Null Detectors/Voltmeters - how useful for a turbo-hobby metrology lab?

[Rax]

I may have the ability to acquire a null detector/voltmeter and would love to hear some thoughts on exactly how useful such an instrument is in today's "extreme hobbyist" home metrology lab.

To note, I have a Kelvin-Varley divider about to reach me soon (an ESI RV 722). I also have a Data Precision 8200, a set of ESI SR1 resistors, and a hoard of other pieces of instrumentation (these not being standards). I am thinking to pair with the KVD I rather need a voltage standard, though the null meter, from my understanding, is critical for applications where it can help avoiding any current draw through the KVD. Exactly how that works in practice I'm still studying.

It'd be great to hear some thoughts on how all these instruments can be put to work. I've read Conrad Hoffman's excellent synopsis and also (not thoroughly) Fluke old edition "Calibration - Philosophy in Practice."

[alm]

A null detector has some advantages over a DMM for adjusting dividers, like adjusting a KVD so the output voltage is equal to that from another source (like a differential voltmeter), or when nulling a bridge (like matching resistors in Conrad's article, but also when for example adjusting a Fluke 752A or 720A). You could use a null meter, KVD and a voltage standard to form an old-school differential voltmeter (something no doubt discussed in the Fluke book). You might also be able to use it to compare the ratio of the KVD to the DP 8200 voltage standard for linearity verification / adjustment. It wouldn't be the first on my list, because you can do an okay job with a DMM. But if you an get a good deal, then I think you'll find uses for it.

[The Soulman]

It depends on what null-meter you're looking at and what dmm you already may have (that could also be used as a null-meter) and more importantly,
what problem are you trying to solve?

[Rax]

You might also be able to use it to compare the ratio of the KVD to the DP 8200 voltage standard for linearity verification / adjustment. It wouldn't be the first on my list, because you can do an okay job with a DMM. But if you an get a good deal, then I think you'll find uses for it.

I think this is something that seems pretty useful. The KVD provides a .1ppm resolution and I think that's why these seem to still be used for primary calibration purposes - Fluke still seems to sell the 720A for both arms and both legs in this day and age, which is pretty similar with the ESI I got - and probably also where it far exceeds what a DMM can do (right?).

[Rax]

It depends on what null-meter you're looking at and what dmm you already may have (that could also be used as a null-meter) and more importantly,
what problem are you trying to solve?

Starting with the problem I am trying to solve... a metrology itch?

But seriously, it's repairing and calibrating DMMs and other instrumentation up to whatever level I can possibly get (currently, I'm dissecting a 7.5 digit DMM, as it pertains to its 10V scale...).

The null meters I'm looking at are an HP 419A and a Fluke 845AR.

[alm]

I think this is something that seems pretty useful. The KVD provides a .1ppm resolution and I think that's why these seem to still be used for primary calibration purposes - Fluke still seems to sell the 720A for both arms and both legs in this day and age, which is pretty similar with the ESI I got - and probably also where it's far exceed what a DMM can do (right?).

If depends on which DMM. The HP/Agilent/Keysight 3458A has a better linearity than any KVD, and Fluke has a document where they describe how their 8508A can supposedly replace the old system consisting of a KVD, null meter, standard cell and some other components. But you're unlikely to find such a DMM for the kind of prices you might find a KVD on the used market for.

[Rax]

If depends on which DMM. The HP/Agilent/Keysight 3458A has a better linearity than any KVD, and Fluke has a document where they describe how their 8508A can supposedly replace the old system consisting of a KVD, null meter, standard cell and some other components. But you're unlikely to find such a DMM for the kind of prices you might find a KVD on the used market for.

Thank you for the article. Yes, that's the thing, isn't it - what is affordable at hobby level lab and how can one maximize their capabilities on a budget? I'm not currently in an affordability position for anything 8.5 digits. I barely am able to tackle 7.5 (= afford broken), and we'll see if I can reliably repair, adjust, and calibrate it.

That sweet spot may equal what a primary metrology lab was doing in the '80s, but that's nothing to sneeze at, is it?

