34401a >10G Ohm input resistance, slow voltage drop after disconnecting leads

[hp3310a]

Hi all,

  I'm playing with my "new" Agilent 34401a and an LM399H based voltage reference and I noticed something very strange:

I set the input resistance to >10GΩ (increases precision as per the manual), connect the DC voltage reference and let it sit for a couple of hours (don't know if it's relevant how long the measurement lasts). When I then disconnect the leads, the voltage doesn't drop immediately to next-to-zero. Instead you can watch the reading on the meter decrease at a rate of ~25 seconds per Volt, as if a big capacitor was discharging. The reference outputs ~7V.

This does not happen with the input resistance set to the standard 10MΩ. The reading goes to next-to-zero the moment the leads a unplugged from the meter.

I didn't find anything searching for it, neither whether this is a normal nor (if not) what causes it. The manual doesn't mention it so I guess it's not normal, so....

Question: Is this normal? Question 2: What could cause it?

[Kean]

Completely normal.

The very high input impedance means that any charge will take a while to dissipate.
In fact, if you short the inputs so it reads zero, once you remove the short you will see a charge build up slowly.  The charge will increase faster the more you move anything near the inputs.

[coromonadalix]

there was discussions here

on very high impedance,  even no connections to anything you could slowly charge the probes inputs  etc   ...  and  could discharge slowly .....

for sure this behaviour is not on 10 Megs input impedance


you don't need to go on the high impedance  if you measures resistances under 100Meg, and remove the auto impedance mode

for me  on 4 wires mode you are more precise  for ohms

for voltages  i never go in high impedance,  don't need to at 6.5 digits  on a 34410a  i have ?

https://engineering.purdue.edu/~aae520/hp34401manual.pdf      is that manual you talk ?

to check references voltages stability, you do it on longer acquisition times, play with nplc measurements settings you'll see,   you could even grab the measurements into a graph  ....


chapter 7
Low-Level Measurement Errors

When measuring ac voltages less than 100 mV, be aware that these
measurements are especially susceptible to errors introduced by
extraneous noise sources. An exposed test lead will act as an antenna
and a properly functioning multimeter will measure the signals
received. The entire measurement path, including the power line, act as
a loop antenna. Circulating currents in the loop will create error
voltages across any impedances in series with the multimeter’s input.
For this reason, you should apply low-level ac voltages to the
multimeter through shielded cables. You should connect the shield to
the input LO terminal.

....

A high-impedance source is
more susceptible to noise pickup than a low-impedance source.

[tggzzz]

Instead you can watch the reading on the meter decrease at a rate of ~25 seconds per Volt, as if a big capacitor was discharging.

Time constants are the product of capacitance and resistance. You have a very high resistance, so time constants will be long.

There might also be leakage currents in the DVM input. Depending on the sign, they could cause the reading to either increase or decrease over time. Have you tried inverting the inputs to measure a negative voltage, and observed what happens?

Whether or not you need to use the HiZ setting depends on your source's output resistance and the accuracy you wish to obtain. The source output resistance and the meter input resistance form a potential divider; you can use basic analysis to determine the consequences.

[hp3310a]

Thanks for the comments. As for the output impedance of the source, I don't know. It is one of these:


Reversing the leads does change the reading of the least significant digit from +7.02643 to -7.02644. What this tells me I'm not sure.

According to the specs of the voltage reference, the output is supposed to be +7.02645V, so my DMM is about -20µV off. If I am not mistaken that is one order of magnitude better than the DMM spec'd accuracy (DC 10V Range, 1 Year,  0.0035+0.0005 % of reading/range ~ 251µV). The last calibration of the unit was 2009. Looks pretty OK to me if the reference voltage is really what it says it is.

[Simmed]

there is i think 470p + 220p inside the front end that soaked up this DCV
assume your description 25s per volt drop = 40mV/s
(40mV/s * 0.69n) = about 28pA leak
assume 7v vref (R = V/I)
input impedance is over 200G = pass 

[Kean]

Thanks for the comments. As for the output impedance of the source, I don't know. It is one of these:
(Attachment Link)

Reversing the leads does change the reading of the least significant digit from +7.02643 to -7.02644. What this tells me I'm not sure.

