Null voltmeter

[Echo88]

Full ack on the last post Kleinstein and now i get why you mentioned a MOV: the symbol that is used in front of the main OP is a MOV-symbol, i intended to use it as a symbol for a thermal jumper but forgot to add the Theta (dont know if theres a more fitting one) 
We´ll see in the end if they aid the project.

Heres the changelog for the current version, that is attached:

Low leakage diodes are now BAV199, typical leakage with <1pA is good enough, see:
https://www.eevblog.com/forum/projects/forward-leakage-of-a-diode/msg1287718/#msg1287718

Added LDO-regulators

Detailed the battery indicator section, taken from the Datron 4910. Current consumption with suitable
LEDs is marginal compared to the OP quiescent current. Only implemented for pos battery, should be good
enough as an indicator supposing nullmeter quiescent current draw is sufficiently symmetrical during operation
and recorderoutput draws current from positive supply, making it the fastest draining battery and therefore the most reasonable
to choose for "battery low"-warning.

Improved recorder-section, should now be bipolar input capable and draw absolutely neglegible current, see
https://www.elektronikpraxis.de/galvanisch-getrennte-emi-freie-messschaltungen-a-1102943/
Before i assumed the IL300 needs mA LED-drive current and V->f-converters that i found also needed at least 1.5mA

A 6V-lead acid battery enables 2.5V and 5V supplyvoltage, as the LDO-dropoutvoltage and deep discharge voltage
of the battery is fitting for the purpose. With 1.3Ah they should last 13d at expected 4mA max current draw.

What still needs to be done:

Finalizing all component values and producing a BOM
Checking the correct function of the IL-300-section

I found a few nice cases from Hammond and suitable rotary switches, gotta do the mechanical planning now.

If someone has bipolar capable very low quiescent current (<1mA) V->f and f->V-converters im also interested.
Im wary of digital isolators (capacitive and inductive) and similar stuff as a solution, since id like not to worry about the resulting EMI.

[Kleinstein]

If all amplifier parts get away with a +-2.5 V supply (or maybe a little less) one could opt. for a single battery and virtual ground. So only one 5 V regulator and than a OP-amp to buffer half of that as a virtual ground.

The IL300 part should get away with a less expensive amplifiers. The signal is +-1 V range and I don't think one would need a 5 digit dynamic for the output. Some offset can be compensated relatively easy.
For high resolution use one may want a little more current there. It is a compromise and still relative easy to change later if needed.
For the output side It may help to have only 3x1.5 V (or oven just 2x1.5) so that 5 V range amplifiers are possible.

The 10 K for the input protection are relatively small compared to the 2x22 K in the filter. For ESD spikes it would help if the 10 K can withstand higher pulses too.  So the combination looks a little odd. I would try to get away with smaller capacitors, so that the input does not get too slow with a high impedance source.

The LP filter is quite slow responding - it may help to have the option to bypass the filter, especially for the analog output or higher voltage ranges. The low speed is mainly needed with the low ranges to suppress hum and noise.

[JohanH]

What still needs to be done:

Finalizing all component values and producing a BOM
Checking the correct function of the IL-300-section

I found a few nice cases from Hammond and suitable rotary switches, gotta do the mechanical planning now.

If someone has bipolar capable very low quiescent current (<1mA) V->f and f->V-converters im also interested.
Im wary of digital isolators (capacitive and inductive) and similar stuff as a solution, since id like not to worry about the resulting EMI.

This is interesting. I like the approach. Any status update?

I'm building a KVD and already have the Hoffman Null detector adapter for a DVM. But it would be awesome with an analog display. I have to finish the KVD and some other projects first, but after that I will take a look at this. The meter and switches will be most expensive. I have some Elma switches, but they are only specified at 35V DC. There are also Chinese rotary switches specified for 150V DC. Could be a better alternative.

As Kleinstein mentioned, a single battery would be nice, too. Which op amp are you using (planning to use?) in the latest revision (Main OP, low pass and Post OP)?