[alm]

I think the main reason for replacing the old equipment like KVDs with 8.5 digit DMMs was convenience, speed, and needing less less trained staff. So for a hobbyist using technology as described in the first edition of the Fluke calibration philosophy and practice book, other than standard cells, might be a reasonable alternative.

[bastl_r]

I remember an event where I wanted to check and adjust the reference voltage of my Solartron 7065. The measuring current of the 34401 was sufficient to falsify the reference voltage. However, this only became apparent when I found the switching option to 10GOhm of the 34401...
With the measurement and counterholding with a null detector no measuring current would have flowed.

[Rax]

But that hobby point made, why are 720As and the bunch being sold for a pretty penny these days? What are they're being used for?

[Conrad Hoffman]

I can't imagine having a KVD without a null meter. The question is what null meter. The Fluke 845 is a classic but often has service issues. The HP 419 is an older classic, very nice to use, but often has service issues. ESI made some null meters, harder to find, but very good. Yes, they often have service issues. Anything that used NiCd batteries will have service issues, and that's most null meters. See a trend here? Still, if you're capable of servicing the stuff and can find one for not-crazy money, get a null meter!

[Rax]

I can't imagine having a KVD without a null meter. The question is what null meter. The Fluke 845 is a classic but often has service issues. The HP 419 is an older classic, very nice to use, but often has service issues. ESI made some null meters, harder to find, but very good. Yes, they often have service issues. Anything that used NiCd batteries will have service issues, and that's most null meters. See a trend here? Still, if you're capable of servicing the stuff and can find one for not-crazy money, get a null meter!

Thank you for these pointers. I actually typically invite (reasonable) servicing issues (such as: nothing leaked, etc.), as it adds to the fun, and can make a unit more affordable. These null meters don't seem very complex either, so that's another aspect.

What does matter to me is size, as I'm beginning to run out space very badly at my bench (=...mancave...garage...). So in that sense, the HP would be better. But the Fluke is more sensitive (1uV scale) which I think matters when paired with a .1ppm KVD. So my current pick is the Fluke.

[guenthert]

[..]
What does matter to me is size, as I'm beginning to run out space very badly at my bench (=...mancave...garage...). So in that sense, the HP would be better. But the Fluke is more sensitive (1uV scale) which I think matters when paired with a .1ppm KVD. So my current pick is the Fluke.

   Null meters aren't typically limited by their scale (most have an amplified output, originally intended for strip chart recorders, to which a meter with a larger scale could be attached), but by their noise.  Some (surviving) units are considerably better in this regard than originally specified. 

   Further, on the lowest scale, settling time might be fairly high (in an attempt to suppress noise): 3s for the HP 419A, 5s (iirc) for the Fluke 845 and Keithley 155, which might be a bit taxing when adjusting a pot.

   The AVM-2000 (the null meter of the 21st century) offers to (mostly) independently select input impedance, sensitivity and settling time.  It isn't produced anymore though and I haven't seen it on the secondary market (and likely would be out of my reach anyhow).

[Rax]

[..]
What does matter to me is size, as I'm beginning to run out space very badly at my bench (=...mancave...garage...). So in that sense, the HP would be better. But the Fluke is more sensitive (1uV scale) which I think matters when paired with a .1ppm KVD. So my current pick is the Fluke.

   Null meters aren't typically limited by their scale (most have an amplified output, originally intended for strip chart recorders, to which a meter with a larger scale could be attached), but by their noise.  Some (surviving) units are considerably better in this regard than originally specified. 

   Further, on the lowest scale, settling time might be fairly high (in an attempt to suppress noise): 3s for the HP 419A, 5s (iirc) for the Fluke 845 and Keithley 155, which might be a bit taxing when adjusting a pot.

   The AVM-2000 (the null meter of the 21st century) offers to (mostly) independently select input impedance, sensitivity and settling time.  It isn't produced anymore though and I haven't seen it on the secondary market (and likely would be out of my reach anyhow).

This is helpful, thank you. On a more specific note regarding the settling time, I think you're correct on the data provided, though on an "apples to apples" clarification, I think the 845 is 5s on the 1uV scale, while on the 3uV scale (lowest on the 419A) it records the same 3s as the HP.

