Very low bias current Op-amp to buffer a Kelvin Varley divider

[Kleinstein]

The resistance dependent offset is just a different description of bias current. One they compensate for the bias, the offset is not effected by the input impedance.

With AZ OPs the offset can also depend on the AC impedance at the internal sampling rate. So one may need to have some capacitance at the input.

The 30 pA or so for a modern meter is not only the amplifier but also includes protection circuits, range switching and the layout. With such low bias ratings it's important to look at the temperature and maybe humidity for which this is specified.
High end meters often use a very slow auto zero cycle and a discrete JFET amplifier. Like with integrated AZ OPs its finding the right balance between low bias and low noise. The overall performance seems to be not significantly different.

Even with a noisy amplifier there may not be much need for filtering, as most meters are not that sensitive to noise. The usually long sampling time is already a very efficient low pass filter.

[Gyro]

You have to admire the designer's skills given these date to the 70's (60s?)

Absolutely. Early 70s certainly. The days when men were men and sheep were nervous   It's all too easy today. Lots of lessons still to be learned from the challenges overcome in vintage stuff.

As for the input current, yes I guess leakages start to dominate - lots of PTFE standoffs in the Datron, don't know about the Agilent, but both would have active guarding.

As for the modern AZ amps, Ib and Vos compensation probably have much less importance but supply bootstrapping would probably still give a significant improvement in CMRR and probably other parameters, as well as allowing a greater i/p swing (outside the supply limits of the AZ amp).

[TimFox]

To avoid current from common-mode voltage, the guarding referred to above is essential.  Even a PTFE standoff can have leakage resistance from humidity (although PTFE is naturally "hydrophobic"), so having a guard ring with appropriate drive at its base is a good idea. 

[Kleinstein]

Having a bootstrapped supply, bootstrapped protection circuit and driven guards can be often combined to to together in one circuit.
With guards and a few cuts, one may get around extra standoffs.

Whether you need bias compensation depends on the AZ amplifier. The very low noise versions have a bias that is to high to be tolerated - in the 100 pA range. With supply bootstrapping the bias compensation may be as little as a very high value resistor and a trimmer.

[Echo88]

Maybe someone can help me out with the understanding of the KVD-buffer mentioned here and in the AN86 Page 9: http://cds.linear.com/docs/en/application-note/an86f.pdf

I only ever thought about CMRR in difference amplifiers which measure tiny signals. In this case its mentioned together with a voltage follower; how does the CMRR manifest as an error in this circuit?

[Kleinstein]

A buffer made from an OP makes the op see the input voltage as common mode voltage at the input. So if there is a common mode error, this would would cause the amplification to be of from 1 by that small amount. A 120 dB CMRR would thus translate to a 1 ppm gain error. So like open loop gain, the CMRR is an important parameter for a precision buffer.
The usually very good CMRR of AZ OPs helps a lot.

Bootstrapping the supply can be a nice trick to improve a buffer, as the critical OP will not see the full common mode voltage, but a much smaller one (the error of the 1st approximation).

[David Hess]

Maybe someone can help me out with the understanding of the KVD-buffer mentioned here and in the AN86 Page 9: http://cds.linear.com/docs/en/application-note/an86f.pdf

I only ever thought about CMRR in difference amplifiers which measure tiny signals. In this case its mentioned together with a voltage follower; how does the CMRR manifest as an error in this circuit?

For an operational amplifier, the common mode rejection ratio represents the change in offset voltage for a change in voltage at the non-inverting input.  So a 120dB CMRR means that for every 1 volt change at the non-inverting input, the offset changes by 1 microvolt.

So CMRR is a critical parameter when dealing with non-inverting amplifiers and can limit performance in precision applications.  With inverting amplifiers, the non-inverting input is fixed so CMRR ratio is irrelevant; audio circuits often use common mode suppression to remove distortion caused by limited CMRR.  With difference amplifiers, the non-inverting input changes by some fraction of the common mode input voltage.

[Echo88]

Thanks! Since all other topologies apart from the voltage follower necessitates resistors, seems we no choice but to use the voltage follower. I also took a look again at the AoE-book and there was also the note that inverting amplifiers dont suffer from CMRR-problems. But i couldnt find a description how supply bootstrapping (like done in the Fluke 5700A Page 542 with the LTC1052 https://xdevs.com/doc/Fluke/5700a/5700A_old_sm.pdf ) affects CMRR, only that bootstrapping allows the use of low supply OPs (with its sometimes better specs) while allowing bigger voltage input/output-swings.
Is it because CM-voltage is related to the middle of the supply voltages and always follows the input signal, so that the CM-voltage is constant?

Ive read that diodes, which are illuminated to generate a minimal current, could be used to cancel the remaining bias current. Are there any existing designs which use this concept?

