HP 3478A 5.5 digit meter as null detector

[retroware]

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!!

When using the 2182, do you measure and account for the bias current using a technique similar to that described in (using a DMM to calibrate a 752):

https://www.eevblog.com/forum/metrology/hp-3478a-5-5-digit-meter-as-null-detector/?action=dlattach;attach=942832

[Kleinstein]

The Fluke instructions about using a DMM with the 752 is a little odd: At least from the drawing inside this PDF the auxiliary bridge is 2x120K and thus 60 K output impedance. So the source impedance for the self adjustment would be 20 K + 60 K = 80 K and not 40 K. So the bias compensation should be done with 80 K and not 40 K ! If in doubt - use a 0 V source (short) for the excitation and measure the resistance.

For other tests (especially in the final 1:100 mode) the impedance can be higher / different.

[David Hess]

I wonder what the input current of the HP 419A DC Null Voltmeter which uses a chopper (not chopper stabilized) amplifier is.  It also uses a single ended input.

[Kleinstein]

The HP419 is relatively similar to the Fluke 845, but using lower impedance.
I would expect a slightly higher input current and less noise.
On the positive side the HP419 has a well defined dead time and an extra zero adjustment for the DC amplifier part (1 V range zero adjustment). So one may be able to adjust the input current to some degree and reduce the residual AC needed at the input.

The specs for the current ranges suggest < 1 pA for the input current.

[retroware]

I have two HP419s, one working (part of an ESI 242 resistance measurement system) and one that's not.  I've been struggling a bit to measure the working one's bias current. The method of simply measuring the difference between the zero mode and regular operating mode doesn't work for the HP419 due to its low input impedance of 100K. The meter moves around 100~200nv in the operating position which isn't that much different than when I have it shorted. Thus, the noise floor seems to be limiting this particular procedure. Also the lack of a shorting switch on the ESI doesn't help either as I'm always leery of my low thermal short. Ideas for measuring the current in light of the 100K input impedance would be greatly appreciated.

With regards to DMMs, their bias current and suitability for balancing high impedance sources, there are some DMMs that require one to adjust the bias current as part of the cal procedure.  The Datron 1061,71,81 series comes to mind. Indeed, my Keithley 155 meter also has this adjustment. Mine was originally quite high (thanks mrdiode for pointing this out) and adjusting it brought it down to the femtoamp range. I'm guessing with careful adjustment, DMMs with adjustable current one could do the same.

[Kleinstein]

A 100 K input impedance makes it tricky to measure the input current in deed.
To really measure the input current would need some modification.  One could get away with a switch in series to SIR20 to optionally add a large series resistor that. The low voltage ranges could than have some 10 M or similar input impedance. So the bias current test would be much more sensitive.

The internal 1 V zero adjustment should also have some effect on the input current.

For adjusting the input bias: this only works well if the input current does not change that much with temperature. The old Datron 1061/1071 use a BJT bases input stage and thus need the compensation to get rid of quite a bid of input current (e.g. 1-5 nA).
I would consider some compensation possible for an AZ OP.  Here the input current can be quite small to start with (e.g. 10s of pA).

[guenthert]

I have two HP419s, one working (part of an ESI 242 resistance measurement system) and one that's not.  I've been struggling a bit to measure the working one's bias current.
[..]

  The procedure as far as I understand goes like:  apply a (low thermal EMF) short (like a copper strand between inputs), set the meter to SET NULL, adjust bucking voltage until needle dances around 0, set meter to READ NULL and the then indicated voltage is the voltage generated by input bias current across the input resistance (100k in the lower ranges, i.e. for each 10pA expect 1uV deflection).

Edit: unfortunately neither is the input bias current of the HP419A specified, nor a performance test described which would measure it.  After having another look at the schematics, I realize that above won't yield the desired result (on READ NULL the input is shorted and the polarity of the bucking voltage reversed, hence it'll yield exactly the opposite of the SET NULL with inputs shorted).  I'd think now that just ZERO'ing the meter and the setting it to VM mode with open inputs should show the input bias current (of the amplifier) across the input resistance (SIR20 through SIR24, total of 100kOhm) plus SIR27 (15k).

[retroware]

The problem with that procedure is that one runs into to the noise floor of the instrument when one gets down to around 2pA. 2pA across 100K resistor is 200nV which is less than the noise spec of the instrument. So, if the bias current is less than that, one won't be able to measure it. On my unit, that seems to be the case.

[SilverSolder]

The bias current could be measured with an electrometer?

That's what I did with my meters, and found them to be to spec.

[Kleinstein]

The relatively low 100 K resistance across the input makes it essentially impossible to measure the input current as one has one the very low voltage of some 100 nV. A typical electrometer would not work on this case as the TIA will also have some offset on its own. The measured current would be more like the TIA offset divided by 100 K, often orders of magnitude more than the real input current.

From the circuit I would have expected less voltage noise than from the Fluke 845. The visual comparison could be tricky with much more filter action in the 845 and thus a slower response.

[SilverSolder]

One could perhaps chop the current and use a lock-in amplifier to detect it?

[guenthert]

Keep in mind that the input-bias current will vary with time.  I observed peaks in the nA range for the input-bias current of a HP3456A, while the average was in the order of tens of pA.

[SilverSolder]

The relatively low 100 K resistance across the input makes it essentially impossible to measure the input current as one has one the very low voltage of some 100 nV. A typical electrometer would not work on this case as the TIA will also have some offset on its own. The measured current would be more like the TIA offset divided by 100 K, often orders of magnitude more than the real input current.

From the circuit I would have expected less voltage noise than from the Fluke 845. The visual comparison could be tricky with much more filter action in the 845 and thus a slower response.

