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

[RoadDog]

Fluke wrote the article on replacing null detectors with 8 1/2 digit meters. At their Josephson junction though sits a Fluke 845ab. I’ve used it. Not the junction they didn’t let me touch that thing but I borrowed their 845 for my 720 bench a couple times. They trust it over the 3458a and their 8 1/2 digit meters. It has worked for over 50 years. It’s too bad they quit making them.

I recently picked up an 845 and it doesn’t seem to have any issues. The one I regularly used at Fluke took forever to settle which is a capacitor that needs to be replaced. It had occasional issues with certain ranges too. There are troubleshooting steps in the manual for the 845. I was just lucky and got a newer model that has the newer redesigned circuit in it so no neons or old leaky caps yet. 

[Dr. Frank]

Fluke wrote the article on replacing null detectors with 8 1/2 digit meters. At their Josephson junction though sits a Fluke 845ab. I’ve used it. Not the junction they didn’t let me touch that thing but I borrowed their 845 for my 720 bench a couple times. They trust it over the 3458a and their 8 1/2 digit meters. It has worked for over 50 years. It’s too bad they quit making them.

This is due to the high bias current of any DMM, compared to the 845A. Usage of a DMM might be ok for several ppm error and under certain input/output conditions, but not if you go @ 1ppm or below. Additionally, the 3458A shows high AZ current spikes, which would affect the JJ array.. Other DMMs and the 845A might be lower.

Frank

[Kleinstein]

For just replacing a nullmeter it does not need 8 digits - with slightly reduced performance (1 µV resolution) a AN8008 handhald DMM could be an option.
The point is that the 8 digit DMM can in quite some cases replace the old combination of  precision 10 V reference + KVD and null-meter.

For some strange reason there seems to be no good modern direct replacement for the old fluke 845.  It is a rather small market (mainly the users of the Fluke 752) and there seem to be still enough of the old one around. I would not consider the AVM2000 a valid option - it looks like the bias/current noise is similar / worse than with many DMMs. At least it offers battery operation and thus the option to get low leakage to ground, which is the real show stopper for many bench DMMs. The meters bias courrent could be corrected by polarity reversal, but this is time consuming (idelly aim for symmetric +- reading instead of a simple 0).

[lugaw]

I still don't understand the sensitivity specification of Null  detectors measured in millimeters.

0.6 x 10^-3 microvolt per millimeter.

Where can I read about it?

[alm]

I still don't understand the sensitivity specification of Null  detectors measured in millimeters.

0.6 x 10^-3 microvolt per millimeter.

Where can I read about it?

Try old texts about galvanometers. The sensitivity helps you determine how much voltage is necessary to produce a readable deflection of the meter. If a meter claims a sensitivity of 1 nV, but this only produces 1 um of deflection, then this may not be very usable. Or it might mean you need optical aids to read the meter, like in the old sensitive mirror-galvanometers. See page 7 of NBS monograph 39: Calibration Procedures for Direct-Current Resistance Apparatus.

[Conrad Hoffman]

Old tech like a mirror galvo, plus new tech like laser pointers, can give you a very long baseline, even more if you fold it (and double the sensitivity) with an additional mirror.

[iMo]

What would be the resolution of a vintage mirror galvanometer, a strong laser, and the beam projected on the Moon surface? 

[TimFox]

There is a simple system described on p. 490 of Valley and Wallman, Vacuum Tube Amplifiers, vol 18 of the Radiation Laboratory Series, McGraw Hill 1948
A laboratory galvanometer projects onto a differential phototube (two vacuum photodiodes in a single envelope), driving a simple 6SF5 high-mu triode amplifier, with feedback to the galvanometer coil.
As drawn, this is to make a very low burden resistance microammeter, but it could be re-configured to make a null voltmeter.
The very high gain from galvanometer coil current to grid voltage easily compensates for the drift in the DC input bias of the 6SF5.

[lugaw]

I still don't understand the sensitivity specification of Null  detectors measured in millimeters.

0.6 x 10^-3 microvolt per millimeter.

Where can I read about it?

Try old texts about galvanometers. The sensitivity helps you determine how much voltage is necessary to produce a readable deflection of the meter. If a meter claims a sensitivity of 1 nV, but this only produces 1 um of deflection, then this may not be very usable. Or it might mean you need optical aids to read the meter, like in the old sensitive mirror-galvanometers. See page 7 of NBS monograph 39: Calibration Procedures for Direct-Current Resistance Apparatus.

Thank you very much! I forgot it was part of Physics class.  I had a null detector for a year now and could not find out about the millimeter ratings and google was not much help.

[ivo]

I am wondering if anyone has any experience to answer this question:

I am looking at the possibility of building Conrad Hoffman's nullmeter using a dual precision opamp. One side could do rail-splitting, and the other side can do the actual difference multiplication.

