[sdiy] Walt Jung and Richard Marsh's Capacitors article

[Grant Richter]

> It had a 16 bit ADC, and ahead of was a single
> pole low pass filter.  The problem turned out to be the low pass filter, made
> with a ceramic capacitor.  They were only getting 12 bit performance rather
> than the 16 they were expecting.  Changing the cap to a polystyrene fixed the
> problem.  It was difficult to explain to the engineer that a $0.05 ceramic
> capacitor was not the same as a$2.00 polystyrene.
> 
Could this be dielectric absorption?

I once had to fix a problem with a data acquisition design that where
readings were slew rate limited. The problem turned out to be the ceramic
capacitors in the input filter. Replacing them with polycarbonates fixed the
problem. The symptoms were on a step change from 10 to 0 volts, the system
would read 0.015, 0.009, 0.005 etc and "converge" to zero rather than step.
That was a 14 bit system and the caps were the Kemet golden ceramic.

As I understand it the symptoms of dielectric absorption are this "storage"
of electrons in the dielectric which then slowly bleed off. Ceramics have
about 1% dielectric absorption while polystyrene and polycarbonate have
about 0.1%. Jims experience would suggest the ceramics are good up to about
12 bits and a better cap is needed for higher resolution systems.

I suppose technically, this dielectric absorption would appear as
non-linearity? That is produce distortion in the waveform from the storage
effect?