I'm not sure I understand this.  Won't the electrometer input impedance (in current measuring mode) be much lower than the 100K resistance, so the 100K essentially won't matter - and so, the current seen by the electrometer is substantially all of the bias current?

[Kleinstein]

The electrometers measure small currents as a trans-impedance amplifier. So the input is essentially at the ground potential. However no OP is perfect and there usually will be a small offset (e.g. mV range). One can still measure sub pA current from a higher impedance source, and get a good zero with an open input. However with a low impedance current source there will be current due to the offset voltage. The offset voltage is an unwanted small voltage source in the circuit.

[SilverSolder]

The electrometers measure small currents as a trans-impedance amplifier. So the input is essentially at the ground potential. However no OP is perfect and there usually will be a small offset (e.g. mV range). One can still measure sub pA current from a higher impedance source, and get a good zero with an open input. However with a low impedance current source there will be current due to the offset voltage. The offset voltage is an unwanted small voltage source in the circuit.

Aaah, thank you, yes, I see it now. 

A Keithley 614 here has 0.62mV offset, which would result (does result!) in 6.2nA of current in a 100K resistor...  which will of course "drown out" any current one might be looking for at the pA level - the current suppress function doesn't have enough range to compensate.

[retroware]

Would bucking this offset with the 419's NULL feature work?  Unfortunately, my working 419 is a modified version that ships with the ESI 242 resistance measurement system and it has neither the zero short switch or null feature. I tried using an external voltage source together with a Keithley 616 electrometer but was unable to get any results that I trusted.

[Kleinstein]

The offset of the usual very low current TIA in the electrometers is not very stable - so even if one would adjust or compensate it would another voltage source the TIA is still not suitable to measure the current from a low impedance source.
 
If at all one could use an external nV meter (e.g. Keithley 2182) to measure the voltage at the 419 input.

The other option would be to modify the 419 to allow a high impedance measurement:
This could be an additional resistor (e.g. 10 M) in the ground connection of the input divider from (SIR20) with a switch in parallel. With the high resistor active one would have high Z for the lower few ranges (til 3 mV), but non working higher ranges. With closed switch there might be a minimally higher error for the higher ranges (maybe + 1% at 1000 V, less with lower ranges) - so probably not a big deal.  The input current would than be better measurable with some 100 times finer scale than the original.

[SilverSolder]

Would bucking this offset with the 419's NULL feature work?  Unfortunately, my working 419 is a modified version that ships with the ESI 242 resistance measurement system and it has neither the zero short switch or null feature. I tried using an external voltage source together with a Keithley 616 electrometer but was unable to get any results that I trusted.

The offset is simply too big for the nulling / compensation features to be able to kill it off.

I'm thinking that an external voltage source to cancel out the offset might work, provided the source is stable enough...

[retroware]

Would bucking this offset with the 419's NULL feature work?  Unfortunately, my working 419 is a modified version that ships with the ESI 242 resistance measurement system and it has neither the zero short switch or null feature. I tried using an external voltage source together with a Keithley 616 electrometer but was unable to get any results that I trusted.

The offset is simply too big for the nulling / compensation features to be able to kill it off.

I'm thinking that an external voltage source to cancel out the offset might work, provided the source is stable enough...

I was using a Fluke 5440B but I wasn't particularly careful with my connections. Maybe I'll try again.

[Kleinstein]

Even of the external source is stable to compensate the electrometer offset, the electrometer offset itself is not stable as it is from a high Z JFET or CMOS amplifier. So expect at least 10s of µV of variations. If in doubt just measure the voltage at the TIA with a reasonable good DMM.

[SilverSolder]

Even of the external source is stable to compensate the electrometer offset, the electrometer offset itself is not stable as it is from a high Z JFET or CMOS amplifier. So expect at least 10s of µV of variations. If in doubt just measure the voltage at the TIA with a reasonable good DMM.

I did notice the instability that you are talking about.  -  but the magnitude of the instability is much less than the magnitude of the offset voltage?

[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!!

Yes, we used this!!!!


When using the 2182, do you measure and account for the bias current using a technique similar to that described in (using a DMM to calibrate a 752):

https://www.eevblog.com/forum/metrology/hp-3478a-5-5-digit-meter-as-null-detector/?action=dlattach;attach=942832

[Kleinstein]

Even of the external source is stable to compensate the electrometer offset, the electrometer offset itself is not stable as it is from a high Z JFET or CMOS amplifier. So expect at least 10s of µV of variations. If in doubt just measure the voltage at the TIA with a reasonable good DMM.

I did notice the instability that you are talking about.  -  but the magnitude of the instability is much less than the magnitude of the offset voltage?

How large the absolute offset is, depends on the individual units. One could even have an adjustment for the offset. Without adjustment I would expect up to a few mV, so considerably more than the usual variations. Just have a look at the specs of an OP suitable for an eletrometer (e.g. LMC6001 : some 1 mV range offset and 10 µV/K drift - 0.1-10 Hz noise is expected to be in the 10 µV range). A TIA is just the wrong way to measure the bias current for a low Z input. It is kind of hard to separate a bias from an internal offset. If there is some thermal EMF at the resistor, these can also cause some kind of bias. With 100 K it only takes 0.1 µV to get 1 pA. So if low bias is needed it helps if the null meter is also high impedance. The 100 parallel resistance can be a limiting factor for the HP419.

[Kleinstein]

Measuring the open circuit voltage is the way to measure the bias. However the resolution is limited, as 1 pA corresponds to only 0.1 µV. So it needs a good meter - ideally better than the 419, but than there is bias from that other meter  .

It would be easier if the input resistance would be higher: than the 419 could have sufficient resolution.