HP 3478A 5.5 digit meter as null detector

[Conrad Hoffman]

I've found the 3478A to be a handy bench meter with good low range sensitivity and low ohms capability. It does have to warm up a few hours for minimum drift; mine drops about 30 uV over an hour or three. After that it's pretty good. I logged data for half an hour on the 30 mV range to see what the noise was like. Excel file attached. Curious what people think of it as a null meter.

[Kleinstein]

The 3478 is nice for low voltages. The 30 mV range gives sufficient resolution for a low level measurement. It is a nice meter to buy used: only 5.5 digits, but still with sub µV resolution.

The possible weak point is the input current. One should at least first check the input current, e.g. with a large resistor at the input, or watching a 1-10 nF cap charge up.
Chances are the simple null-meter circuit with AZ OP (LTC1050/LTC1052 or simialar) and a modern handheld DMM could give better performance as a null meter. Battery operation reduces common mode leakage to mains.

The choice of DMM can make a difference as old ones only sample for a small part of the time and this way can increase the noise. This is also a slight weakness of the 3478: in AZ mode it the input is read only 50% the time.

It depends on the signal sources how important input current is.

[retroware]

I tend to use my null meters (keithley 155, fluke 845a and hp 419) to balance resistance bridges and also to compare the outputs of KVDs. Thus, input bias current is very important to me. But almost as important are the user interface advantages of an analog meter. Nulling typically involves very small movements of a pot, be it on the bridge or the meter itself.  Even on the slower 1 and 3 uV scales of the analog meters, one can get a feel as to whether they are getting close.  Having a DMM set to an NPLC of 10~100 makes this much trickier. The other trick one can play with an analog meter is null at say the 30uV scale where things are a bit less noisy. On a DMM, once you are on its lowest scale, you are sort of stuck. But, to be honest, I haven't actually tried using my 3458a as a null meter, maybe there's a NPLC setting that is suitable for this purpose.

[Conrad Hoffman]

I have an 845AB and two HP 419s. I still use the 845 but it's a bit noisy and needs service (again). I really liked the 419s but mine have been idle for a long time and need serious service before I could use them. I've also got a couple top-hat L&N galvos, but that's probably going too far back! OTOH, I can imagine that an old galvo like that, combined with a laser pointer and a couple passes across the room, could be seriously sensitive. My 3478A is always right there ready to go. I'll check the input current next.

[The Soulman]

I've been looking for a meter like this, didn't know the 3478a had a 30mV range, that combined with 5,5 digits means 100 times
the resolution from what I'm currently using.. 

Is it me or is something wrong with the excel sheet? can't open it..
Noise is something to consider, curious about that.

[Conrad Hoffman]

I couldn't get any useful numbers with resistors but the meter changes by about 1 mV per 10 seconds with a 0.1 uF cap on the input. Check my math, which is almost always wrong on the first three tries, but is that 10 pA?  BTW, I was wrong about the initial offset. It's not 30 uV, but 3 uV.

[Kleinstein]

0,1 mV/s and 100 nF would be 10 pA.  However such a large capacitor may show some extra dielectric absorption effects if it was charged before. So this looks reasonable good.

The offset of the meter should be lower. At least there should be a kind of user zero adjustment so one could set the offset back to zero.

It's quite some time I last used a 3478, but I do no remember much noise when measuring thermocouples.
The amplifier is a precision JFET OP of that age.  There may be a little more noise as with modern meters, as the ADC is likely relatively slow and thus using a low frequency AZ cycle.

[Conrad Hoffman]

I think I used a polyester, so will try again with styrene or propylene, but I think the number is close.

While riding the exercise bike this morning, and waiting for the torture to end, it occurred to me that one might build up a little plug-in circuit module with a battery and high value trimpot and higher value resistors, to cancel out the input current. Is that something people have done?

