From: firstname.lastname@example.org (Tom Bruhns)
Subject: Re: How to measure C+stray to 0.01pf
Date: 2 Sep 2002 10:51:26 -0700
NNTP-Posting-Date: 2 Sep 2002 17:51:26 GMT
So, Sally... To give you a little encouragement on this (as opposed
to those saying you have no hope of doing it), I have a simple little
breadboard hack that measures capacitance to the sort of resolution
that you're looking for. When I get more of those "round tuits," I
should be able to make it a bit more stable by getting it off the
breadboard. It doesn't work like the one you describe, however.
With repect to capacitors, there are some factors that you should
consider. DA, if you run the capacitor through the same cycle each
time you measure it, should (very nearly) cancel out. DA is an effect
you see mainly at low frequencies, so if you make your measurements
quickly, you won't see it.
At 10fF resolution on a 10nF capacitor, that's one ppm, and most
common low-tempco capacitors still have 10 or more ppm/C drift rates.
A fraction of a degree can change your reading by your resolution,
then. One _can_ make very stable caps, but it gets tough. We used to
build caps in a vacuum, with plates spaced by sapphire balls, and of
course they were pretty darned expensive. We needed low-value caps
(just a very few pF) with ppm/C stability. You CAN get vacuum
capacitors in the 1000's of pF that are physically not very large;
they are typically done with interleaved cylinders of thin copper.
You can get them in fixed values or as variables. You likely could
find some on ebay even.
Some capacitors also have noise. I've noticed this in high-K ceramic
caps especially. With no voltage across them, there's no noise, but
with a voltage across the cap, there's a noise which looks like a
broadband noise current in parallel with the capacitance; at higher
frequencies, the capacitor shunts the current, and you don't see it,
but at relatively low frequencies, it's quite apparent. (It's easy to
discover things like this when you're building spectrum analysis
equipment that goes down to DC and has sub-millihertz resolution...)
Since polystyrenes are strongly discouraged in new equipment design, I
haven't bothered to look at how their DA is. I know that polyester
(Mylar) is poor and polypropylene is at least an order of magnitude
better. I understand that mica is quite bad! I would expect that
high-K ceramics would be bad. Teflon is supposed to be good, and so
is polyethelene (but where do you find that?).
Was the ceramic you tried a high-K type, or a C0G? Have you tried any
C0G 10nF caps?
If you have a way to metalize things, you could try making your own
stable caps. You can get thin fused quartz sheets at a pretty low
price. But still, making a 10nF reference cap that way would take a
significant stack of plates.
For pF values, beware of other effects: changes in air pressure and
humidity cause changes in the external capacitance, and in fact, that
stray capacitance can perhaps be significant (at the fF level) no
matter what the capacitor you're measuring.
"sally" wrote in message news:...
> "Issac Asimov" wrote in message
> Dear Sally,
> Read Bob's comment, he is right. But even if we ignore stray
> capacitances(which we never can), how could you expect such a high
> precision? If you are measuring 10 nF and expect down to.001 pF
> precision, then all your measurements(current, voltage and time) must at
> least have a precision of 0.0000001 (=0.001pF/10nF). If you read
> datasheets of the capacitor, you will notice a degree Celcius change in
> room temperature will cause much more deviation in capacitance than
> 0.001 pF .. and even your position in the lab actually can easily change
> your capacitor value much more than the precision you expect . Things in
> Physics are not like in Mathematics !
> By the way, how could you reach this high expectation ? Are you sure
> there is not a flaw in your design ?
> Best wishes,
> I should have made things a bit clearer. I'm using a 3-measurement
> technique to measure the mutual capacitance between the leads in a
> shielded twisted pair. The measurement process involves measuring the
> total, including stray-capacitance. The 10nf cap is an onboard cap used
> to check the circuit. The measurement requires that the sum of this
> capacitance and stray can be measured consistently to 0.01pf (not
> 0.001pf). When I try it, there is 1pf noise. I have verified that this
> is partly down to mains pick-up by measuring 1/2 a mains cycle later and
> averaging the two results to get 0.1pf noise. When I put in a ceramic,
> the problem is a lot worse so it perhaps is DA; polystyrenes have a DA
> of 0.01%. Finally, the meter measures the stray capacitance only as 4pf
> to an accuracy of 0.01pf, and when shorted, 0.00pf to 0.01pf so it seems