[Kleinstein]

AFAIK the Hp419 and Fluke 845 are suitable for relatively low impedance signal sources due to low current noise and low bias current. The lower current than a more normal DMM makes them still attractive, despite the age.

In contrast the AVM2000 is using an amplifier with rather high bias and high current noise. So it is a different class of instrument and not that suitable for higher resistance source (e.g. KVD) or higher impedance dividers. For the input current noise chances are the AVM2000 is worse than many long scale DMMs. If at all there can be an advantage for use with low resistance sources.

[Dr. Frank]

One year ago I wrote articles about Fluke 720A KVD and 752A Reference Divider, using either 3458A or 845A as Null detector for divider alignment  and Differential VM against a 10V voltage reference, e. g. FLUKE 7000. Search in Metrology section, please.
I also described, how to use the dividers, and how to make very precise measurements, competing with the 3458A, and especially beating its 100V range. The 1kV range is known to suffer from power non linearity, visible in it's mediocre spec.
Under certain configurations, the 10..20pA bias current of the 3458A created errors which were bigger than specification or expected uncertainty.
The 845A has fA bias currents, which was superior over the 3458A.
I don't remember, if somebody determined the bias current of the hp419A.
Currents higher than 1pA is not recommended for use with these dividers.
Frank

[donlisms]

I'm ramping up on this stuff; I think it's great, and fun, and I have nothing to impress you with other than enthusiasm.

A hundred years ago, people wrote about what they were doing with studying resistance, and wouldn't you know... those simple techniques still work!  If you believe in learning from the past at all, you might be interested in many documents about resistance from the US National Bureau of Standards, freely available.  A nice collection of a few important ones is in Precision Measurement and Calibration, Electricity and Electronics, Handbook 77, with papers from Frank Wenner (in charge of resistance at NBS for many years, and a good writer/explainer/teacher), and "Precision Resistors and Their Measurement" by James Thomas, who designed the 1 ohm resistors upon which all our resistive hopes were based for so many years.  Wenner also has a good paper on four-terminal conductors and the Thomson bridge, where he walks right up to the door of tetrajunctions using some Maxwell equations and other things I don't understand, then veers off to measurement current for the power company rather than exploring the application to bridges.  Northrup also wrote an interesting book that may be available online somewhere, about designing and using the gear, including lots of galvanometers.

Just replace "galvanometer" with "null meter" as you read.  But the galvanometer actually doesn't have some of the issues that any amplified null-detector must deal with, including input bias current and many other influences, so in a galvanometer you can truly can go for the "infinite impedance" (or zero current) balance state.  And imagine something kind of like a little featherweight coil and tiny mirror attached to a long taught vertical string, with magnets nearby, and bouncing light of the mirror so you can watch the motion of the reflected spot a meter away, or a room away -- very (very) greatly amplified, with close to 0 current.  No friction, just a little mass to move.  Easy nanoamps detectors.  A hundred years ago.  With no digitals.  No nothing, almost.

And then there are clever techniques like Hamon's implementation of the concept of high-precision ratios using moderate-precision resistors and different series/parallel combinations... a simple but very useful idea, and that's why I see some SR1010's sitting on the bench in the picture on the (modern) calibration lab's web site.  Easy-pease better than 1ppm ratios with stone knives and bearskins.

Bridges and dividers are really neat, and can be almost free.  You need a null detector to use them, and the miracle is you're working with ratios, so the absolute voltage or current doesn't get in the way very much.  (You need a voltage reference for other things, of course.

[alm]

The common phrase "infinite impedance" relating to galvanometers and null meters is misleading. What they mean is that as the voltage across the meter approaches zero, the current approaches zero. But this is just Ohm's law. The same happens if you connect a low value resistor between the voltage sources. So much for 'infinite'. Compared to a DMM with 10 GOhm input impedance, a null meter behaves more like a resistor at low voltages, while the DMM behaves more like a current source. A more correct and useful statement is that the bias current is much lower.

