I looked up the DC voltage characteristic of the TDK X5R cap in question. YIKES!
Capacitance loss is way more than I had been led to believe.
Here's the smaller 2012 package size Jules listed above:
Usually you can reduce MLCC temp and inductance loss by picking a larger 3216
package, but for 22uF / 35V you still have X5R dielectric, and it doesn't help a lot.
You lose "only" half as much capacitance at 22V, but your 22uF is now only guaranteed
to be around 2.5uF. That's where the 9Hz 3dB point comes from if that was an actual
in circuit measurement. By the way, these are useless above about 1MHz. For RFI,
you need to parallel them with smaller class 1 ceramics.
OK, back to distortion - the curve after the great fall-off in the first 12 volts is pretty flat.
So the delta in capacitance with a 100mV change around 22VDC resting voltage won't?
be that drastic. And that's why distortion measures low. A few millivolts in 22V isn't
significant.
I learned something today. Thanks, guys!
