In article , Don Pearce wrote:
On 07 Jan 2005 07:31:51 GMT, John Phillips
wrote:

I remain puzzled about the postulated audible superiority of teflon
dielectrics (and occasionally paper-oil - for the right type or oil
I assume).

1. The published curves for dielectric absorption etc. show what I
think of as trifling differences compared to polypropylene,
polyethylene and some other dielectrics, There are several
dielectrics which should be audibly good enough to be
indistinguishable.

Do bear in mind that the tan d loss is in series with whatever circuit
impedances are present. A tan d of the odd ohm or so is massively
swamped by the many kilohms of the average circuit coupling situation.
So in most circumstances of capacitor use, tan d is simply not a
relevant issue.

I think we are in agreement here.

2. DA and the other usual capacitor defects are primarily linear effects
regardless of what people imply when they talk about smearing of
pulses. They result in ripples in frequency response (which are
mathematically equivalent to the "pulse smearing"). These ripples
can be made so small by good design of the circuits surrounding the
capacitor so it should be a non-issue. Loudspekers have ripples
in frequency response that are orders of magnitude worse.

Pulse smearing or dispersion is nothing to do with the quality of a
capacitor, but its value - or more specifically the type of filter in
which it is used. A Bessel filter and a Chebyshev filter might both
use the identical capacitor. The Bessel will not smear, the Chebyshev
will. This effect is found particularly in loudspeaker crossovers,
where filters are typically operating in the centre of the audio band.

Well, "pulse smearing" as used by those writers who see dielectric
absorption as a problem comes from the capacitor model with series R-C
networks in parallel with the capacitor. A certain proportion of the
LF capacitance comes from the parallel network and retains charge from
pulses which then "trickles out" or "smears". I use their words and
meanings here.

This parallel network certainly modifies the circuit's
frequency-domain behaviour a little but if 1% or less of the
capacitance is in DA then the modifications to frequency response
usually drop below what is normally thought of as audible. See
http://www.national.com/rap/Applicat...570,28,00.html, figures 2
and 4 for examples of capacitor models, and figure 7 for the effects of
different dielectrics.

--
John Phillips