Hi-Z High Voltage Probe

[Odysseus]

I recently built a HV DC supply and I wanted to measure it's output to a relatively high degree of accuracy without loading it down appreciably.  At worst it could be producing 10-20KV, but I might want to measure more than that in the future.  If I went the resistor divider route, and I only wanted to dissipate 0.1W in the probe at 10KV, I would need a 1GOhm impedance.  While perfectly doable, it's either expensive or requires a lot of soldering and volume.

Instead, I'd like to make the measurement capacitively. The first idea I had was to buy or build something like this: http://en.wikipedia.org/wiki/Electrometer  While this works great for measuring charge, it will not work all that well for voltage, for reasons described in this paper: www.eng.auburn.edu/~deanron/PPA.pdf

So then I figured that instead of letting the capacitor plates move, I'd hold them fixed and measure the electrostatic force instead. I haven't figured out a clean solution to measure such a small force cheaply, though.  My best idea is to use a voice coil in combination with a short-range distance sensor (ultrasonic, LVDT,etc) in a feedback loop to maintain the distance between the plates.  The measured voltage would then be represented by the required current in the voice coil to so.

But that is still pretty convoluted, so I think I'm just going to make an opamp charge integrator.  As long as the opamp and capacitor leakage is low enough, it will be practically DC coupled for the timescales I'm interested in making the measurement over (less than a minute).

Any other ideas, or improvements to mine?

EDIT: Am I just dense, or is it not possible to post the symbol for Ohm? This is an electronics forum, after all.  All that results is:  ?  I'm copying the symbol from Wikipedia.

[w2aew]

As you well know, these voltage levels can be dangerous. Probe design requires special construction and materials to be safe.   You'd be much better off with something like this...
http://www.fluke.com/fluke/usen/accessories/high-voltage-probes/80k-40.htm?PID=55357
There are other similar options available.  They were once quite common for measuring the 2nd anode voltage on the picture tube of TVs.

[c4757p]

EDIT: Am I just dense, or is it not possible to post the symbol for Ohm? This is an electronics forum, after all.  All that results is:  ?  I'm copying the symbol from Wikipedia.

is about the best you'll get.

[edpalmer42]

Why not a commercial electrostatic voltmeter?

http://en.wikipedia.org/wiki/Electrostatic_voltmeter

ebay has lots of them for sale.

Ed

[ElectroIrradiator]

The (inexpensive) resistive voltage divider, 1 GOhm from several series connected resistors, likely wouldn't be even remotely accurate. At least not without special precautions. It would be quite difficult to reach useful accuracy in many cases, due to corona discharge caused by the very high voltages.

Professional HT dividers are either dipped/encased in oil or a similar insulating compound, or they use individual, high voltage capable precision resistors. Either of which makes them expensive, as you have discovered.

[Odysseus]

As you well know, these voltage levels can be dangerous. Probe design requires special construction and materials to be safe.   You'd be much better off with something like this...
http://www.fluke.com/fluke/usen/accessories/high-voltage-probes/80k-40.htm?PID=55357
There are other similar options available.  They were once quite common for measuring the 2nd anode voltage on the picture tube of TVs.

You're quite right about the dangers involved, but for what I'm interested in, my supply and any sources I intend to measure in the future will be severely limited in the amount of available output current and discharge energy.  Just to put everyone's mind at ease, the short circuit current of my supply is 7mA and has about 200pF of capacitance at it's output.  Worst case that's 40mJ of discharge energy at 20KV. Quoted on Wikipedia, IEC 60065 states that consumer products cannot discharge more than 350 mJ ( ) into a person. Under what conditions I have no idea, however.

That probe is exactly what I need.  1GOhm, 1V per KV, and 40KV rating.  For what I'm doing, though I don't think the cost is justified, and then I would't have any fun making it

http://en.wikipedia.org/wiki/Electrostatic_voltmeter

See, I knew something like that existed.  That's what I was trying to think of after I gave up on the resistor idea.  Thanks!

Also, I'm wondering how to construct a high voltage capacitor to use in a charge integrator.  I think leakage is going to be incredibly difficult to avoid.  I guess something encased in a special potting compound or oil and evacuated in the vacuum chamber we have at work would suffice. 

[branadic]

EDIT: Am I just dense, or is it not possible to post the symbol for Ohm? This is an electronics forum, after all.  All that results is:  ?  I'm copying the symbol from Wikipedia.

The correct copied symbol looks good in the preview but returns into a ? in the post, I agree, it's inacceptable in an electronic forum.

[Bored@Work]

EDIT: Am I just dense, or is it not possible to post the symbol for Ohm? This is an electronics forum, after all.  All that results is:  ?  I'm copying the symbol from Wikipedia.

The forum software is broken in that regard (and in few others  ). I think the latest theory is that the database backend uses another char encoding then the web frontend.

