Input protection for high impedance voltmeter

[Signal32]

Hello,

I'm curious how exactly meters that have 10GOhm input impedance manage to obtain such high impedance while also providing input protection over a high input voltage range.
The impedance itself without any type of input protection would not be hard to obtain, just need to use a JFET op-amp as a buffer.
What I'm curious is how to add input protection so that the input impedance would remain constant no matter the input voltage.
Practical question:
How would a circuit look like that:
- Takes in an input voltage in the range 0-200V
- Maintains a high input impedance through the 0-200V range (100MOhm+, preferably 10GOhm)
- Feeds the input voltage accurately to an ADC with input impedance of 100k
- If the input voltage is >5.1V, it will only feed 5.1V to the ADC (anything 5.1v+ will be truncated to 5.1v)

I would appreciate if someone can post concepts / schematics  / explanations of how such a circuit would look like.

Thanks!

[tggzzz]

Many manuals are available at http://bama.edebris.com/manuals/

[Gyro]

I can only tell you how Datron meters do it:

- A chain of four 22k 3W resistors in series to make 88k 12W total, these are mounted on ptfe standoffs for low leakage,
- followed by a Zener diode in series with a JFET to each of the input amplifier's bootstrapped supply rails. The JFETs maintain low leakage (the input signal doesn't normally 'see' the Zeners).
- Two more Zeners to clamp the bootstrapped supply rails to input common.

This arrangement can survive 1kV indefinitely, even on the 10mV range. The meters maintain 10Gohms, input current <50pA, up to +/- 20V (10Mohms resistive divider above that).

You ought to be able to find a 1061 or 1065 service manual on the web.

[Signal32]

Many manuals are available at http://bama.edebris.com/manuals/

I tried looking at the 34401A schematic, but it's more than just an ADC with input protection. It's not very clear which components have to do with input selection / switching between 10M/10G input impedance, self test, function switching etc. I'm trying to find a simple version to understand at first. Perhaps someone can suggest a specific device where the input protection circuit is obvious / isolated in the schematic ?

I can only tell you how Datron meters do it:

- A chain of four 22k 3W resistors in series to make 88k 12W total, these are mounted on ptfe standoffs for low leakage,
- followed by a Zener diode in series with a JFET to each of the input amplifier's bootstrapped supply rails. The JFETs maintain low leakage (the input signal doesn't normally 'see' the Zeners).
- Two more Zeners to clamp the bootstrapped supply rails to input common.

This arrangement can survive 1kV indefinitely, even on the 10mV range. The meters maintain 10Gohms, input current <50pA, up to +/- 20V (10Mohms resistive divider above that).

You ought to be able to find a 1061 or 1065 service manual on the web.

Thanks, I didn't know that the 10G was not maintained up to the maximum voltage that could be applied, is this how all meters work, or is it just those Datron models do ? I'll have a look at the service manual.

[Gyro]

Thanks, I didn't know that the 10G was not maintained up to the maximum voltage that could be applied, is this how all meters work, or is it just those Datron models do ? I'll have a look at the service manual.

I think you'll find that it's probably like that for all of them. They usually maintain the high input impedance by bootstrapping the amplifier supply rails to follow the signal. There are several that only maintain 10G up to 2V (probably not bootstrapped). At some point you have to resort to a resistive divider just because of the difficulty in handling the voltages involved.

[Signal32]

I've tested the 34401A and it only maintains >10G impedance up to 10V input, then if falls down to 10M.
New challenge though: According to the datasheet the Keithley 6514 has an input resistance of >200TOhm(>200,000,000MOhm) for voltage measurements up to 200V  .
Unfortunately no schematic is available. How are they doing that (Excluding high impedance signal isolation techniques such as Teflon mounts)?

[Kleinstein]

Usually the > 1 G, high impedance only applies to relatively low voltages (e.g. up to 3 to 20 V) and the lower ranges (e.g. up to 10 V / 15 V). The higher voltages usually use a divider. Only very few meters keep the high impedance to more than 20 V. AFAIK some Keithley electrometer meters do this up to the 100 V range, but using an input stage with a >100 V supply. So the higher supply allows bootstraping to a higher voltage. The diodes used for protection only see the offset voltage of the amplifier (mV range and less).

If the voltage is higher then the about 15 V limit, impedance drops to the 100 K range and a current up to the mA range can flow.
This current limiting resistor can be just a resistor, a PTC and / or FETs that do current limiting.

