high impedance voltmeter?

[FrankBuss]

In one of Dave's videos he used his Agilent meter in high impedance mode. How does the input circuit look like for it? I would like to build something similar. My idea is to use a FET OpAmp in non-inverting mode as an input buffer, with +/-15V supply. This would allow me to measure +/-15V (minus some volts, depending on the OpAmp).

But would be nice if I could measure higher voltages as well. Should I use relays to switch between the direct input and a voltage divider, or is it possible to build a high impedance meter with higher input voltages as well? I would like to have at least 1GigOhm input resistance, and most CMOS switches have higher leakage, or cost a fortune, compared to a relay.

[Marco]

Depends what you want noise and acquisition speed wise.

You can just put a 1:X divider in front and handle everything else with a VGA ... but with a 1 GOhm resistor that divider will be just a bit noisy.

If you just use that divider for the high voltage path you will always need a relay (to select between the high voltage path and the low voltage path, a MUX won't provide sufficient isolation). You can use a MUX to select between the output of the divider and other paths though, from the leakage current point of view the two arms of the divider are in parallel, together with zeroing calibration that's not a huge issue.

PS. without capacitive compensation your accuracy will drop off a cliff after only a few 10s of Hz AFAICS.

[FrankBuss]

Why do I need always a relay? My cheap multimeter doesn't need one and can measure in auto-range mode from 1 kV to millivolts. I was thinking of an input protection like this:



Maybe I'm wrong, but a 1 megohm resistor should be sufficient for protection, in combination with the diodes, even if 1 kV would be applied to the input (and maybe an additional varistor for really high voltage peaks). And then a relay for disconnecting the ADC input from the 1 megohm resistor, and connecting the resistor to a resistor divider, which goes to a mux for the different divider stages. And another relay connects the output of the mux to the ADC.

I don't plan to be super accurate, maybe a 12 bit ADC, but with high samplerate (maybe 100 kHz) to do interesting things like calculating the RMS of low frequency signals (mains) in software.

[jpb]

How about one of these:

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

if I understand it correctly it draws no current at all so has infinite impedance!

To be a bit more serious, I notice on my DMM (a 5 1/2 digit ADC) the more than 1G ohm option is only available up to the 3V range, so I guess measuring high voltages with very high input impedance accurately is not trivial.

[David Hess]

The problem involves the input common mode range of the high impedance buffer which is why typically only the lowest range or ranges have input resistances greater than the resistive divider at the input.  Getting around this limitation requires designing a high impedance buffer with a larger common mode input range which is a straightforward but not trivial task.  In addition, any such design should not compromise input overload capability.

There are a couple of ways to go about this:

1. Use a vacuum tube for the input buffer.
2. Use a higher resistance input divider.  Some old instruments did this but high value resistors have compromised precision.
3. Use transistors which support higher breakdown voltages and a higher supply voltage.
4. Use a cascode design so the input stage supply voltages follow the input.  This works well even with operational amplifiers.

[free_electron]

notice how your input is at 15 volts ? an 1n4148 is completely unsuitable. it leaks in the order of 1 to 10uA ! into a 1 gig impedance ... that is a lot of volts ! (of cousre it clips against the 15 volt supply.)

you need to use a diode with a LOT LESS leakage. typically an n channel jfet is used. gate is anode, drain-source is shorted and becomes cathode.

[David Hess]

A cascode design neatly solves the issue with leakage in the input protection diodes although you still need low leakage diodes.  If the clamp voltage follows the input voltage, then the reverse voltage across the diodes can be controlled so low leakage diodes that have low reverse voltage ratings can also be used like LEDs that are painted black or bipolar transistor base-emitter junctions.

[drtaylor]

There are two basic methods to accomplish higher voltage input ranges at very high input impedances. 1) Use a high voltage Op Amp for the input device. 2) Bootstrap a lower voltage Op Amp so it can handle far higher voltages.

Of these two approaches #1 is fairly easy to accomplish. The LTC6090 from Linear Technology has +- 70V rails. With a very low bias current of 50pA typ it could easily serve as a DMM front end. However to make it truly precision might take some work.

Bootstrapping is the tried and true approach used by almost all precision DMMs. I have designed many of these and they can be tricky. If anyone is interested I can post bootstrapped DMM front ends that Fluke used long ago.

[HighVoltage]

Bootstrapping is the tried and true approach used by almost all precision DMMs. I have designed many of these and they can be tricky. If anyone is interested I can post bootstrapped DMM front ends that Fluke used long ago.

Very interesting, I would like to see, how this was done by Fluke.

[David Hess]

Bootstrapping is the tried and true approach used by almost all precision DMMs. I have designed many of these and they can be tricky. If anyone is interested I can post bootstrapped DMM front ends that Fluke used long ago.

I used the term cascode instead of bootstrap but we were thinking of the same thing.

Some oscilloscope input amplifiers use this technique to increase their common mode range in differential applications.  I am sure modern high voltage differential probes do it as well.  The old Tektronix 7A13 and 7A22 differential amplifiers which do this also happen to support operating in a high input resistance mode where the input resistance is >1 gigohm instead of the standard 1 megohm.

[FrankBuss]

I think a lower voltage range in high impedance mode would be no problem. Usually for higher voltages like measuring mains the input impedance is not important anyway, or even better to have low impedance mode, like the 3 kohm input mode of the Fluke 114, to avoid "ghost voltages", as the Fluke manual calls it. Nevertheless, could be interesting to see such a boostrap circuit, and if there would be a simple way to avoid relays.

notice how your input is at 15 volts ? an 1n4148 is completely unsuitable. it leaks in the order of 1 to 10uA ! into a 1 gig impedance ... that is a lot of volts ! (of cousre it clips against the 15 volt supply.)

Good point. What about this OpAmp with FET input? http://www.ti.com/lit/ds/symlink/opa140.pdf On page 17 it shows some input protection diodes and says up to 10 mA is allowed. So with this OpAmp I don't need any other external protection, other than the 1 meg series resistor, to protect it up to 1 kV (using an 1 W / 1 meg resistor)?

[Marco]

Why do I need always a relay?

You don't always need it ... I said if you just use the divider for the high voltage modes you need it (but I guess if I said "if you only use" it would have been more clear). With a relay you can have low voltage measurement paths with only enough series resistance as necessary for input protection instead of the noisy X MOhm resistor you use for a high impedance divider.

The 34401A for instance has only 100KOhm series resistance for it's lower voltage ranges, it's normal input impedance in those ranges is determined by it's active input buffer and not that series resistance. As suggested above you can of course do this for higher voltage ranges as well ... but then you need a high voltage power supply in your meter.