Null voltmeter

[ap]

The AVM2000 seems to be discontinued. The 3458A, at least with high impedance sources such as the Fluke 752A, is not a solution, I did check with one of my 3458As and the bias current is just too high. Would be interesting to see how a 8508A behaves. So seems the only items still available are the EM Electronics one. Other than that, older meters such as the K155.

[Echo88]

https://doc.xdevs.com/docs/Valhalla/2720GS/2720GS%20VOL.%202%20SCHEMATICS.pdf Page22 Nullmeter

@jfphp: Would be a pleasure if you could share more infos/pictures about your Valhalla 2720GS, since the forum and generally the internet is lacking that kind of info of this very interesting device. AFAIK only the user Kosmic has another 2720GS.

[HighVoltage]

I have been using the Keithely 155 Null Detector for some time now and it gives repeatedly good results.

It is still on the first et of new batteries, after I repaired it here:
https://www.eevblog.com/forum/repair/keithley-155-null-detector-repair-and-restauration-attempt/

[Kosmic]

https://doc.xdevs.com/docs/Valhalla/2720GS/2720GS%20VOL.%202%20SCHEMATICS.pdf Page22 Nullmeter

@jfphp: Would be a pleasure if you could share more infos/pictures about your Valhalla 2720GS, since the forum and generally the internet is lacking that kind of info of this very interesting device. AFAIK only the user Kosmic has another 2720GS.

Was about to start a thread about my repair progress. With full disassembly pictures 

Should be able to find some time next week to post something.

I also have a Fluke 845ab in the repair queue, so I might be able to do a comparison at some point.

[Echo88]

I was indeed thinking about asking you for potential teardown-photos and infos Kosmic. 
Such a thread about the Valhalla 2720GS would be much appreciated by all members here i think. 

[Kleinstein]

The Nullmeter part from the Valhalla looks really simple:  Mainly just an AZ OP with switchable gain. So this one could have a bias current in the 3-10 pA range, so maybe a little better than a good DMM, but not much.

A null-meter of that class could be a relatively simple DIY project, maybe even dead bug or on a raster board (The LTC1052/ICL7652 is available in DIP). As a very small circuit, battery powered operation is very practical.  I would even consider adding bias current compensation, so one should be able to manually adjust the bias to the low/sub pA range.

[ArthurDent]

There was some discussion in the 'TEA' group about the HP 419A starting around post #25975. I posted this with a couple of photos of my replacement batteries in post #25993.

"Like beanflying in post #25975, I had to do a lot of cleaning and replace the leaky and dead battery packs in my HP 419A. I replaced the original batteries with packs that were designed as phone replacement batteries and they fit with little difficulty. The new batteries have been in the HP 419A for a couple of years on what was fast charge for the original smaller batteries but is trickle charge on the new larger ones. " 

[Conrad Hoffman]

I've got a couple 419s and it's probably time for new battery packs again. One is a bit older than the other and I remember it having some mysterious issue that turned out to be corrosion on the front panel where the input jacks go through. It was causing leakage. Removed the jacks, cleaned everything up, and no more problem. They're really quite a good meter with good ergonomics.

[guenthert]

[..]
A null-meter of that class could be a relatively simple DIY project, maybe even dead bug or on a raster board (The LTC1052/ICL7652 is available in DIP). As a very small circuit, battery powered operation is very practical.  I would even consider adding bias current compensation, so one should be able to manually adjust the bias to the low/sub pA range.

Like https://drive.google.com/open?id=1yzzUKX1S4_AR7HiQNTvz-Z_wjBTFUkpp
https://drive.google.com/open?id=1NALNYXngEHq8iBmX4gUHewEG81J5gXoT ?
(Well, yes, I'm not paid for my handiwork  -- this is just to show that anyone can build a null meter

[Dr. Frank]

Sorry for unearthing this thread, but maybe we can discuss the following question here:

I often read the claim that a Nullmeter like K155 or Fluke 845AB have the property of an infinite input resistance when Udiff=0 at its input-posts.
...

Therefore it should be possible to use any suitably high resolution, low noise and low bias current DMM as a nullmeter, as long as the bias-current from the meter doesnt introduce new errors due to high bridge-resistance ..
Fluke mentions this in an appnote, but also states that Nullmeter-input-resistance is infinite at null, which i dont understand. ->

Can anyone point to my error in thought or is the appnote wrong in claiming "Input resistance infinite at null"?

