[sdiy] ASM Capacitors/Bench Power Supplies

[Magnus Danielson]

From: Ian Fritz <ijfritz at earthlink.net>
Subject: Re: [sdiy] ASM Capacitors/Bench Power Supplies
Date: Thu, 31 Jul 2003 12:12:13 -0600

> At 11:34 AM 7/31/2003, Magnus Danielson wrote:
> > >
> > > Does anyone know of a reason not to use NP0 ceramics for the integrating
> > > cap? I briefly tried both NP0 and polystyrene in my last VCO and didn't 
> > see
> > > any difference in waveshape or tracking.  The NP0 has a smaller
> > > tempco.  Also it seems to me that the high-grade mica units, although
> > > expensive, should work better than polystyrene.
> >
> >In general, ceramics is bad due to leakage, memory effects, tempco etc.
> 
> Well, these are the words we always hear, of course, but what are the 
> numbers???

... and I agree with that! This is exactly why I wrote "In general" so we had
the backdrop clear and could go on to the specifics...

> Certainly the NP0 tempco (+/- 30 ppm/K) is significantly better than 
> polystyrene (-100 ppm/K).

Yeap!

> >If this is critical for NP0 in oscillators I can't say
> 
> Well, but that's what I'm asking!

Yes, I know! Cool it a few dB!

> >, but memory effects you
> >want to avoid and leakeage is certainly a low-freq problem.
> 
> How do you think memory effects would show up?

Well, at first I only thought "and memory effect". The next thought was
naturally "How the hell does it manifest itself here?".
My answer is this:

The memory effect works such that you will not completely "empty" the "state"
of the capacitor, or more correctly, the (nonlinear) dielectrum. When you empty
the capacitor at reset time, you may not be able to fully relax the strain the
molecules in the dielectrum had, so the voltage may rise again. Now, this rise
would make the chargeing to start early, i.e. there will be a increase in
frequency since from the release of the reset to the acheivement of threshold
has a shorter time than intended passed, due to the decrease in voltage needed
to be charged up.

The memory effect is sometimes modelled as a number of additional capacitors
in series with resistors all shunting the main capacitor (see for example
Bob Pease both in Pease Porrige and his book), but that remains to be a model
of the electrical behaviour and not the physical behaviour.

> I would think of memory involving some small change in capacitance from the
> cap being "soaked" at high voltage for a long period of time.  But would this
> be significant in a VCO where the voltage history is always a 0-5 V ramp (and
> the cap is rated at over 50 V)?

You assume that "soakage" is only in existens at near full saturation, i.e. at
about the marking voltage. Now, what if it also exists for lower voltage
levels.

Another thing, if it is a longer time issue, then would lover frequencies be
more affected than higher, since the oscillator is staying longer at near +5V
than for higher frequencies. This causes the problem to cause worse error than
just a pure scale problem (which he handle anyway).

>  If we can figure out what the effect should be, maybe we 
> can figure out an experiment to see it.

Indeed.

> I agree about leakage coming in at low frequency, and I didn't look at the 
> extreme low end when I compared ps vs. NP0.  From what I've seen, the 
> dissipation factor may be about a factor of 5 better for 
> polystyrene.  Usually, however, dissipation is modeled with a series 
> resistance, which does not produce leakage. Rather, it would have the same 
> effect as a Franco series resistance.  So it seems to me, it would be 
> compensated out in the high-frequency tracking process.

Pure leakage is a resistive shunt over the capacitor, memory effect is a bit
worse due to its distributed properties, so a bunch of resistor-capacitor
series shunts is needed for that.

> So to characterize leakage, it would probably be good enough to look at the 
> dc resistance.  Anybody have an electrometer handy?

Yeap! It sits next to my computer on top of my powersupply, why? ;O)

I got this lovely high resistance meter (as doubles as a low current meter)
which precharge capacitors with up to 1 kV and allows for leakage measurements.
Handy in analysing dielectrics... ;O)

Cheers,
Magnus