Circuit design of branded electrometers

[Kirill V.]

Hi friends,
I want to study the question of building a voltmeter with a high input resistance. You will help me very much in this if you name the models of industrial voltmeters that have manuals with diagrams in the public domain. You just know a lot more about Western measuring equipment than I do
The main requirement - the device must be from the transistor era, even better if it is built on operational amplifiers. This is much easier to analyze than dealing with vacuum tubes circuitry.
Well, I'll tell you what I'm most interested in. The biggest problem is to provide a path for the operational amplifier's bias current to drain. I want to know how this is implemented in industrial voltmeters.
Thanks.

[tkamiya]

There is a very nice application note here.  While this is not about a production model, it meets your needs almost exactly.

http://www.ti.com/lit/an/snoa620/snoa620.pdf

[Kirill V.]

Thanks, tkamiya!
Let's look at a simple buffer

This is so primitive that it does not need comments, and if I wanted to make my voltmeter exactly like this, I would not ask for help on the forum
Previously, dear David Hess recommended me app note 242, which describes a trick to increase the equivalent input resistance:

We disconnect the lower end of the leakage resistor from the ground and feed it an auxiliary potential, in-phase and proportional to the input voltage. For example, if the auxiliary voltage is half the input, the equivalent input resistance is doubled with the leakage resistor unchanged.
Is there a brand voltmeter that uses this trick?

[Conrad Hoffman]

Download the manual for the Keithley 610 or similar model. They use a floating scheme that's very effective and it will also measure charge.

[MiDi]

Take a look at the Keithley 617 electrometer.
Design is from mid 80s, couple of things are a bit outdated, but principle remains.
The manual includes circuit description and schematics.
Valuable information and some improvements to the Frontend are described in the K617 Thread.

[Kirill V.]

Ok, thanks
Oh, the K617 circuit is what's causing me a headache. I've looked at it several times, trying to understand I much prefer the old HP manuals, everything is absolutely clear, clear flowcharts, detailed notes on the diagram, and so on.
Yes, floating topology and bootstrapped supply are classics, I know that. There are no alternative solutions?

[MadTux]

I find Keithley 617 input stage rather easy to understand, schematics are great IMO, too.
Basically isolated gate FET pair as differential amplifier for input, that keeps itself at input potential by driving +/- 210V transistor amplifier.

Perhaps find a broken K617 on ebay, fixing it usually helps understanding it 

Keithley electrometers are rather simple in principle, if you want something more hardcore, try understanding HP 3456A AC converter for example 

[Kirill V.]

Yes, I understand the operation of the device in General terms. But I'm trying to trace the path of the signals through various sheets on the diagram and alphanumeric links. I want to understand the nuances
I think I'll try it again

[Kirill V.]

Yes, apparently, you are right, it is easier than it seems. A 250 GOM resistor is used to drain the bias current.
It remains to understand what COM and Signal Ground are and why there is a 100 Ohm resistor between them. And perhaps a few more nuances.
K617 input buffer manages its own power supply and guarding. But sometimes a separate buffer is used for supply and guarding management, and input signal buffering is assigned to a separate operational amplifier. What is the difference between such topologies?

[Stray Electron]

Hi friends,
I want to study the question of building a voltmeter with a high input resistance. You will help me very much in this if you name the models of industrial voltmeters that have manuals with diagrams in the public domain. You just know a lot more about Western measuring equipment than I do
The main requirement - the device must be from the transistor era, even better if it is built on operational amplifiers. This is much easier to analyze than dealing with vacuum tubes circuitry.
Well, I'll tell you what I'm most interested in. The biggest problem is to provide a path for the operational amplifier's bias current to drain. I want to know how this is implemented in industrial voltmeters.
Thanks.

