On 2006-05-25, Jim Lesurf wrote:
In article , John Phillips
wrote:
Also I would point out that I believe the hysteresis observed is not,
per se, a linearity issue. I think hysteresis will arise from parasitic
series inductance or resistance, and also from dielectric absorption.
These may well be defects from ideality but in spite of the article's
title are not capacitor linearity issues.

Indeed. And may also in practice be orders of magnitude less significant
than implied by the curves when we move to a more relevant set of
conditions of use.

This is certainly true of the linear effects of dielectric absorption
(DA). I see audio cable and capacitor sales literature pronouncing
that energy storage in DA and its later release "smears" transients.
Well, if it's a linear effect [1] it does cause ripples in the frequency
response so it may do what's stated. However the magnitude of the effect,
when I plug in some real numbers for DA, is several orders of magnitude
below what should be audible [2].

If there are audible differences between capacitors there's no evidence
yet that DA is the (or even a) culprit.

[1] The DA models are all linear, but I have searched occasionally
for any evidence of non-linear effects arising from DA. These may
be audible if big enough. However the one mention I found was for
semiconductor-insulator (e.g. Silicon - Silicon Nitride) boundaries but
it provided no details and no references to follow up.

[2] Based on reported levels of audibility arising from frequency response
differences. However, recently I have become interested in the human
sensitivity to audio arrival time differences. Detectable differences
as reported seem to be smaller than implied by 20 kHz hearing limits.
I am looking out for more evidence on this.

--
John Phillips