Aging and De-Aging of Ceramic Capacitors

[amspire]

I was looking at this information sheet on Digikey:

http://www.digikey.com.au/Web%20Export/Supplier%20Content/JohansonDielectrics_709/PDF/JDI_Training_MLCC-Basics.pdf

and the thing that interested me was not just the fact that these capacitances suffer a significant decrease in capacitance over time, but that this capacitance drop can be reversed by heating to the curie point of about 125 degC.

This seems to imply that if you take a board such as a PC motherboard that has had thousand of hours of usage and bake it, you may get slightly better supply filtering. Or bake an aged surface mount switching supply and the switching noise may reduce.

Does anyone know much about this effect?

It looks like this drift is related to capacitor quality because the MuRata data sheet shows about a 3% drift of a X7R over 1000 hours compared to 10% drift in the graph above. It is a logarithmic effect so that the drift from 0 to 1000 hours is about the same as from 1000 to 10,000 hours. If you design for a 50% reduction in X7R capacitance over its lifetime, you can probably ignore this effect.

[mariush]

This seems to imply that if you take a board such as a PC motherboard that has had thousand of hours of usage and bake it, you may get slightly better supply filtering. Or bake a a surface mount switching supply and the switching noise may reduce.

Does anyone know much about this effect?

Motherboards use ceramic capacitors mostly for decoupling (where actual capacitance doesn't matter much). Where the capacitance matters, they usually use polymer capacitors or electrolytics, or ceramic capacitors with slightly higher capacitance because they also have to consider the effect on voltage on the ceramic capacitor.

There's lots of plastic parts on a motherboard that would be damaged if you "cook" the motherboard at 125c, it's hardly worth the effort. Like I said, that +/- 2.5% is nothing, even +/- 10% is considered in advance when designing a motherboard.
Switching power supplies (and I'm thinking of ATX and similar computer power supplies when I say this) have even fewer ceramic capacitors and again, besides some very specific places (frequency setting, or to prevent oscillations), the actual capacitance is not really important, i'd say even +/-20% would be ok for ceramic capacitors.

There are switching power supplies or dc-dc converters which run at much higher frequencies (1mhz+) and on such power supplies the minimum capacitance may be more significant..

[BravoV]

That baking and toasting is quite long, from Kemet.

[sarepairman2]

so this is an effect that "starts" after soldering?what is the age trigger? some temp x degrees above curie?

do C0G capacitors experience this at all?

[T3sl4co1l]

so this is an effect that "starts" after soldering?what is the age trigger? some temp x degrees above curie?

Trigger implies a threshold or stepwise phenomenon ("new" vs. "old").  There isn't, it's only age related, as the graphs above show.

Same for the "last heat" process; the closer to Tc, the faster the "healing" goes (in an inverse exponential fashion: at Tc, time goes to zero).

do C0G capacitors experience this at all?

Both pictures say no ("NP0" == C0G).

Here's an anecdote: a random Z5U from my collection.  Probably had been sitting around for 5-10 years after having been desoldered once before.  Ceramic disk, possibly 16 or 50V rating, "203Z" the only marking (i.e., 20nF).



Annealing was performed by heating each leg with the soldering iron for ten seconds, at full soldering temperature.  (The disk melted and bubbled a little, as absorbed moisture bubbled through the sealing wax.)

Tim