[..

Hi!
We had exactly this discussion 2..3 years ago.
My criticism about the 845a was, that its bias current was NOT specified, instead the false expression of 'infinite' input resistance @ very small input voltages was chosen.
It's obvious, that <1uV over 1MOhm creates <1pA, which gives virtually very high impedance, when comparing 10V sources, for example. 10V/1pA gives 10¹³Ohm, zero difference infinite resistance, virtually.
A really low bias current is the important feature, and another owner really measured that bias of the 845a to be lower than 30fA, due to galvanic separation by this optical input chopper.
As this had been proven, I also bought an 845A, because my 3458A had a few pA of bias, really influencing the trimming/calibration of instruments like the 752A, or the 5450A, or any other instrument relying on the Wheatstone Bridge alignment method.
Therefore the Fluke app note that long range DMM can replace these old null voltmeter, is simply wrong. Even the 8508A has pA of bias. The trimming accuracy of these old bridges might be just sufficient using a DMM, but it's really no fun, as you really notice the limitations by their bias currents.

I could also test my 845A for fA bias AND leakage to case, so the alignment of my DIY Hammon divider now is much better and easier than using the 3458A .


The hp419 might also have low bias, due to optical chopper, but that has to be measured.
I have no idea about the other instruments.
It's clear, that all the chopper ICs have much too high bias, also. Mostly, low bias currents and low noise are contrary requirements.
Frank

[e61_phil]

Does anyone looked into the input current over time on the 8508A?

Perhaps, it is much better than the 3458A, because it is not doing auto zero. But if one measures zero there shouldn't be much charge due to auto zero.

[1audio]

How do you measure the input bias current of a meter when its that low? You need to separate the bias from leakage current which may be difficult. Would it me meaningful to measure the voltage across an open input?

Traditionally a sensitive galvanometer would be used as a null detector. Internal resistance in the 10's of ohms BUT when the voltage on both sides is the same no current and effectively infinite impedance. Switching to an electronic indicator should be easy if you can keep that really high leakage requirement. However a null detector is all about the null. Once you go off null the infinite is no longer infinite.

I have been playing with one of these: https://quantasylum.com/products/qa350-microvolt-dc-volt-meter as a nullmeter. I checked its isolation between the USB (its power source and display host) and the circuit when I got it on my HP 4329 and I got something like 5 X 10E14 Ohms, essentially the limits of the instrument at 100V. I did not want to go further then.

Pluses- 24 bit at 5V FS or 50V FS. Seems pretty stable. Running off of a PC is nice if you have enough isolation. It seems the resistive isolation is really high and the capacitive is in the pF. Software is all open source so you are free to hack away.

Minuses- Input is a BNC connector (not good when you consider thermals( I added a pair of banana jacks). Input impedance is not that high at 1 MegOhm. Software does have a bug in the logging, it locks up at about 8000 records. Software's slowest logging is 2.5/second so a long session may have a lot of records.

Odd-  is that its converting the binary output of the ADC to decimal so at the lowest steps they are a little odd with skipped decimal steps. The software display is very similar to a DVM and it graphs so you can get an immediate display of fluctuations. Attached is a plot measuring my 732A. I believe most of what is shown is internal noise on the 732A.

[beanflying]

There was some discussion in the 'TEA' group about the HP 419A starting around post #25975. I posted this with a couple of photos of my replacement batteries in post #25993.

"Like beanflying in post #25975, I had to do a lot of cleaning and replace the leaky and dead battery packs in my HP 419A. I replaced the original batteries with packs that were designed as phone replacement batteries and they fit with little difficulty. The new batteries have been in the HP 419A for a couple of years on what was fast charge for the original smaller batteries but is trickle charge on the new larger ones. "

Actually the 419A is back on the bench this morning fitting in a floating 1.35V Mercury Battery Replacement board, the main packs have already been done. Also in the list for refurb is a Fluke 895A and also the HP 740B and sort out the Null meter section on my 335D.

I don't have a Null meter problem, I don't have a Null Meter problem ..... 

[MegaVolt]

(Well, yes, I'm not paid for my handiwork  -- this is just to show that anyone can build a null meter

Can you show your scheme?