However most opamps also ask for say, a couple of 100n caps around their supplies. The problem is that given those caps will then be both loading the output of one opamp, while also being used to stabilise it as a whole, mostly it seems this would produce oscillations from capacitive loading, from what I can see trying to set up the circuits in LTSpice.  Also the datasheets tend to warn against that in general as well. This is one advantage of the original component, TLE2426, it is designed to be stable after you add enough bulk capacitance. If interested, I was looking at an OPA2186 currently, which seemed to have a nice combination of decent uV offset, low current bias, and low noise. So I have some queries-

1. Anyone know if you're likely to see these oscillations in practice, trying to operate mostly a simple DC circuit with seconds of stabilisation time?
2. If operating off some simpleton batteries, like 1x9V or 4xAAs, would it be possible to forgo the capacitors all-together? Would this be likely to ruin performance or noise of the opamp / circuit in use?
3. Or anyone know of a scheme of compensation to allow for this design to work?

If this design can be stabilized you could design a seriously tiny board to implement this instrument.

[iMo]

You do not need a precison opamp for the rail splitter with that null meter. Any opamp is ok.
You may add a small resistor, like 22ohm at the opamp's output, but it is not needed (no need for a large cap at its output).
The decoupling capacitor C1 at the first opamp (and the C2 at the splitter) should be wired between its Vcc and Vee (see below). Mind the 9V battery is floating.
EDIT: improved the schematics for clarity

[ivo]

You do not need a precison opamp for the rail splitter with that null meter. Any opamp is ok.

I know it is definitely not "needed", but these days the dual version of a lot of precision OPs are barely $0.5-1.50 more than than the single part. If it can be put to use, it saves size, BOM, part count, ease of design, and probably battery life.

The decoupling capacitor C1 at the first opamp (and the C2 at the splitter) should be wired between its Vcc and Vee (see below).

I know a lot of references suggest caps from Vcc - Gnd and Gnd - Vee, I'm not sure how necessary that is for opamps generally (instead of straight across Vcc - Vee), or specific to this application? Maybe a confusion with my understanding in their proper bypassing lies in that 80% of cases looked at these days, Gnd is also Vee?

[Echo88]

As a suggestion:
Heres my working implementation based on a OPA170, chosen for its very low quiescent current.
Simulation suggested 100R on OP-output is fine for stability, in reality it was of course not with such a big capacitive load and it oscillated, indicated at first by its high current draw of about 8mA instead of the specced 110µA.
Thats why the 3k3 and 10nF were added, though i did not do a stability analysis and the values might be improved.
I dont know if omitting caps like in ivos design is having any downsides.
My AD8628-based nullmeter achieves about 200nVpp in a 0.1-10Hz bandwidth, still needs to solve some minor bugs and test it fully for every spec that can be found in usual nullmeter-datasheets.

[Kleinstein]

With some 50-100 Ohm most OP-amps are OK for the rail splitter. So this could be something cheap like LM358, TLC271 or similar. Some capacitance from the virtual ground to the supply can be a good idea, but one usually does not need much.

If looking for a precision OP-amp that can tolerate capacitance at the output the OPA202 is a good candidate.

A single OP-amp for the input would have less effect from heating of the amplifier. With a low power part and a realtively low voltage this would be only a small effect.

[leighcorrigall]

I recently bought a null detector for my home lab.

Combining a Fluke 5440B, 752A and an 845AR makes an excellent differential voltmeter with self-calibration capabilities based on the 732A. The thermal stability and linearity are good as well. When combined with a digital voltmeter, the output of the null detector can be measured to acquire the mean and standard deviation of the null. Since I already had everything but a null detector, the purchase was obvious. I paid approximately 100 USD for my unit. Anyone can get a decent null detector at a realistic price given enough patience and help from friends.

Good luck! 

[ivo]

Anyone can get a decent null detector at a realistic price given enough patience and help from friends.

Good luck! 

Eh, I think this can depend a lot upon where you are living in the world 😅

[lugaw]

Anyone can get a decent null detector at a realistic price given enough patience and help from friends.

Good luck! 

Eh, I think this can depend a lot upon where you are living in the world 😅

I got my working Keithley 147 for $35.2USD shipped.

[Rax]

I got my working Keithley 147 for $35.2USD shipped.

That's fine for bragging rights, but Ivo has a perfectly good and considerate point. I used to live in the US Pacific Northwest and everything was expensive and hard to come by. I now live in Southern California and as I walk out the door, I'm tripping over some free instrumentation that needs minimal repair (if even that) and goes right back in spec once fixed. And I still regard both places being relatively generous with the opportunities vs. others.
And everyone has a budget, which needs to be managed so opportunity definitely makes a sizable difference.

[lugaw]

I am just adding data points to ivo about test equipment price depends on where you are on part of the world. In the previous two country I lived at, I couldn't even find a local seller for a null meter.

[Conrad Hoffman]

A decade or more ago when many local companies were replacing equipment or just shutting down, cheap test equipment was plentiful. Now the supply has dried up completely. We're starting to have hamfests again, but they're nothing like they were pre-covid.

[iMo]

As a proof of concept I dusted off my ADI's ADuC845 board (the 845 chip includes 2x 24bit SD ADCs and an 8052 mcu, ++), I put it into 160mV FS differential mode, 2x 1k resistors for input's protection with 47nF foil for blocking, I hanged an HC-10 BT on it, I powered it all from an old 9V wall adapter (transformer one), all I put into a small paper box. With some basic sw it reads the voltage, it does an adaptive EMA, sends out the voltage, stddev, chip temperature. I replaced the ADI's ref with a REF5025, but I doubt it is relevant in this app. I see the same as with my 34401A (but with a "better" resolution)..

PS: added a quick measurement