[Cerebus]

I've tested the 34401A and it only maintains >10G impedance up to 10V input, then if falls down to 10M.
New challenge though: According to the datasheet the Keithley 6514 has an input resistance of >200TOhm(>200,000,000MOhm) for voltage measurements up to 200V  .
Unfortunately no schematic is available. How are they doing that (Excluding high impedance signal isolation techniques such as Teflon mounts)?

They bootstrap the power supply to the input amplifier. They take a, say, +/- 200V power supply and feed it to a +/- 5V power supply regulator via a couple of constant current sources. The +/- 5V supply powers the input amplifier. The common (i.e. 0V) rail of that power supply is driven by the output of the input amplifier so as the output of that goes up or down the 5V power supply rails follow it. Thus the input amplifier hunts up and down the range of the 200V supply so that it appears to see only 0V on its input and output. Look at older Keithley or Datron service manuals (i.e. the ones with full schematics) for concrete examples.

[Signal32]

They bootstrap the power supply to the input amplifier. They take a, say, +/- 200V power supply and feed it to a +/- 5V power supply regulator via a couple of constant current sources. The +/- 5V supply powers the input amplifier. The common (i.e. 0V) rail of that power supply is driven by the output of the input amplifier so as the output of that goes up or down the 5V power supply rails follow it. Thus the input amplifier hunts up and down the range of the 200V supply so that it appears to see only 0V on its input and output. Look at older Keithley or Datron service manuals (i.e. the ones with full schematics) for concrete examples.

Thanks for the explanation. I did understand the basics of what's happening in the very high impedance meters.
However I still don't get how protection + high impedance is provided in case of sudden jumps (input is at 0v, then suddenly it jumps to 200V). Eventually the input amplifier will raise it's power supplies to 200v, but in the meantime, what is providing the protection ? I'm assuming that at least 10M input impedance is maintained during the ramp-up period. I guess I will have to look in detail at the schematics.

Or if someone could give me a schematic of 10M input impedance + protection for an ADC ( without bootstrapped power supply ), I would very much appreciate it. Or is 10M input impedance only possible with bootstrapped power supply ?

[Morgoroth]

pH sensors have very high impedance (up to 10GOhm if I remember well), and to obtain the measurement you need a very low bias opamp as a buffer. The problem with this designs are that the currents are so low, that diodes leakage current become a big problem, so probably at low voltages there is no protection except for opamp internal protection.

take a look at INA116 and LT6268 for very low bias currents opamps, and for schematics I suggest to download this reference design CN0326.

Now, if you gonna measure that level of currents, you NEED an isolated and low noise power supply, with isolated digital and analog power paths filtered the best you could, and very special considerations for PCB manufacturing, like no solder mask, no mid layers, ring guard, etc etc...

In a second though, if the power supply is isolated, the opamp become quite immune to common mode problems, so probably the internal protections are enough. Not 100% secure, but I could bet on that.

Probably all this is not the solution you need, but could be a clue for another.

JP

[Gyro]

Thanks for the explanation. I did understand the basics of what's happening in the very high impedance meters.
However I still don't get how protection + high impedance is provided in case of sudden jumps (input is at 0v, then suddenly it jumps to 200V). Eventually the input amplifier will raise it's power supplies to 200v, but in the meantime, what is providing the protection ? I'm assuming that at least 10M input impedance is maintained during the ramp-up period. I guess I will have to look in detail at the schematics.

Or if someone could give me a schematic of 10M input impedance + protection for an ADC ( without bootstrapped power supply ), I would very much appreciate it. Or is 10M input impedance only possible with bootstrapped power supply ?

Yes, you are correct, sudden jumps don't see the same input impedance. As you say, the bootstrap supply takes a finite time to track. At that point you are down to the input series resistance of the input circuit (88k in the Datron case) and the protection components. There will always be parasitic capacitances too forming low pass filters. The >10G ohm input resistance applies to reasonably steady state measurement, there is the question of when a sudden large step voltage becomes an input protection event. Note that the input resistance is >10G, not 10G ohm, ie. there is no 10G ohm resistor defining the input resistance - the input resistance is made up leakage, input bias currents etc, so the spec will be the 'guaranteed' minimum, not a fixed value.