[Kleinstein]

How do you measure the input bias current of a meter when its that low? You need to separate the bias from leakage current which may be difficult. Would it me meaningful to measure the voltage across an open input?
....

For measuring the bias / input current there are 2 common methods: one is measuring the zero voltage with 2 different resistors at the input, usually a short (or near short) and something like 10 M or 1 M. If the meters own input impedance is relatively low (e.g. 10 M) one can also use the open circuit voltage.

The other method is using a small, low DA capacitor (e.g. 1 nF or whatever is suitable for the current range)  and measure the voltage drift rate. With a 1 nF cap 1 pA of current give about 1 mV/second. As one usually only need the rough size of the current even a manual reading is often suitable. Not so relevant to a null meter, but with a DMM this method also allows to get the input current at a different voltage.

[Bill158]

How do you measure the input bias current of a meter when its that low? You need to separate the bias from leakage current which may be difficult. Would it me meaningful to measure the voltage across an open input?

There is an interesting App note from Fluke about this subject of correcting for input bias current when using a DVM/DMM for a null meter.  See attachment.
The F845A takes care of it's input bias current by having you turn the "OPR/ZERO" switch to "ZERO" before making a measurement like balancing the bridge adjustments in the F752A or the F720A.  If you look at the schematic of the F845A you can see that the 1 meg resistor is in the circuit so doing a zero null takes out any offsets from the measurement when in OPR.
I have used the method in the App note and have made excellent adjustments to the F752A and then compared the adjustments using a F845A.  I use a HP 3458A.  I find that the 3458A has around 8 pa of input bias current, at least mine does.  I guess the biggest issue here is that your DVM/DMM has an extremely high isolation resistance from both the high and low inputs to either the guard or ground depending upon how the guard is connected.
Bill

[guenthert]

Can you show your scheme?

Uh, oh.  I was thinking to put it in a project page (it's rather a beginner's project than belonging in the metrology section) or add it to my new project: the Null Meter, Mark II (now digital 

It's in Dave CAD format and I really need to look at the device to verify it, but unfortunately, it's in a storage facility half way around the globe.

There isn't much to it though.  I used a LF411 (not the best choice, but what I had laying around) as trans-conductance amplifier driving the meter for the mV range.  Making sure the MCM wasn't over-driven while simultaneously minimizing gain error was one of the greater challenges (one of the requirements was a 2V range, as this null meter ought to do double duty display for my Keithley 617 analog output -- that turned out to be a questionable idea, as it made input protection considerably more difficult).  For the uV range I used good-ol LTC1050 (I wanted to use a OP27 with a discrete matched FET pair as input buffer, but I couldn't get my hands on a reasonably priced one) as fix gain (1000x) pre-amplifier.  For the burden [bucking] voltage generator I used a single AAA cell and had a good look at the schematics of the HP419A.

Noise I did combat with heavy handed low pass filtering.  I went back and forth changing the capacitors as the first version was overdampened (response time around 13s to 90%) taxing my patience.  Overall the performance is just so-so (1 and 2uV ranges are unusable, 4uV is ok-ish, but the LF411 drifts too much;  resistance input to ground about 250GOhm as it is battery powered and unlike commercial products provides no output), but was a great learning experience.  It's really something to do to understand what not to do next time 

[Kleinstein]

There are different null meters. Some have lower voltage noise and other have lower bias / current noise. Which type of amplifier is best, depends on the signal source.  For a simple, low cost version AZ OPs can be an option. One would still want some input filtering against spikes and EMI.
Besides the AZ OPs just listed, I would consider the AD8628 as an intermediate between LTC2057 and LTC1052.
The higher bias OPs may want a compensation for the bias current. However this does not help against current noise - so the use with high impedance sources is still limited.

[guenthert]

i must have mis read that null meters need very low ibias

  Well, it might be the wrong way to look at things.  When comparing the potential of two voltage sources with low output resistance, it is not needed and you might as well use any DMM or use a long scale DMM to measure those voltages directly.  If however, you want to compare two voltage sources of which at least one has a 'high' output resistance, then the input bias current of your measurement instrument might make a significant difference. 
  One of the remaining use cases of a traditional null meter is the calibration of Fluke's reference divider (752A).  I believe its output resistance is about 40kOhm.  An input bias of e.g. 10pA (not untypical for a long scale DMM) yields then an error of 0.4uV, while the calibration procedure requires a null detector which can determine balance to within (i.e. the total error required to be less than) +/-0.5uV ...