   Go look for that subject in the archives of the Hewlett Packard Journal and the Hewlett Packard Application Notes. I can't say what they have regarding electrometers but they most likely covered them.  The other good thing about the HP stuff is that there is often a FULL service manual available so you can study every inch of the schematics and get exact descriptions of the parts and usually even who supplied them to HP.  The old General Radio (aka GenRad aka GR) articles are also very good but harder to find.

[Kirill V.]

Okay, I'll look at these sources
I will describe my own idea in General terms.

I have a lovely large laboratory voltmeter with a very low input resistance ( starts at 100 Ohms at the 7.5 Volt range). At first I wanted to make a simple external buffer prefix, but then I wanted more, and to achieve as high an input resistance as possible with my home capabilities and zero experience
Now I have started studying the K617 circuit again, and suddenly I see that this is very similar to my idea. In General, it is very simple. But if the voltmeter input floats... the ideal voltmeter should show zero in this case. In order for this to be the case in reality, it is necessary to provide a path for the bias current, while the voltage drop on the leakage resistor should not be shown by the voltmeter. That's my problem

[MiDi]

For example if there is 1fA input bias and 1Gohm input resistance, this contributes only 1uV offset voltage.
So it depends on your requirements for max voltage error and min input resistance when "floating".

[Kirill V.]

So, in 2016, I bought AD549LHZ for my future projects. All this time it was lying in my box with nothing to do. Suddenly, today I wanted to test it and conduct some experiments. Especially for this purpose, I had prepared a Board with aluminum foil, an antistatic bracelet and a Bernstein ESD tweezers.
I put together a circuit with a unipolar supply and a resistive divider to form the midpoint of the supply. The non-inverting pin is bent 90 degrees and does not touch anything around it.

I also have a homemade cable Assembly made by RG-174 coaxial cable and alligator clips.
You can see the 4700 pF capacitor and the clips from the alkaline battery (gives 6.1 Volt, because used) connected to the wires insulation. The leakage current charges the capacitor and the AD549 buffers the voltage. I measured the current using the integration method.

Current about 1.4-1.5 pA, therefore, insulation resistance more 4 POhm.
 I've been want to make picoammeter with this amplifier. Suddenly, quite unexpectedly, I made it in half an hour. WOW!!!
Now I can experiment with the design of the electrometer. And picoammeter. And ohmmeter too

[Kirill V.]

For example if there is 1fA input bias and 1Gohm input resistance, this contributes only 1uV offset voltage.
So it depends on your requirements for max voltage error and min input resistance when "floating".

Yes, of course. but the K617 has an input resistance in the TOhm range, while the circuit does not have a single resistor with this rating. The floating resistor is the secret. But I don't understand why this idea doesn't lead to strong drift and negative feedback persists even with floating input

[MiDi]

The 250G resistor is only able to deliver a couple of +-10s fA, so 1% drift in this current source converts to drift of sub-fA only.
Do not understand what you mean by nfb persists - in voltage mode the frontend is just a buffer and the power supply is referenced to this.
Do not expect that the voltage stays near 0V when the input is floating...

[Kirill V.]

If the buffer operates within a constant supply voltage, then we can force the bias current to flow through the external resistor, rather than through the signal source, if the potential difference between the ends of the resistor corresponds to Ohm's law. This can be achieved using a potentiometer. And any drift related to the input will appear at the output of the buffer with a transfer coefficient of one. The buffer output does not affect the potential at the right end of the bias compensation resistor.

And if the supply voltages now follow the buffer output, how does this technique work in this case?

[MiDi]

And if the supply voltages now follow the buffer output, how does this technique work in this case?

Simply connect OP1 output to the GND of supply rails.
This GND is than bootstrapped and should not be confused with Signal GND.

[Kirill V.]

Now it is necessary that the OP1 output controls the external output stage, and the inverting input monitors the load voltage. I showed the diagram above. When nothing is connected to the input, the op-amp cannot compensate for its own drifts because the inverting input is outside op-amp power rails
I redrawn the diagram

Now the leakage resistor is located between the op amp inputs and any drifts are amplified by open loop gain
Or do I not understand something?

