0.056uF
> Panasonic high voltage pulse caps. The total value has gone from 27.9nF to
> 27.7nF. Each individual string seems to have drifted the same amount
> (-0.7%). This is close to the limits of my LCR meter but the drift seems
> real. After lasts week's run, it seemed to drop by about 0.3%. I ran it
> very hard this 4th and it seems to have drifted another 0.4% with that use.
> I run this 16000 volt cap at about 22000 volts so it is stressed in
> voltage. The strings seem to be fine current wise. I wonder if the dv/dt
> thing is degrading the internal connections?? Perhaps it will level out
> with more use but it seems to be continuing in a linear way with the amount
> of use. If it continues, eventually it will fail and I can dissect the
> remaining caps to see what the cause is.
You may be seeing the effects of internal shorting and clearing if
you're using a metalized PP dielectric. Certain newer style high energy
density storage caps use a thin (partially transparent) metalization
layer which has been evaporated onto the dielectric. These are designed
to be "self-clearing" during dielectric breakdown by actually
evaporating/removing the metal around the effected failure area rather
than simply arcing and shorting as with metal foil plates.
While this extends the life of a highly stressed energy storage
capacitor, multiple breakdowns eventually will reduce the effective
capacitance over time. Because of the construction of the cap, you're
actually seeing longer life than if you were to use a metal foil
construction - and your caps actually may have more than nine lives!
The downside of this style capactiro construction is that these caps
will have a higher ESR, and a lower dV/dT rating. If you're using
Panasonic's ECW-H style metalized PP cap, this is most likely the
-- Bert --