As for the noise, it's only moving a few counts, with a periodic spike a bit higher and lower. Not terrible. The offset, as I mentioned above and is wrong in the spreadsheet text, is about 3 uV at turn-on, coming down  to maybe 0.5 uV after a couple hours warm-up, as the graph shows. I'm sure there's an adjustment, but I've never calibrated the unit as it's not far off enough to worry about.

There's an interesting Fluke document on using a DVM as a null meter when working with the 752- can't seem to find it again, but it was in this forum.

[Kleinstein]

One could build a compensation current source. However I don't think this is really needed at 10 pA.

For only 10 pA or a similar small current one could also consider a combination of photodiode and LED to generate the small compensating current. The get 10 pA with less noise from a resistor is needs more than some 50 mV at the resistor and thus at least some 5 Gohms.

[retroware]

10 pA is quite a bit larger than what one typically finds on an opto chopper null meter such as the Fluke 845a.  I measured around 100 fA on my 845A which is important when measuring high impedance sources.

What's interesting is that my Keithley 155 (which uses a mosfet chopper) seems to have a much higher bias current - around 3 pA. Looking at its schematic and adjustment procedure, they are using a voltage source through a 1 Gohm resister to adjust the bias.  I'm going to play around with that to see if I can get it lower.  But, if I'm reading the adjustment procedure correctly, it seems that anything within 5 pA is acceptable (5uv across 1M) so for balancing bridges, the 845a seems to be a better choice.

I have two hp419a's, one with a dead chopper.  I'm considering a couple of different retrofits for the dead one ranging from replacing the chopper with an H11F3 to replacing the amp with a modern AZ amp.  The problem with the latter is getting a low noise, low bias current amplifier is challenging.  I had contemplated the voltage through large resistor solution but like Kleinstein I was worried about the resistor noise.  I do however like his idea of using a photodiode.

[Conrad Hoffman]

I like the photodiode solution. I think this comes down to the impedance you want to measure. The 752A is on the high side, so it really matters. Comparing my voltage references with low impedance it shouldn't matter much, if at all.

[Kleinstein]

With a good quality high value resistor, there is no real problem from resistors noise. The photodiode has shot noise from charge quantization. The Johnson noise of a resistor gets lower, once the voltage across the resistor is larger than some 50 mV (I may be off by maybe a factor 2 with this number). So if a voltage source is available or doable the resistor tends to be better. The photo-diode may also be part of the clamping diode - photo-diodes also tend to be low leakage, though also expensive. The noise from the resistor is not that high: a 1.5 GOhms resistor would be at the 1 fA/sqrt(Hz) level, less than most amplifiers. Something like 10 GOhms resistors seem to be still affordable.

An AZ OP and current compensation sounds very plausible. The AD8628 would be may favorite as it is reasonable low noise (~ 3 times lower than the resistors at the 845 input) and still not that much bias and current noise.
It still needs some filtering at the input, but not that much capacitance. Filtering for noise purpose is anyway better not done directly at the input.

For the 752 the divider output impedance during adjustment and use should be at some 40 K. So 10 pA of bias current would be only 400 nV.

[MegaVolt]

.There's an interesting Fluke document on using a DVM as a null meter when working with the 752- can't seem to find it again, but it was in this forum.

This is all that I managed to find from Fluke, mentioning DVM as a null meter.

[calija]

For me the best is Nanovoltmeter Keithley 2182A, range 10 mV.

Best regards!!!

[MegaVolt]

For me the best is Nanovoltmeter Keithley 2182A, range 10 mV.

Best regards!!!

Keysight 34420a  range 1mV best!

[retroware]

.There's an interesting Fluke document on using a DVM as a null meter when working with the 752- can't seem to find it again, but it was in this forum.

This is all that I managed to find from Fluke, mentioning DVM as a null meter.

Thanks for posting those links. I had seen the first and third but not the second.