Here is the discussion Dr. Frank referred to about using a null meter or DMM to adjust a high-impedance bridge circuit, in this case the 752A self-adjustment circuit, but the same could also be relevant for other bridge circuits if they have a high output impedance.

Keep in mind that the HP 419 and older Fluke 845 meters can have problems with the neon-based photo choppers. There are designs for replacement circuits. Search the forum and xdevs.com.

[Kleinstein]

With a galvanometer the zero point is with zero current. So even if there is some offset from thermal EMF zero reading is zero current.

With an electronic nullmeter a zero reading does not automaticly mean zero current. Ideally they also have a low current, but it depends on the quality and adjustment. The HP419 has a separate adjustment to get near zero current. How low the current for the Fluke meter is likely also depends.  Especially with an old instrument it would be definitely worth checking before use (is an open and shorted input reading the same ?). The test on the bias may resolve a bit better than 1 pA  (1 µV at 1 Mohm input resistance).
One may get around most of the bias effects by reversing the whole meter. This would average out the effect of bias current.

The actually tricky part with no easy work around is the isolation from ground. Ideally a nullmeter would be battery powered to allow for good isolation by design.

[guenthert]

[..]
One may get around most of the bias effects by reversing the whole meter. This would average out the effect of bias current.
[..]

or reverse the source, which one is meant to do in any case in order to compensate for remaining thermal EMV.   

So, within limits, the input bias isn't so much a problem (here *).  On the one HP419A I managed to get (mostly) working (on AC only, replacing the (dead) batteries with a zener/capacitor assembly as suggested by Conrad Hoffman), the input bias is smaller than I can currently measure (less than a few pA), alas the impedance to ground is smaller than specified (more like 6 GOhm).


*) in AN86 Jim Williams et. al. used a 419A to determine the offset voltage of a Op-Amp in voltage-follower configuration.  Clearly this works only for low impedance sources.

[Rax]

I think for me, personally - but others may as well be in my position - one of the biggest benefits of having a KVD handy is the capability of generating exact higher voltages for adjustment purposes.

I don't feel my 8200 is particularly precise in its 100V range (not anywhere near what's going on in the 10V), and so dialing in a dead-on 100V and inputting into the KVD to be sampled down is very beneficial for the far better accuracy.

[Kleinstein]

The KVD can also have a thermal effect from heating, especially in the top most decade. It depends on the quality and resistance on how good the KVD is at higher voltages.
Depending on the model 100 V may already be too high or at least with a reduced accuracy.
In the 100 V there is than also the loading problem for the KVD - not many DMMs have a very high impedance (e.g. > 100 Gohm) in the 100 V range. Usually high Z mode is limited to some 12 or 20 V.

[bdunham7]

In the 100 V there is than also the loading problem for the KVD - not many DMMs have a very high impedance (e.g. > 100 Gohm) in the 100 V range. Usually high Z mode is limited to some 12 or 20 V.

I'm confused by that.  I thought if you wanted to check the 100V range on a DMM usign a KVD you would use a stable but adjustable (or at least trimmable) source in parallel with both the DMM and the KVD which is set to 0.1 and then nulled to a 10V reference.  Thus the impedance of the DMM would not be important unless the source was overloaded, and since the KVD is only 100K, the 10M or even 1M of the DMM would be relatively unimportant.  Likewise, you would calibrate a source using the same idea.  No?

[TimFox]

Nulling the output of a KVD against a stable reference, where at null there is zero current flowing between the KVD output and reference, is normal.
I believe Kleinstein is pointing out that if the output of the KVD is > 10 to 20 V, the finite input impedance of the DVM connected from KVD to common (perhaps 10 megohm) will load the KVD output.
The high-impedance mode of the grounded DVM at lower voltages should not load the resistors, but the DVM's input bias current will give some error.
If the DVM is the null meter, the important error is from the bias current.

[Conrad Hoffman]

You can certainly do HV calibrations with a KVD. My Julie unit is good for 700V with a power coefficient of 0.00005%/watt. Still, that sort of thing is better done with a Hamon divider. The 752 is such, or you can DIY one.