[Echo88]

Hi Johan,

im not quite finished with it. But so far it works nicely apart from 3 small bugs, and needs some cosmetic changes in the mechanical pcbs and i'll exchange the knobs for more fitting ones.
Noise is about 150nVpp with the used AD8628 which are used for the complete analog path for simplicity.
It differs slightly from the schematic in the last recent post:
-only one 12V-battery is used together with a railsplitter and a ultra low quiescent current DCDC-converter for the isolated output
-with the used 7Ah battery the runtime is about a month at roughly the same case size as nullmeters like HP419A/AVM2000
-a battery charge/protection circuit was added
The panelmeter was taken from a damaged K155, but the circuit has provisions to also use normal (non +-) panelmeters.
The idea for the binding post thermal baffle was taken from the AVM2000. It can be screwed on the case-backside when not used.
 

[JohanH]

That's so awesome. I already found a null meter (with mirror scale) on ebay and ordered it. M265M 100 - 0 - 100 µA. Not as nice as yours, but will do.

What devices did you connect to the guard post? I think in null meters such as Keithley 155 only potentiometer shields are connected. Not sure about that, though. Manual states that guard should be tied to circuit gnd (-) if not used (they do it on the rear terminal). And is your inner box connected to the chassis?

[Echo88]

The guard is connected via two capacitors to outer case and negative input like in the K155, the inner case (isolated with standoffs) is also connected to guard. That should aid CMRR in my understanding, but i could be wrong.
See for reference http://hparchive.com/Application_Notes/HP-AN-123.pdf
Guard connection depends on the specific use case as lined out in the manuals of nullmeters like the K155.
Specific measurements like CMRR, NMRR, noise, etc. to completely characterize the nullmeter werent done yet though.

[ecclectrics]

Dear All,

I understand that the topic is not very recent, but hopefully not completely inactive. Thanks to all of you for sharing the design and knowledge! Please excuse if my questions are somewhat elementary as I am still rather new to precision design.

I am particularly interested in the bias compensation circuit shown in Echo88's schematic above (https://www.eevblog.com/forum/metrology/null-voltmeter/msg4623649/#msg4623649). I wasn't aware of this possibility, which might be useful also for other applications.

I wonder how the adjustment will be made: First shorting the inputs together and trimming the output voltage to zero with the offset compensation and then leaving the inputs open and trimming the output voltage to zero with the bias compensation? Possibly the two adjustments are mutually interdependent an need to be repeated iteratively?

And is the 10M resistor to GND in the lower ranges of the input divider still absolutely required even after the opamp gets its bias current through the compensation network? Or could a 10G input resistance like for the lower ranges of many precision DMMs be easily obtained? (Whether it would be desired for the present application is a different question of course.)

Finally, what is the purpose of the 47k resistor to GND in the bias compensation network? To maintain a definite state if the pot wiper looses contact?

Help is greatly appreciated!-)
Roland

[Kleinstein]

There should be essentially no interaction from the bias compensation on the zero offset. A repeated adjustment would be more to cope with offset drift.

The 10 M resistors to ground at the input is not really needed at all. It would be OK to have an essentially floating input with only parasitic resistance (e.g. in the Tohms range). Many high impedance meters are made this way. The specs are something like > 10 Gohm, but the actual input resistance could well be > 100 G. How large the input resistance will be depends on the details of the build, like parasitic leakage and the amplifier chip. > 10 Gohm should be relatively easy.

The 47 K to ground in the bias adjustment are there to divide down the voltage before the 1 Gohm resistor.

[ecclectrics]

Thanks for your explanation Kleinstein!

So no interaction between the adjustments, good. But the basic procedure is correct: open inputs for bias and shorted inputs for offset?

Concerning the 47k resistor, I should think of Thevenin equivalent of the 3 1Meg's (see image), which would be very roughly around 470k, so with the 47k it gives roughly a divide by 10.
Is there any advantage for doing it that way rather than simply using ~100k instead of 1M for the pot?

[Kleinstein]

There is a slight advantage with the resistor to ground. The 1:10 attenuation also applies to thermal EMF generated at the POT. I don't know if this gets relevant.