With regards to using a DMM to calibrate a 752, it seems like a lot of extra work (calibrate for the bias current, possibly resort to a cap to increase the filter time) over the 'true' null meter approach. I'd be curious to see an uncertainty comparison. Each of the bias calibration steps seems to add to the overall uncertainty.

Plus, the first line in the summary of the 'migrating from dc voltage dividers to modern reference multmeters' document was a bit amusing:

  "In summary, a comparison of measurement uncertainty between dividers and 8508A will often favor traditional dividers."

It always amazes me what one can do with a bunch of well connected resistors and a relatively simple, but well designed, null meter.

[Tomorokoshi]

...

I have two hp419a's, one with a dead chopper.  I'm considering a couple of different retrofits for the dead one ranging from replacing the chopper with an H11F3 to replacing the amp with a modern AZ amp.  The problem with the latter is getting a low noise, low bias current amplifier is challenging.  I had contemplated the voltage through large resistor solution but like Kleinstein I was worried about the resistor noise.  I do however like his idea of using a photodiode.

Mr. Carlson just repaired the photo chopper in an HP 419A:
https://youtu.be/vrmwql2msbU

[David Hess]

Chances are the simple null-meter circuit with AZ OP (LTC1050/LTC1052 or simialar) and a modern handheld DMM could give better performance as a null meter. Battery operation reduces common mode leakage to mains.

A bench meter also has considerable common mode capacitance to ground.

[retroware]

...

I have two hp419a's, one with a dead chopper.  I'm considering a couple of different retrofits for the dead one ranging from replacing the chopper with an H11F3 to replacing the amp with a modern AZ amp.  The problem with the latter is getting a low noise, low bias current amplifier is challenging.  I had contemplated the voltage through large resistor solution but like Kleinstein I was worried about the resistor noise.  I do however like his idea of using a photodiode.

Mr. Carlson just repaired the photo chopper in an HP 419A:
https://youtu.be/vrmwql2msbU

Thanks for the link to the very nice video.

I've pretty much convinced myself that the problem with my unit are the actual photocells themselves.  My neons appear to be firing correctly plus I've tried several different types. The chopper waveform compared to my working one suggest that they have become slow. I don't think its the circuit board, if I swap boards with the working unit, the working unit continues to function.

Still, there might be something to be had by trying to drive the photocells with LEDs. If I remember correctly someone else with supposed photocell issues tried that with little success.

[retroware]

For me the best is Nanovoltmeter Keithley 2182A, range 10 mV.

Best regards!!!

Keysight 34420a  range 1mV best!

Have you measured the input bias current? I'd be curious how it compares to say a 3458A. I have the same question with regards to the Keithley 2182A.

[MiDi]

Attached the specs for input bias currents.

[retroware]

Attached the specs for input bias currents.

Hmmm, not so great for balancing high impedance sources.

[MiDi]

Hmmm, not so great for balancing high impedance sources.

I do not have expertise on that, but what about balancing by current?
Perhaps a high resistor in series with electrometer in amps mode between sources could do the job.

[Kleinstein]

Hmmm, not so great for balancing high impedance sources.

I do not have expertise on that, but what about balancing by current?
Perhaps a high resistor in series with electrometer in amps mode between sources could do the job.

The electrometers low current range use a TIA and thus can have an offset (several mV and not very stable). So they are not an alternative.

The input current of the DMMs makes thinks complicated, but in theory one could compensate for it: do a zero reading with the same impedance. The usual bridges are not that high impedance. Some 40 K and 400 K for Fluke 752 is about the highest one would encounter. 20 pA trough 40 K gives 0.8 µV of offset - that is not very much, so that the compensation does not have to reduce it that much.

[calija]

For DCV calibration of DMM 3458, 8508 and 8081 (8.5 digit) and for calibrators 5720 and 5520 we used 732 + 752 + 2182.

A lot of national metrology institutes do that and don t have problem.

When you do ACAL and REL on 2182A it is excellent on range 10 mV (temperature must be +-1C)

Best regards!!