Seeking schematic or advice on Keithley 6103C HV Probe

[GEOelectronics]

A schematic would probably be best - what little info readily available indicates 10^12 Ohms which is perfect but is there a divider or is is a series resistor only?

Is it possible to adapt this probe to a KS 34470A set to high impedance input, and use MATH function to correct
the reading?

Mostly I need to measure a low current, high impedance 1kV to 2.5kV bias power supply.

Thanks

George Dowell

[e61_phil]

https://www.tek.com/manual/model-6103c-high-voltage-probe-instruction-manual-28273b

In the manual everything is described. It is a divider and the lower resistor has 375Meg

[e61_phil]

The 34470A is specified with > 10G in high Z mode. That will give 3.6% of additional error (max).

[GEOelectronics]

Thanks for all, that will work nicely.

George Dowell

[GEOelectronics]

When the probe comes in I'll measure the resistors inside, test it with a known 1kV, then calculate my correction factor and maybe even the true input resistance of the DMM.

George Dowell

[GEOelectronics]

The 34470A is specified with > 10G in high Z mode. That will give 3.6% of additional error (max).

This probe wound up being a 6103A model, which is 10^12 Ohms input impedance, even better for my needs, but the readings on my DMM will be off even more because the bottom resistor is about 940M Ohms.

Thanks for your help.

George Dowell

[e61_phil]

The input impedance of a meter isn't in high Z mode really an impedance, it is more an input current which can vary with the input voltage.

To test the behaviour of the 34470 (this is only valid for the one I tested and could be different on another 34470A) I put 1Gig in series with the meter to mimic your probe (940Meg isn't that far from 1Gig). I run through 0 to 10V and plotted the error.

[e61_phil]

And -10V to 10V in smaller steps. Clearly no constant error.

[GEOelectronics]

Thanks for that analysis, which is now printed out and kept with my documents as a reference on this project.

Did you by chance run one with H-Z mode, 1G Ohm in parallel and 0-2V across the 1G?

George Dowell

[e61_phil]

Did you by chance run one with H-Z mode, 1G Ohm in parallel and 0-2V across the 1G?

Sorry, I can't follow you. 1G in parallel to the input of the meter? That would be shorted by the voltage source.

If you're aiming for a comparision of your 1:1000 divider at 1000V to 2000V than you can compare the graphs above. That is measured in high Z with 1Gig source impedance like your divider has (Thevenin’s Theorem).

[GEOelectronics]

Yes in parallel, the ~1G is across the input impedance and the 999G is in series with the HV, 2 resistor external divider, yielding 1V at the DMM terminals per kV.

If the Hi-Z is really changing that much, I'll have to use the internal 1G and correct the readings, .2% is nowhere near close enough for my testing needs.

Thanks, the charts are real eye openers.

George Dowell

[e61_phil]

Could you draw a schematic? I'm not sure if I really understand the setup you would like to be measured by me.

I attached what I meant:

V1 is a 10kV source divided by 1000 which gives 10V with 1Gig output impedance (999G in parallel with 1G).
V2 is 10V (1000 times less) source with also 1Gig output impedance.

If you view the two circuits only from the two output ports they behave equal. It is like I had measured -10kV to 10kV with your divider

[GEOelectronics]

V1 is the circuit I have and will try to use. That's what's inside the Keithley 6103A divider, it is a 10^12 Ohm divider, not 10^11 like the 6103C (thanks to advice here). My thinking is 10^12 is a million MegOhms, or 1000 G Ohm (correct me if not)


My measurements are in the 0 to 2kV range only for this project. The readings that must be taken is series with an unknown M Ohms resistor in series the the already low current HV bias supply. Therefore it is imperative to not drawn much if "any" current.  The actual load does not draw any current unless it avalanches, when it draws max current (which is limited to microamps by the PSU). We then derive a specifically shaped pulse from that rapid change via a pF capacitor.

With my KS DMM, there are two possible settings concerning input impedance (resistance), which will be in parallel with the 1G resistor of the divider.

Hope that's enough info.

George Dowell

[e61_phil]

You have to use high Z. If you switch your meter to 10Meg you will have a ~1:100.000 divider

[GEOelectronics]

You have to use high Z. If you switch your meter to 10Meg you will have a ~1:100.000 divider

Correct. Since most my measurements are at 1kV, 1kV = 0.01 V (I think),

 the question really is, how many usable digits would that leave me and is it withing the capabilities of my DMM to read say 0.01001 or oven 0.010005?

George Dowell

[Kleinstein]

A 1 GOhms resistor in parallel to the high Z meter, or the meter in an 1 G ohms mode does not improve things, It is more like making things worse. The extra 1 G resistor is affectively parallel to the lower leg of the divider, changing the divider to a slightly different type. The error from the amplifier would be still there.

For better accuracy is just need a divider with lower impedance - so more like a 10^9 ohms divider and not a 10^11 or even 10 ^12 ohms one. So the diver should be more like 1 GOhms and 1 M.

The error from the DMM input current would generally also include some bias current. The impedance itself can vary between units and with temperature / humidity.

[GEOelectronics]

For better accuracy is just need a divider with lower impedance - so more like a 10^9 ohms divider and not a 10^11 or even 10 ^12 ohms one. So the diver should be more like 1 GOhms and 1 M.

Understood, thanks and that's how I've been doing it, with 1G 1% resistors, trimmed.

Looks like I need an electrometer?

Would a Keithly 616 suffice to use this 10^12 divider with stability and confidence?

Before going that route we have several Keithley 35050 electrometer adaptors with triax input connectors, battery operated, they take the signal and provide a 0-2V output to any DMM.

I use several electrometers by Keithley, Fluke, Victoreen, Femto-Tech and others but they are married into testing systems and read out in various radiation dose rate or kVP instead of volts.


George Dowell

[Kleinstein]

The higher the resistance the more problems with leakage currents. With a 10^12 ohms divider every pA of leakage current from the meter would result in an 1 V error (independent of the divider ratio). It would still help to not have too much of a divider, so that the reading amplifier can be optimized for low currents with less care about voltage offsets / noise.
One can see the initial high resistor as a current source - so it's about measuring some 1000 pA in one way or the other. Unless one needs the TC matching of the divider, one could also use the meter in current mode and just a series resistor.

The other problem is that the very high resistors tend to be less stable/accurate than lower ones.

The Keithley 616 is old, but should be still good for a really high impedance signal.

[GEOelectronics]

The higher the resistance the more problems with leakage currents. With a 10^12 ohms divider every pA of leakage current from the meter would result in an 1 V error
The other problem is that the very high resistors tend to be less stable/accurate than lower ones.

The Keithley 616 is old, but should be still good for a really high impedance signal.

Thanks, this board has been very helpful. I'll try several approaches and report my final "industrial" system on the Test Equipment board.

George Dowell