The offset adjustment can still effect the bias trim, though not much (one should aim for the same reading for the short and in the bias step).

[ecclectrics]

Ok, thermal EMF of the pot - what all needs to be taken into consideration!

One other question I have in particular to you: in some other threads on the Null-Voltmeter from Conrad Hoffman (e.g. here https://www.eevblog.com/forum/projects/conrad-hoffman-null-detector/msg2893432/#msg2893432) you suggested smaller values of the caps for the input filter, which were at 0.47µ there. Here they are with 1µ even bigger. I wonder what the pros or cons are for having smaller (i.e. 47n .. 100n) or larger (i.e. 0.47µ ... 1µ) caps in the input filter?
I have actually cobbled together a version of the Hoffman circuit with some jellybean parts, which was a valuable lesson on 1/f noise. With the opamp I used it was several µV and severely limited the possibility to get the offset really low. Now the autozero opamps seem to have it in the order of several 100nV depending on the type, but is still an issue if I want to get the offset lower than that. So I think a lot of filtering will be necessary for me, and I am looking for the best way to do it without messing up something else.

[Kleinstein]

The capacitors at the input give a filter response, where the source impedance adds to the resistor in the circuit. With a high impedance source this can become sluggish and dielectric absorbtion can make it even more sluggish than the RC time constant suggests, if there was a much larger voltage before.
With an analog display one may want the analog filter fuction, especially for the highest gain range. Thus the filter at the amplifier that changes with gain.
For use with a digital display it would be better to do the final filtering digital with a more FIR type filter and thus a finite response time instead of exponential decay type. Digital filtering is also easy to make flexible.

The filter capacitors also are part of the ESD protection to reduce the peak voltage in addition to the diodes.

[Vtile]

Has the parasitic resistances ( cabling ) talked about. This is a subject most of the times overlooked when the null detection is spoken with wheatstone bridge example in academia and totally forgotten the wiring resistance from equation given in these examples (which is much more complex with wire resistances) this of course is mostly for bridge setups using null methods.

[guenthert]

Has the parasitic resistances ( cabling ) talked about. This is a subject most of the times overlooked when the null detection is spoken with wheatstone bridge example in academia and totally forgotten the wiring resistance from equation given in these examples (which is much more complex with wire resistances) this of course is mostly for bridge setups using null methods.

   I presume you mean the leakage of the insulation of those cables, as the resistance of the conductor is very low (expect it to be less than 100mOhm) compared to the input resistance (100kOhm in case of HP419A, 1MOhm in Keithley 155 and Fluke 845A) and irrelevant at null.

   I'd think the leakage of the cabling is implicitly included in the leakage of the case.  Many bridge circuits have the detector at ground potential in order to minimize leakage effects (however then requiring very good isolation of the generator, see e.g. ESI 801).

[acstd90]

Have tried the AVM. Maybe someone else may have had better luck than myself but they seem to have problems with Noise, Drift and the battery is always dead (self discharge)
Mine holds down a spot Under the workbench.

[acstd90]

There is a paper written by Neil Faulkner on the Fluke website that discusses the errors associated with replacing the Fluke 845AB.
Also in the October 2023 issue of Cal Lab magazine is an updated article written by David Martson (1A CAL GmbH) on the pitfalls of replacing the analog null meter with a DVM as a null meter for use with a Fluke 752A and calibrating a Fluke 5720A

[e61_phil]

There is a paper written by Neil Faulkner on the Fluke website that discusses the errors associated with replacing the Fluke 845AB.
Also in the October 2023 issue of Cal Lab magazine is an updated article written by David Martson (1A CAL GmbH) on the pitfalls of replacing the analog null meter with a DVM as a null meter for use with a Fluke 752A and calibrating a Fluke 5720A

Hi,

do you have the 1A Cal article? I also started a discussion with them, because they said, that they are using a Keithley 182, but in my experiments there where several problems by replacing the 845A simply by the Keithley 182.

Best regards
Philipp

Edit: found it