In your second paragraph you seem to have switched from 10G to 10M, from what you've said before I assume this is a typo. If you really want a simple 10G ohm input without going to the trouble of bootstrapped and isolated supplies, then you could try just using a 10G ohm input voltage divider. Use a high quality, high voltage 10G ohm series resistor (yes, you can get them) together with a 10.01M shunt resistor. This would give you 1mV per Volt output with a 10G Ohm input resistance. Obviously there are drawbacks with this scheme...

- It is likely to be a bit noisy, both from resistor noise and pickup. Careful assembly and screening would help.

- Temperature stability and tracking of such high value resistors will probably not be good, limiting your accuracy.

- Stray capacitances would make this a DC only measurement solution, unless you add a parallel capacitive divider.

- You would need to buffer the output with a low drift, Very low Ib, low Vos opamp, but that is doable.

- You then need to scale the voltage to your ADC range, not too bad for 100mV at 100V, doesn't need very much gain but 1mV at 1V (or lower) will be more troublesome.

As I say, if you want the 'simplest' way of getting your 10G Ohm input resistance, this will do it. On the upside it would need little or no input protection, dependent only on the maximum voltage rating of the 10G Ohm resistor.

[David Hess]

Others have already covered how it is done so I will skip further explanation except to say that an added input series resistors limits destructive current while the bootstrapping catches up.

Offhand I do not know of any voltmeters which bothered to bootstrap their input circuits but this is a common feature of electrometers which support much lower input currents and high input voltage ranges at high input resistance.  Voltmeters input circuits tend to top out from 2 to 10 volts although I think I remember some that were good to 20 volts which is handy for development of +/-15 volt precision designs.  I do not know if these later voltmeters used bootstrapping or not.

Also take a look at how the overload protection for oscilloscope inputs works.  Some of them also bootstrap the input amplifier and protection circuits for greater input voltage range.  This is common with dedicated differential inputs to increase their input common mode range.

Most operational amplifiers bootstrap their input differential pair to increase common mode range and common mode rejection ratio but low input bias current devices may also bootstrap their input protection networks.  Super beta input transistors only have a breakdown voltages of around 5 volts (or lower!) so bootstrapping is mandatory to make them usable.  An interesting exception to this is PNP input operational amplifiers; their crummy low gain PNP transistors have high emitter reverse breakdown voltage so the ubiquitous 324/358 design has no input bootstrapping and requires no differential input protection.

[mirbagheri1122]

high I couldn't understand any thing from datron 1061 manual for making gigaohm input.could you send me schematic?

[Gyro]

You can find the Datron 1061 Service manual with schematics on the KO4BB Manuals site here...

http://www.ko4bb.com/getsimple/index.php?id=manuals&dir=06_Misc_Test_Equipment/Datron

[tggzzz]

Have a look at TAoE3, Fig5-80 p361 and text explanation. It is a differential input amplifier protected to +-250V.

[mirbagheri1122]

hi gyro
thanks for your very useful schematics on datron.but I saw in manual discrete buffer(instead of using a zero chop opamp) for voltages under 20v.this hardens calibration very much that I think I should do short circuited for estimating offset.and refrenceing input for estimating gain.(due to changing in resistors and transistors variations in relation with temperature.
could you guide me?

[Gyro]

You're welcome. Yes, the input stage is bipolar rather than FET / chopper. The designs date back to the '70s and IC choppers simply weren't available. Datron used bipolar rather than FET because bipolar has more stable Ib across the operating temperature range. In addition Ib and Vos are much easier to temperature compensate, so they achieved a much lower (typ <20pA) input current over the instrument operating range than the could have with FET input.

I had forgotten that I previously posted input stage schematics for the 1041/1051 previous generation meters  . These did not have microprocessors so they had no s/w assistance with zeroing etc. Even so they were capable of stable 100nV resolution. As you will see, the actual circuit is virtually identical to the 1061 (minus the added uC control complexity). Part way through production of the 1041/51 they switched from using a bipolar LM312 opamp as the input stage to the dual matched transistor + opamp that you see in the 1061. The result was a little more stable and easier to trim (less need for extended opamp burn-in and selection) but at added complexity. Both use the same bootstrap supplies as the 1061. This was very much an evolution of the same topology over the years.

Anyway, here is (are) the post(s), hopefully the schematics are easier to follow without the added complexity of the one in the 1061 manual...

https://www.eevblog.com/forum/projects/very-low-bias-current-op-amp-to-buffer-a-kelvin-varley-divider/msg694593/#msg694593

In terms of input Vos and Ib, yes, these are adjusted with input shorted, and 10M resistor in parallel with a low loss 100n capacitor respectively. The 10M resistor reads 10uV / pA of Ib.