[MiDi]

Now it is necessary that the OP1 output controls the external output stage, and the inverting input monitors the load voltage. I showed the diagram above. When nothing is connected to the input, the op-amp cannot compensate for its own drifts because the inverting input is outside op-amp power rails
I redrawn the diagram

Now the leakage resistor is located between the op amp inputs and any drifts are amplified by open loop gain
Or do I not understand something?

There are two power supplies involved, the bootstrapped one for Frontend (+-5V referred to buffer output ~= Signal HI) and the second one for the LOAD (+-210V referred to Signal GND/LO) - the LOAD is input impedance of ADC.

Edit:
When nothing is connected to the input, the effective input bias-current will load the input impedance and create a voltage drop on it.
Lets assume the input impedance is 1T and the effective input bias-current is 1pA, then the output voltage will settle to 1V, for 1P it would theoretically settle to 1000V.
1000V is outside the range of Signal GND referred power supply (+-210V) and the output will clip at around this rails.

[Kirill V.]

Yes, separate coil of transformer with one-half rectifier and 78/337 regulators. It's like two small batteries on my circuit, right? Potentiometers of offset voltage and bias current adjustment refers to this supply.
In other words, the voltage drop on the isolation of the input circuit is perceived by the buffer as a input signal and gives to the output?
By the way, what can I make a multi-gigaohm resistor from? Two lengths of insulated wires intertwined? Or a piece of FR-4?

[Conrad Hoffman]

I'd also like to know how to DIY high value resistors. Making a bad one is easy, but making one that's stable with time, temperature and humidity is much harder. As a first go, try mixing some powdered pencil lead with epoxy or lacquer and painting it between two traces on a piece of PCB. The type (#) and quantity of pencil lead in the mix will determine the value.

[Kirill V.]

Cool, this will be a thick-film resistor. This is how they were made in hybrid integrated circuits
I have a graphite suspension with glue

[MiDi]

Yes, separate coil of transformer with one-half rectifier and 78/337 regulators. It's like two small batteries on my circuit, right? Potentiometers of offset voltage and bias current adjustment refers to this supply.
In other words, the voltage drop on the isolation of the input circuit is perceived by the buffer as a input signal and gives to the output?
By the way, what can I make a multi-gigaohm resistor from? Two lengths of insulated wires intertwined? Or a piece of FR-4?

Correct.

First I was a bit confused what you meant by drift related to the input, to be clear, this is not drift - it is interaction of input bias current with input impedance.
This impedance consists of the resistive part and the capacitive part, the capacitive part will be charged by the bias current, so the output "drifts" - or better said settles - to the final value given by the resistive part.

As you already noticed, if the output hits the high voltage rails, the NFB does not persist.
This leads to a voltage difference of the op amp inputs and finally will be clamped either by protection circuit between the inputs or to the bootstrapped power supply rails.
When the protection kicks in, the input impedance will drop dramatically.
To protect the protection circuits from to high currents there is the need for current limiting the inputs, this is simply achieved by putting a suitable resistor at each input (most op amps are rated for 10mA absolute max).

If a specific resistor with specific properties is needed, I would recommend to buy one - up to 10G are awailable for ~1EUR with not to shabby specs.

[serg-el]

Кирилл, посмотри здесь.
Схема похожа на то, что ты ищешь.

Cyril, look here.
The circuit is similar to what you are looking for.

http://bbs.1ppm.cn/topic/276/%E8%87%AA%E5%88%B6%E6%8C%87%E9%9B%B6%E4%BB%AA%E7%9A%84%E6%80%9D%E8%80%83-some-thoughts-on-diy-null-meter/16

[guenthert]

[..]
Current about 1.4-1.5 pA, therefore, insulation resistance more 4 POhm.
[..]

On a breadboard?  That is, er, quite remarkable.