What about this: 10M? ? This looked pretty well in the preview and is no compromise like using headphones.

 

[ElectroIrradiator]

Also, I'm wondering how to construct a high voltage capacitor to use in a charge integrator.  I think leakage is going to be incredibly difficult to avoid.  I guess something encased in a special potting compound or oil and evacuated in the vacuum chamber we have at work would suffice. 

Depending on capacity required, then you may be able to make one from standard FR4 double side PCB material. You would still need to remove around 20mm of copper at the edges, in addition to immersion in oil or whichever. However the flash over voltage *through* the PCB core is some rather ridiculous value. You will get about 2pF, or a bit more, per cm^2 of overlapping copper area.

[Lightages]

an example

http://www.ebay.com/itm/BK-Precision-PR-28A-40kV-HIGH-VOLTAGE-DMM-PROBE-/330951715759?pt=LH_DefaultDomain_0&hash=item4d0e426baf

[alm]

Note that high-value resistors will often have a voltage coefficient: the resistance will change as a function of the voltage across the resistor. Not sure if this is constant enough between resistors that the voltage divider remains accurate.

[Marco]

The death of the CRT has created a glut of high voltage probes ... unless you're dead set on making something yourself just get one like the one Lightages linked.

[Odysseus]

I think I'm going to buy one of these multimeter probes so that I have some sort of reference to compare anything I make myself against.  That used BK Precision probe is looking awfully tempting.

Using copper clad board to make the structure of a capacitor seems reasonable.  Is it worth the trouble to shield the inverting input of the opamp from stray external fields and leakages?  If anything other than my HV capacitor couples to the input, then it will throw off my measurement.  Just for clarity, the attached circuit is what I'm thinking about.

[Marco]

Wait, let me get this straight. You suddenly apply the high voltage, the opamp output goes to -Vi/1001 and there's no integration since there is no continuous current? What's the point of using an integrator at all?

[Odysseus]

Wait, let me get this straight. You suddenly apply the high voltage, the opamp output goes to -Vi/1001 and there's no integration since there is no continuous current? What's the point of using an integrator at all?

Yep, except it'd be a -1/1000 gain, but I doubt the absolute accuracy of the ratio between the capacitors will make a difference.

Are you wondering why I don't just use a capacitor divider configuration and an opamp as a buffer?  I don't guess there is a good reason not to.   I could just as easily do this, instead.  Still need a way to adjust it if I want an exact ratio at the output, either a trimcap or another opamp stage with a trimpot.

[ElectroIrradiator]

Your first suggestion for the electrometer schematic would probably have killed the opamp in very short order.

When you connect the HT probe, the 100 pF cap has to be fully discharged, so you get full accuracy of the meter. But this means the cap needs to be charged. It will do this by drawing current from the opamp/integrator, of course. However, it is being charged from a 20 KV supply, which means that the initial charging current will be equivalent to the 20 KV short circuit current! The only limit to the height of the - very brief- current spike will be the resistance and inductance of the probe wires. The anti-static protection diodes in the opamp will probably have to work overtime for a few moments there...

The second schematic is better, but still problematic, as the charging current spike remains. If the inductance of the circuit layout isn't just so, then there is still a risk the probe cap will draw its charge through the opamp, instead of charging the electrometer divider. The current spike will be very fast...

If you happen to have a high value HT capable resistor, exact value or wattage is irrelevant, then wire it in series with the 100 pF cap, in between the probe tip and the cap. It needs to be able to withstand the full 20KV for a brief moment, while the HT cap charges. It will also help when you are done with a measurement, and wish to discharge the HT cap.

If you build the input circuit in free-hand wiring, dead bug style or something similar, then I wouldn't bother with any form of electrostatic screening of the non-inverting input. Air is a plenty good insulator at around 20V or so, which you will reach at that point. I'd only bother with guard tracks, if you mount the opamp on a PCB.

[Odysseus]

I hadn't even considered that.  I would have figured it out quickly had I actually tried it, though!

And you aren't kidding about the speed of those discharges. My $50 multimeter goes off in the weeds whenever I start making even tiny sparks.  No EMI shielding to speak of.  It didn't like my Tesla coil very much, either. 

It should be possible to make a four wire connection to the divider cap so that any inductive spikes across the op-amp are minimized.  I don't have any HV resistors for the electrometer, but I could put resistors in each stage of my CW multiplier supply.  Doesn't really solve the problem though.

I'm not worried as much about surface leakage as I am about additional capacitive coupling of the "earthed" plate of the 100pF cap.  Well that and just the leakage of the 100pF cap itself.  It needs to be below 100pA if I want a only a 1% error after measuring 10KV for 100 seconds.  The leakage of my first crude attempt at a HV cap was near 50nA.