I hope this helps.

[Marco]

You're welcome. Yes, the input stage is bipolar rather than FET / chopper. The designs date back to the '70s and IC choppers simply weren't available. Datron used bipolar rather than FET because bipolar has more stable Ib across the operating temperature range.

Doubt it was more stable than the current through a micron range thick layer of silicon oxide (ie. MOSFET). Hard to beat defacto 0.

Vgsth stability and tracking is a different matter.

[Gyro]

Good point, but back then they would have been comparing against JFET rather than MOSFET opamps... To quote from the LM312 datasheet: "with very low offset-voltage and input-current errors - at least a factor of 10 better than FET amplifiers over a -55'C to +125'C temperature range".

Low offset MOSFET amps would have been a mere twinkle in their creators eyes back then, time has moved on. I've never looked at what the 1071 used, I'm curious now.


Edit: From what I can see, the 1071 is similar to the 1061 (as above). The 1081 used a matched JFET pair followed by an LM11, with a 7650 chopper handling DC offset. Ib still still handled by a compensation network. The 1281 looks to use a rather complicated discrete chopper amp rather than a monolithic IC.

[David Hess]

You're welcome. Yes, the input stage is bipolar rather than FET / chopper. The designs date back to the '70s and IC choppers simply weren't available. Datron used bipolar rather than FET because bipolar has more stable Ib across the operating temperature range.

Doubt it was more stable than the current through a micron range thick layer of silicon oxide (ie. MOSFET). Hard to beat defacto 0.

Sure it would.  MOSFET and JFET input current is non-zero and doubles every 10C while a super beta bipolar input stage with bias current cancellation like the old LM11 can be pretty stable in the 10s of picoamps range with a linear change of current with temperature.  Typical MOSFET and JFET input stages back then were not much better than 10 picoamps.

I used to use the LM11 for its low input current at high operating temperatures which only the best and selected MOSFET inputs could match.

[Marco]

Only the gate protection, the MOSFET lets through attoamps. Of course integrated MOSFETs need gate protection and it seems to have taken a couple of decades for manufacturers to put bootstrapped protection on MOS opamps.

[JS]

  I guess you are getting this wrong, very high impedances (and low noise) are archived with bootstrapping, so you have a follower amplifier giving you a low impedance version of the same voltage as the input. You could strap a pair of reversed diodes in parallel from the input to the output if this amp so if the amp saturates you shorted the input. If the amp is operating in the linear zone only a few µV across the diodes will mean no current at all. Then if the amp is saturated, it can't take a lot of juice, so you need something in series to limit the current. There a PTC + suitable resistor would do the job, maybe some inductance in case very fast spikes could be a problem.

  Also the output of this amp should be limited within it's supplies, so a few diodes here will help to deal with that and provide a second path (other than the opamp output stage which is usually pretty wimpy) for the current through the first pair of diodes to go through. For high voltage inputs a spark gap could also be used, probably in combination with the other method.

  In any case, when designing an input protection you need to know how much over range it can handle for how long. You are trying to get a 200V stage with ~1G input impedance, not a big deal for a bootstrapping amp (check the metrology section for exactly this discussion) Then how much else you want to be able to handle indefinitely  without releasing smoke? 300V? 1kV? 10kV? Then you need to dimension the protection for that, or dimension the protection for that for a short period of time and then use the old school relay you need to reset after you go out of range.

  When crunching out the numbers you need to pay attention to the max admissible level, let's say 1kV, to your max input voltage, 200V in this case. That is 800V, a random vishay PTC which could take up to 10mA 1kV, so your clamping must handle that for any time you want, not hard for the diodes clamping the amp but you would need a zenner at the output of the amp, just over 200V and over 2W, the same zenner will deal with the other end as far as the amp is ok going under ground, but it should as it should take grounded inputs without any issues, a volt or two under ground shouldn't be an issue. The 2k resistance of the PTC shouldn't be a problem for the G? input stage, 6 orders higher and shouldn't change more than a few %, only a problem in an 8.5 digit meter, a consideration on a 7.5 one.

  Then, there is one final consideration is the recovery from the over range, could take some time to be back into spec, the PTC to cool down from 10M? to 2k? in the case of a huge over range or at least the diodes and the saturated opamp in the case of a smaller one.

JS