X5R C-V data

That's a curve of dQ/dV versus V_bias, I presume? That's the significant value when you're bypassing a DC bias voltage source. The capacitance could also be considered to be the total charge Q(V)/V versus V_peak, which would represent an AC current through the capacitor to ground better than dQ/dV does.

Capacitance being an intrinsically linear property, the graph alone is ambiguous.

Cool! Thanks, some manufacturers are coy about those graphs.

So at 10V two 2.2uF would be almost as good as a single 10uF?

piglet

It looks that way. I suspect that the root cause is the higher capacitance in a given package forces thinner dielectric films, increasing the voltage stress per volume of dielectric.

Also, different dielectric compositions may be used.

Joe Gwinn

Yes, less inductance too.

Our original question was whether it's always better to use a cap with a bigger nameplate capacitance as a bypass. Probably yes. In this one case, the 10u always wins.

I have some power pours that have noise from switching regs, in the hundreds of KHz range, 1.2 to as much as 10 volts DC. They are filtered by a ferrite bead and a lot of 1 uF bypass caps. The bead is a short at these frequencies. The fix is to use a real 47uH inductor and change the caps to 10u.

Ferrite beads and chips do have finite inductances. LTSpice has the Wurth range in it's list of models, and they go from 3nH to 14uH.

It's a pedantic point, but one that I've had to worry about (and add a damping resistor to deal with).