[sdiy] Re: Kemet Multilayer (was Walt Jung and Richard Marsh's Capacitors article)

[Magnus Danielson]

From: harry <harrybissell at prodigy.net>
Subject: [sdiy] Re: Kemet Multilayer (was Walt Jung and Richard Marsh's Capacitors  article)
Date: Sat, 27 Jan 2001 00:07:58 -0500

> Hi Chris...
> 
> Depends on where you use them...
> For the S/H I'd go with the polystyrene... the dielectric
> absorption is just a little better.  Or do you mean for the
> DAC multiplexer ??? I'd go polystyrene there for sure...

Either polystyrene or polypropylene, but if a teflon cap is at hand,
I'd probably use that.

In a S/H there are a few things to recall:

1) For high precission you *dont* want loss. A successive
   approximation method for instance will decide the lowest bit last,
   and if the voltage is drifting, then you are lost.

2) If possible, keep inductance low, since that would affect the
   settlement speed when the input is open.

3) Source impedance should be low at "open" for maximum settlement
   speed.

4) Source impedance should be high at "closed" for minimum stray
   currents either way.

5) Protect it with guard rings etc so no stray currents is there.

... and a zillions other things which becomes apparent as you are
starting to understand things...

> Are the Kemet a temperature stable variety ???  They ARE
> rugged.  We use them for decoupling... and even to simulate the
> gates of very large IGBTs... they are really tough and take a hell
> of a lot of AC dissipation.  But for audio I'm not sure.

Oh dear... now we are there again....

Ceramics big gain is in their very high relative dielectric values (I
hate to say constants here), you can easilly attain 1000 to 10000.
This dielectric "constant" comes at many prices, it's non-linear, high
loss and temperature dependent (just let me think more on it and I
hand you more ugly dependencies). The big benefit with a high
dielectric value is that you don't need as large area of this material
in order to acheive a cap of a certain value. This translates into
physical size of electrodes and thus "parasitical" inductance. This
later is THE issue for high speed decoupling caps. Plastic caps is
just plain gone (they turn into inductors as you go higher up in
frequency) as you go up while ceramics keeps acting like caps to
higher frequencies. This is where ceramics is *really* usefull.

In lower frequencies, like that of audio, I am much more
reserved. Sure, use them to decouple digital chips and fast
comparators etc.

Plastic caps usually have a dielectric constant of 2-3 or there
about. One of the killers here is the ability of the dielectric to be
doped by humidity. Here polytetrafloureten (PTFE or Teflon(TM)) rules,
it is so "fat" that it doesn't accept any water. The loss is thus low
and nonlinearity is very weak.

Actually, one must understand that dielectric constants other than 1
by definition means non-linearity, the higher dielectric constant, the
more non-linear. The loss is natural with higher dielectric constant
too.

The whole issue about dielectic constant is that the dielectric
contains molecules which is bipolar, but is suspended to neighbor
molecules by van der Waals force. As you apply a electrostatic field
over the material, the bipolarity of the molecules wants to allign
according to the field, but the van der Waals force interact with this
turning action, and thus provides a counter-force. This is the
connection between the electrostatic field and energy storage. The
stronger a molecules counterforce is, the harder it is for a applied
field to turn it, and thus the higher dielectric constant.

Cheers,
Magnus