Calibration Procedure for Guildline 9520 Teraohmmeter (for Electrometer cal.)

[e61_phil]

Hi,

I repaired a Keithley 617 Electrometer and I would like to calibrate the meter. I have build some high ohm resistors (1G, 10G, 100G and 1T). To calibrate the resistance values I used a calibrated Fluke 8508A-7000k 1G resistance standard together with my FEMTO DLPCA-200 transimpedance amplifier. (1:10 steps up to 1T). That worked, but maybe there is some room for improvements.

In all the manuals of the Keithley Electrometers (I know the 617 and 619 only) they use a Guildline 9520 Teraohmmeter to measure the 100G resistor. Unfortunately, I couldn't find a manual for the Guildline meter. I wonder how that is calibrated. Maybe I can mimic the Teraohmmeter to achieve better uncertainties.

Does anyone have the (service)manual of the Guildline Teraohmmeter? Or other ideas to calibrate high ohm resistors?

The problem with the high ohm resistors is their very high voltage coefficient. It really matters which voltage was applied for comparison.

Best regards
Philipp

[HighVoltage]

Hello Philipp

I think with your way of calibrating the 617, you have probably reached the limit of accuracy.

The way I have done the high ohm calibration of my 617 in the past is to measure a 100 GOhm resistor with the Keithley 6517B at different voltages and  then rely on this value for the 617 calibration.

In the past I have also looked at manuals for Guildline 9520 Teraohmmeter and even asked Keithley. But no success.

 

[ap]

Hello Philipp,

I had intended to do such work as well, being currently limlted to 100M, but just have not had time to work on this.
The idea was to charaterize the current input of the 3458A in its 100nA range, which typically is pretty precise, but needs verification first. Can be done with low voltage and e.g. a 100M ohms resistance standard. Internal shunt must be taken into account. The use the calibrated 100nA range to measure higher resistances at high voltages from a calibrator.
I was thinking of either building the Standards myself or buying Russian-type Standards. Any experience with those (stability, tempco...)?

[guenthert]

Zlymex built himself a high impedance bridge.  Perhaps that could be replicated.  You'd still need access to calibrated high impedance resistors though.
 
https://www.eevblog.com/forum/metrology/i-plan-to-build-a-high-resistance-bridge/

[e61_phil]

I played a bit with the Femto DLPCA-200 and the Keithley 617. I used both instruments as a transimpedance amplifier only. Which means I connected the analog output of the 617 to my 3458A. The DLPCA-200 is just an amplifier without display.

All tests are done in the 100nA range (K617) or 1e8 V/A (DLPCA-200). The voltage was fed by my Fluke 5440B. To compare the 100Meg out of my Fluke 5450A against the 1Gig reference resistor I swept the voltage in 21 steps from 0 to 10V (100Meg) and from 0 to 100V (1Gig). That results in the same output on the K 617 output.
With that data I fitted a linear function and the ratio of the slopes should be the ratio of the resistors. This way I get rid of the offsets.

A remaining source of errors is the changing burden voltage of the input. It turned out that the 617 behaves better in that aspect than the DLPCA. I used the 617 as a high impedance voltage follower to measure the input voltage of the DLPCA with different input currents. That revealed one should'nt use the amplifier with more than 7V of output to keep the burden voltage roughly constant. But even in that limited range it changes by more than 200µV.
To test that with the 617 I used my 34401A to measure the input voltage (I haven't a second high impedance meter). The 617 is much more stable over different input currents. To be stable is much more important than the absolute value of the burden voltage, because the constant part is eleminated by the linear fit.

I used the ratio of the slopes and the calibration value of my Fluke 5450A to calculate the value of the 1Gig resistor. This value matches the calibration value of the 1Gig resistor by less than 2ppm. But that is of course just coincidence. I have to repeat that measurement several time to get a feeling of the repeatability.

[doktor pyta]

Regarding *CCCP high ohm resistors (photo below).
They are wirewound hermetically sealed devices.
So You can expect virtually no voltage coefficient and low tempco.
They utilize 'microwire' technology: ultra thin (1um) wire in glass insulation.
Here You will find more info https://www.eliri.md/products/microwires.
Models 1Meg - 100Meg are made as typical two terminal resistor in shielded case.
Model P4085 is universal 1G-1000G and it is made as a resistor divider if I understand correctly.
Their prices are still relatively high because the there are not so many competitors in making microwires.

I only have experience with 1Meg, 10Meg, 100Meg rated 1kV models and their resistance is very stable.
I have to warn You that they are fragile during transport, so ask the seller to protect well the device before shipping.



*They are still in production AFAIK.

[e61_phil]

Hello Philipp,

I had intended to do such work as well, being currently limlted to 100M, but just have not had time to work on this.
The idea was to charaterize the current input of the 3458A in its 100nA range, which typically is pretty precise, but needs verification first. Can be done with low voltage and e.g. a 100M ohms resistance standard. Internal shunt must be taken into account. The use the calibrated 100nA range to measure higher resistances at high voltages from a calibrator.
I was thinking of either building the Standards myself or buying Russian-type Standards. Any experience with those (stability, tempco...)?

Hi Adrian,

I also played shortly with the 100nA range of the 3458A. It seems that the Keithley 617 is superior here as well.

100nA generated by a shielded 1Gig resistor supplied with 100V out of the calibrator.

100NPLC is used for each sample

[ap]

Hi Philipp,
thanks, interesting info... quite a difference.
Not yet having a calibrated 1G, I used a 10M and a 1V source in a quick and dirty test, so not apples to apples.
I see only the last digit jumping by +/-1 digit at 100NPLCs on the 3458A in its 100nA range (so +/- 10ppm).

[Kleinstein]

The 3458 uses the shunt method down to the very low currents. For the 100 nA range, there is a shunt in the 5 MOhms range, that is intrinsically noisy. The 617 uses a transimpedance amplifier and can this way get away with less noise from the resistor as the voltage at the resistor is much higher.  The 100 nA are more like an easy range for the 617, while it is the low end for the 3458.

For the very low currents one can also use the rate of voltage change in a capacitor instead of a large resistor. It is much easier to get a good 10 pF cap than a 1 T ohms resistors.
Attached is an Article about low current traceability.

[volvo_nut_v70]

Guildline is very good about manuals, send them an email.

[volvo_nut_v70]

Current models have manuals, there is a cal routine in them...

http://www.guildline.com/manual

http://www.guildline.com/primary-electrical-metrology/highest-level-standards/high-to-ultra-high-resistance-measurement/6530-series

http://www.guildline.com/primary-electrical-metrology/resistance-measurement-standards/high-to-ultra-high-resistance/6520-intelligent-teraohmmeter

[CalMachine]

It's been awhile, but I think the general routine is as follows :

1) Measure Positive Test Voltage at 1 Volt
2) Press -TV Button to get a Negative Test Voltage and measure
3) Turn Zero dial until the + and - voltages match.
4) Connect 10 MOhm resistance standard.
5) Select Rate and Test Voltage settings which yield the best accuracy for a 10 MOhm setting.  (I think would be a rate of x1 @ 1V)
6) Step Wayyyyyy back and let reading settle.
7) Turn CAL dial until your reading matches the known resistance value.
Select -TV button
9) Step Wayyyyy back and let reading settle. 
10) Turn CAL dial until reading matches the known resistance value.
11) This will change your +TV value, so find a nice middle ground between your + and - Measurements.