According to the specs of the voltage reference, the output is supposed to be +7.02645V, so my DMM is about -20µV off. If I am not mistaken that is one order of magnitude better than the DMM spec'd accuracy (DC 10V Range, 1 Year,  0.0035+0.0005 % of reading/range ~ 251µV). The last calibration of the unit was 2009. Looks pretty OK to me if the reference voltage is really what it says it is.

I am also testing with one of these against my 34465A, two 34401A's, and my K2400 SMU.

The Vz of the AliExpress LM399 reference is shown as 7.04508V.
My 34465A reads 7.04507V (-10uV).
My older 34401A reads 7.04511 (+30uV, or +40uV from my 34465A).
My newer 34401A reads 7.04545V (+370uV, or +380uV from my 34465A).

I also have a new ADR1399 reference from CircuitValley.
The Vz of the ADR1399 reference is shown as 6.93572V.
My 34465A reads 6.93466V (-1060uV).
My older 34401A reads 6.93469 (-1030uV, or +30uV from my 34465A)
My newer 34401A reads 6.93502V (-700uV, or +360uV from my 34465A).

These offsets are pretty similar to what I see when I tested with my PDVS2mini set to 10.00000V, but I don't have those number recorded right now.

I've just been checking on these off an on while they have warmed up over the last couple of weeks, and just using a bunch of banana leads and both front and rear terminals.
I need to set up to collect some statistics as I think the LM399 will have more noise despite being closer to my 34465A calibration.  And also test lead reversal.

I really wasn't expecting the cheap AliExpress reference to read so close.   
I am going to add the suggested snubber circuit to my ADR1399 reference, but that could affect the documented reading.   

Edit:
Forgot to mention the K2400 SMU reading which was 6.93291V on the ADR1399 (-2810uV, or -1750uV from my 34465A - not unexpected as I have no ideal of its calibration history).
I also have another 34401A on the way soon... and a bunch of other 6.5 & 5.5 digit DMMs/DAQs to check.

[tggzzz]

Thanks for the comments. As for the output impedance of the source, I don't know. It is one of these:

The devil is in the (internal) details

Reversing the leads does change the reading of the least significant digit from +7.02643 to -7.02644. What this tells me I'm not sure.

But does the voltage then drift in the same or the other direction as your previous observation? (I.e. towards zero or towards negative).

According to the specs of the voltage reference, the output is supposed to be +7.02645V, so my DMM is about -20µV off. If I am not mistaken that is one order of magnitude better than the DMM spec'd accuracy (DC 10V Range, 1 Year,  0.0035+0.0005 % of reading/range ~ 251µV). The last calibration of the unit was 2009. Looks pretty OK to me if the reference voltage is really what it says it is.

IIRC the 34401 use LM399s internally - but HP/Agilent takes a great deal of care to age them and then only select the best.

[Alex Nikitin]

I’ve described in another thread a way to measure the input current of a high-Z voltmeter.

Cheers

Alex

[bdunham7]

I set the input resistance to >10GΩ (increases precision as per the manual), connect the DC voltage reference and let it sit for a couple of hours (don't know if it's relevant how long the measurement lasts).

10G input resistance doesn't increase the precision directly, what it does is reduce the error caused by circuit loading.  Any time you connect a meter to something it becomes part of the circuit.  So if you were measuring a voltage divider consisting of 2 10M resistors and a 10V input, you would expect 5V across either resistor.  If you connect a 10M voltmeter, the resistor you are connecting across effectively becomes 5M resistor when combined with the meter and you'd measure 3.333333V, not 5.  If  you used a >10G meter the error would be negligible and you'd see the 5V.

For a low-impedance source like your voltage reference this makes little difference to the precision of the measurement.  The residual readings and noise that you see are because those voltages are actually there and the extremely high input resistance of the meter does not drain them off.  The input impedance of your source is irrelevant since it is now disconnected, the input impedance of the meter is the issue.  Your 34401A is very sensitive in the high impedance mode, the input currents are in the 10s of picoamperes.    Everything you have reported is indicative of a perfectly functioning 34401A.

[hp3310a]

Thanks again for the comments, the 34401a is a really cool device and I can't believe my luck buying this one. That's not usually what happens if I buy something used.