[Kleinstein]

The common mode voltage is relative to the supply (middle or one side - if the supply changes there would be an additional effect of the finite PSRR). So bootstrapping to a good approximation avoids a changing common mode voltage seen by the OP.

It is possible to use photocurrent (e.g. from a diode) to compensate. The downside is that photocurrent is inherently paired with shot noise. Current through a high value resistor can (with the right type of resistor) be without extra shot noise. The resistor noise can be smaller. A low leakage diode might be needed for protection anyway - however low leakage diodes are usually in a light tight case.

[David Hess]

I also took a look again at the AoE-book and there was also the note that inverting amplifiers dont suffer from CMRR-problems. But i couldnt find a description how supply bootstrapping (like done in the Fluke 5700A Page 542 with the LTC1052 https://xdevs.com/doc/Fluke/5700a/5700A_old_sm.pdf ) affects CMRR, only that bootstrapping allows the use of low supply OPs (with its sometimes better specs) while allowing bigger voltage input/output-swings.

Is it because CM-voltage is related to the middle of the supply voltages and always follows the input signal, so that the CM-voltage is constant?

Effectively it is although I do not know if it is specifically defined that way.  If the supply voltages follow the signal at the non-inverting input, then it is just as if the operational amplifier has no common mode input voltage change.

If the common mode input voltage is fixed and the positive supply voltage is changed, then the positive power supply rejection ratio (+PSRR) applies.  If only the negative supply voltage is changed, then the negative power supply rejection ratio (-PSRR) applies.  If both supplies are fixed in value but shift together, then the CMRR applies.

Bootstrapping the supply voltages is easy enough and adds the common mode rejection ratio of the amplifier and the bootstrap circuit together.  (1) This could be done to achieve the 140dB or higher CMRR needed in this application while using a non-chopper CMOS or JFET input operational amplifier for even lower input current and lower broadband noise at the expense of increased 1/f noise which is probably what matters in this application.  Chopper stabilized operational amplifiers have flat 1/f noise making them especially suited for low noise DC applications.  Broadband noise is easy to filter out so there is no reason to minimize it.  Electrometer applications do not have a choice and have to put up with the greater drift and 1/f noise of a JFET or CMOS device.

Ive read that diodes, which are illuminated to generate a minimal current, could be used to cancel the remaining bias current. Are there any existing designs which use this concept?

Photodiodes are not required for that and I think they would be a bad choice.  There are a couple of ways to implement fixed or adjustable bias current cancellation externally using diodes or transistors or even a second operational amplifier of the same type.  I have never seen this done though for a chopper stabilized operational amplifier.

(1) It adds the separate common mode rejection ratios in dB.  As a fraction or ratio, they multiply.

[Echo88]

Just like Santa i made a list, to check which OP behaves naughty or nice. Maybe someone else finds it useful. If i made an error please point it out, the prices mentioned may not be all accurate or for the particular op-amp-grade.

Anyway, at the moment im lacking the knowledge to make a better voltage follower than the one made by Jim Williams. So i continue looking at Datron DMM-service manuals (1051 and 1061 for example) for info on supply bootstrapping and the infos from AoE3 and Bob Peases articles about femtoampere- and teflon-stuff.

[Cicada]

Might be worth your while using TI's paramateric search.

Just select Vos, Offset driftand Input Bias current tick boxes. Sort by Vos and scroll down to where you see the input bias is below 1 pA.
https://www.ti.com/amplifier-circuit/op-amps/products.html

Maybe?
https://www.ti.com/product/opa2320
https://www.ti.com/product/lmp2232

You can Analog devices as well
http://www.analog.com/en/parametricsearch/11070

[danadak]

A related topic -


http://www.electronicdesign.com/test-amp-measurement/whats-all-femtoampere-stuff-anyhow


Regards, Dana.

[chuckb]

There are some examples of Bootstrapping the power supplies for improved op amp performance in this old Burr Brown App note.

[MisterDiodes]

Just be very careful picking an op-amp from a parametric list looking at input bias current alone - usually low Ib means higher voltage noise in an IC amp.  By the same token be aware that the typical AZ amp input current noise datasheet spec is generally hundreds of times lower than the -actual- gate current injection noise spikes that occur at the switching freq.  That's not a deal breaker but you have to be ready for it - keep that switching noise current on AZ amp inputs and power rails contained in a safe way inside small area current loops.

Be aware that the ancient '1010 that shows up in old LT app notes:  Even though it gives you some current drive, that comes at the expense of some relatively strong noise density.

It all depends on what your acceptable noise level is on the final output, and how many of those KVD dials are you -really- planning on using?

[MegaVolt]

ADA4530-1  ??