[Kleinstein]

One of the points where a null meter can shine is using battery power and thus essentially avoid coupling to mains ground. So the null meter can be really floating. Otherwise a null meter is more like a sensitive voltmeter (e.g. 100 nV range resolution), that does not need high accuracy or high speed. It is often more about having extra filtering / low speed to reduce the noise.

If the signal source are low impedance (like 2 voltage references) one could as well use a normal mains powered DMM with sufficient resolution.
If the signal source is high impedance (like a relatively high value divider) a classical mains powered DMM could be a problem and a low input current is needed.

So the main need would be more with low input bias (e.g. < 10 pA, maybe less with adjustment).  A higher input bias and lower noise version would be more like nV meter - still useful in some cases. When using AZ OPs the circuit could be essentially the same - just swap the input OP.

Compared with a normal voltmeter one could save on having less protection, no need for a super stable reference and less gain accuracy.  So it could be a relatively simple DIY project:  protection + input amplifier  + ADC chip + µC + LCD powered from batteries.  An ADC with a little more resolution than the old analog scales could simplify gain switching as less gain settings are needed. So 1:10 steps should be ok - no more need for old days 1:3:10.  With low voltages mechanical switches can become tricky and it could be easier to use CMOS switches to choose the gain. So at least for the low end part without a divider one could have some auto-range function.

For low leakage one could consider dead bug style for the input part - so one may not even need a PCB.

[MegaVolt]

The ADS1263 has an integrated chopper amplifier, PGA, input impedance of 1 GΩ and 8 nV RMS noise.

You can play by purchasing these boards:
http://www.yoctopuce.com/EN/products/usb-electrical-sensors/yocto-bridge
or
http://www.yoctopuce.com/EN/products/usb-electrical-sensors/yocto-maxibridge

[Kleinstein]

Quite some SD ADCs have zero drift inputs. However they usually have relatively high input current. Especially the ADC1263 is more like a high end ADC and likely needs a lot of attention to the board to get the full performance.

I think it would be better to have a separate amplifier like the AD8628 ( about twice the noise of the LTC2057, but more like 1/10 the input current). So one could chose between low bias (e.g. max4238) and low noise (AD4528) or intermediates.

For the ADC I would more think about a low end SD converter like MCP3421 or a slightly faster (and external clock to work in sync with the chopper) version if a discrete chopper is used.
At least in theory a discrete chopper could reach a better noise / bias combination than the usual AZ OPs. However the discrete chopper would need more power and more circuit.

[iMo]

Why do you need an ADC and an MCU and a display? With 1-2 opamps and 9V battery you may build a null voltmeter in an hour, the output could be almost any DMM, imho.

[Kleinstein]

Many of the cheaper DMMs sample the input only for a small part of the time and thus have a higher than needed input bandwidth. This would require a relatively slow filter between the amplifier and DMM to avoid extra noise.  So amplifier + DMM is still possible. Instead of a full DMM one could also consider a small LCD type panel meter.  The ADC+ µC/LCD part would be a little like a more flexible panel meter version and one can choose an ADC that does more continuous reading.  Digital filtering in the µC can be of FIR type and thus get faster settling than analog filtering that normally is IIR type.  Depending on how well one knows the µC, the effort for the ADC+µC is not that large.

[guenthert]

  There are many ways to skin the cat, but building a low noise, low drift, low offset amplifier isn't exactly trivial (which I only realized after I tried ;-)

  Taking apart a HP 419A is quite revealing.  Despite its age (early 60s?) and compared to modern units like the AVM 2000 mere modest performance (p2p noise <.3uV, 100kOhm input resistance, unspecified input current), it is quite a sight to behold.  The robustness, attention to detail and effort (gold (!) traces on the PCBs!) is quite impressive.   Best of all, they were once fairly popular, but are long obsolete and with a little luck can be acquired for reasonable prizes on the secondary market (unlike the competing, a little younger, slightly better performing products from Fluke and Keithley).  I was once of the opinion that everyone could build a Null meter (I mean, how difficult can that be?) -- after attempting it myself with rather mediocre results and seeing the HP 419A, I have now more respect for the subject.

  Someday I might attempt again to build one.  Then I probably go the MCU with built-in ADC and external chopper route.  That seems promising and comparatively cheap.