Capacitive displacement measurement

Make the capacitor part of the tank of an oscillator, then measure the resulting frequency - much simpler ! (If 3pF gives you 300 MHz, 0.3fF will give you a frequency delta of 30 kHz)

Of course, making everything stable enough for either approach will be an interesting challenge - are you sure you don't want a mechanical stop !

Dave

Haude Daniel a écrit :

That'll work, but AD had recently released a series of 24bits capacitance to digital converters for this: AD7745/46/47 and AD7142.

Is TSSOP16 small enough?

I would do it the way they do it in that other high-precision positioning realm, CD-players. Maybe you can even gut one to get the necessary positioning and feedback elements. Then it's just a matter of counting pulses instead of trying to resolve analog signals to one part in 10,000.

-- Joe Legris

In message , dated Mon, 4 Sep 2006, Haude Daniel writes

Moving things to places with precision is an application for a micrometer head. All the precision work is done for you, and all you have to do is turn the knob. OK, turning the knob while the head is in vacuo is not quite so easy, but a magnet on the knob and one outside the vacuum will do it.

If I understand this right, your moving peg has two longitudinally-disposed segments and you are sensing a null when the assembly is longitudinally centered in the outer cylinder and the signals are equal.

That seems like a good approach to me because the rate of change of signal is greatest near the sense point. It also should largely cancel thermal expansion because it is a differential measurement seeking a null. It also does not depend on 1 micron stability of the larger structure because all measurement and reference is at the measurement site -- unlike a laser interferometry scheme.

Well, this thing has to work at the end of about 2m cables w/ associated capacitance. No electronics allowed near measurement point.

robert

Not if the thing in question is at the bottom of a cryostat, 2m away, no mechanical connection, in an 8T magnet at 0.5K.

robert

See my other replies about this.

That's not much with sync demux.

robert

Great, but I don't really have a microcontroller around. I'd prefer a simple analog solution.

Usually, yes.

robert

The TC of the inductor will be a big problem. He needs 100 ppm stability, and a typical inductor will run 150 ppm/degree C.

John

I've done signal conditioning for LVDT's in liquid helium at 2K. They are very dimensionally stable at that temp! And they give a lot more signal power than capacitive sensors.

This was developed for the CEBAF/Jefferson Labs electron accelerator, for feedback sensing of the niobium cavity mechanical tuners...

The center (used as primary) LVDT winding is excited with a reasonably stable sine wave, and the two secondary signals are digitized at the sine peaks, and then the position is computed ratiometrically as (A-B)/(A+B) and lowpass filtered. It's ppm stable, and sub-micron resolution and stability should be easy.

John

Do you remember the capacitance-measuring threads I hosted here a few years back? The resulting measurement capability was several orders of magnitude better than you need, so it's certainly possible. I used a three-terminal measurement, which allowed up to 10 meters of coax cables to and from the electronics. The idea was to make the destination a bridge- nulling-point summing junction and to grab all the ac current (i.e. nulling-error current) from the measured capacitance, into the summing amplifier's feedback, which is usually the capacitance around the first stage's dc feedback resistor.

. 5-20pF etc ............. . coax sensor-C coax : shielded box . ,---------x----o--||--o-----x------, : . | : | transresistance . | NULL : SJ- opamp --, . | coarse: ac summer : | : | . +--> ratio ----> G= -1.00 --x--||--' : | . | trans ,-> G= -0.05 : ref-C : 2 stages . | | servo range : 30pF : 10kHz amps . +--> AD734 --' :............ G = 500 . | MPY | . | | | error AD734 | sensitive . | | '--> OUT . | +/-5V | x1 . '--+------>---- integrator ---->----' x10 . | -90 deg phase x100 . 10kHz, 10Vac

I used a ratio transformer for 0.5ppm low-drift performance; an early version used a 10x range switch plus a 10-turn pot. The capacitance gauge measured STM tip displacements in a cyrogenic vacuum chamber.

Well, are you 'robert' (of questionably happy memory) or Daniel Haude? Never mind, it doesn't matter.

You are just making it difficult. If I give you a solution for that, you will then explain that the 0.5 K is because the equipment is on the dark side of Mercury. (;-)

You can MAKE a mechanical connection. Tachometer cable, for example. The

Commercial capacitive sensors are always differential, because single-ended ones are so sensitive to minor errors of centring. I suggest a capacitive divider, i.e. use two cylindrical elements on the ends, driven 180 degrees out of phase with a centre-tapped transformer, and another cylindrical element in the middle. As the dielectric rod moves, you'll produce a voltage between the middle element and the centre tap of the transformer that you can measure very accurately, and that will be zero at the middle of the motion. Then you use the lock-in approach you've discussed. To leading order, decentring and angular misalignment don't cause errors in this scheme, as long as you're measuring the voltage and not the current--i.e. using a noninverting amplifier with a high input impedance. TIAs are not the right choice here, because misalignment will change the capacitances and hence the capacitive current, causing a scale error. Use a high-value bleed resistor between the centre element and the centre tap to set the bias, and a moderate-resistance feedback network to keep the additional voltage noise low.

If you want lower noise, you can replace the bleed resistor with a MOSFET switch to keep the op amp in proper bias. You just hit the switch whenever the op amp output gets too far from 0. (Put the RC highpass on the output of the amplifier instead.)

If you want to get fancy about it, you can get high linearity as well, using interdigitated segments arranged like barrel staves.

Cheers,

Phil Hobbs

Check out the Thompson-Lampard calculable cross-capacitor, as used at all the better national standards laboratories.

Cheaper implementations of the same idea have been proposed and tested

They all depend on precision ratio transformers to make the AC bridge that reads the capacitance to a useful precision.

See if you can find a copy of "Coaxial AC Bridges" by B P Kibble and G H Rayner ISBN 0-85274-389-0. Both authors worked at the British National Physical Labortories at Teddington, and you can still buys copies of the book from there

You can also find the book in university libraries. Amazon is a dead loss.

I have had my own copy for years now, thanks to your suggestion. Reasonable-good custom ratio transformers aren't so hard, once you understand them (e.g., think of multiply-braided windings), and their sub-ppm signal-handling accuracy certainly is useful.

Interesting! Good stuff!

A good option, but I can't have magnetic materials down there.

--Daniel

No. It really is only 2m away from civilization.

Yeah...right. I can make a lot of stuff if I have to. In this case I don't.

--Daniel