Fluke 5522A with a Fluke 8588A

[RYcal]

Hi There

I'm hoping I can get some guidance around what could possibly be causing some issues when measuring resistances above 100 kOhms

My readings seem to be very noisy and the higher I go through the ranges I'm getting far from desirable results. Leading me to believe I have an issue with one of my instruments.

1st question is. Would running these through a UPS (eaton 5s700au) be causing any issues?
2nd. I have tried all kinds of cables etc. without any luck I get similar results so I'm thinking the issue isn't with that.
3rd. In particular the 33 Mohm and 109 Mohm test points are bad, The 109 Mohm is reading high and takes an age to settle down on the 8588A.

Is there any way I can do another sort of test to help identify any issues with the limited equipment I have?

Does anyone have a definitive answer on how I should have them connected in regards to shielding/guarding etc. as i have tried all kinds of different ways without luck the only noticeable change is when I put the guard "plug" into the low on the 8588A and the guard terminal on the 5522A the reading goes from reading high to very low.

Both instruments have been recently calibrated the 8588A is brand new and the 5522A second hand from Transcat.

Thanks

[chekhov]

IMHO, you should be a bit more specific regarding "reading high" and so on, show a filled table with readings (if noisy, some graphs).
Also, 5522A has "Synthesized Impedance" (and maybe there is a fault for high ranges, who knows), so have you tried your 8588A with some convenient resistance standard instead ?

[bdunham7]

I don't have either of those instruments and there are others here with more expertise that may be along soon, but i have just a few quick thoughts.

1) Measuring resistance at 100M and above is always tricky.  You need ceiling fans and circulation off, doors closed and you may need to stand back from the bench to get the readings to settle. 

2) Most test cables won't work very well.  I would try taking two solid conductors out of a phone or LAN cable, connecting them to Pomona MDP-style connectors and then twisting them fairly tightly.  Alternatively, use Pomona 1269 BNC-to-banana and use a BNC-terminated coax cable.  Or simply use the twisted solid conductors directly to the binding posts if possible.

3) Guards are tricky. Some are passive, some are driven and they can cause problems if used incorrectly. Unless you study your manual and understand what the guard actually does in each instance, or you have definitive guidance as to how they should be connected, I would try omitting them entirely for now.

4) Look at the specs for each instrument and figure out what the actual uncertainty is at your measurement point.  The 'noise' may be a significant part of the uncertainty--so if your total uncertainty budget is 5000 counts, don't be surprised to see 2000 counts of noise.  Don't skim the specs, read them carefully.

[alm]

The obvious tests to me seem to, in addition to trying different leads:

• Check the same resistances on the 5522A with another decent (say 6.5 digit) DMM
• Check the 8588A with a fixed resistor of similar magnitude. As already mentioned, the 5522A doesn't actually switch a 10MOhm resistor to the output when set to source 10 MOhm resistance but instead simulates the behaviour using semiconductors. It could be that the 8588A does not play nice with this.

Since both products are current Fluke products, you might also consider asking Fluke tech support about this particular combination. Obviously the 5522A was not designed for calibrating the 8588A, but using the 8588A to check the 5522A makes sense to me.

[Kleinstein]

Twisting the cables is likely not a good idea - this helps against magnetic coupling, but this is not a problem with higher impedance.  Twisting may increase the leakage.

The Probem is more like leakage, capacitance, dieletric absorbtion, capacitive coupling (e.g. mains hum) and possibly a tendency toward osciallation: a simulated resistoance may not be stable with all circuits.
There is a natural voltage limit to a simulated resistance - so it may not be compatable with all meters, espeially not those using a relatively high voltage.
A simulated resistance may show limited accuracy as there is naturally additional uncertainty. So for the extreme values realy reference resistors may be better than the 5522.

The instructions to the instruments should tell how to use the guards - one would likely need them and the implementation of the ohms function can vary, so there is no easy universal rule that applies to all meters.

High resistance is naturally noisy, most meters are kind of a compromise usually not that good for the extremes. Also the simulated resistance can show additional noise.

[branadic]

I remember a last years talk at Metrology Days in Kassel by PTB, stating that they too had issues in their K2 comparison, which turned out to be a problem of the PTFE cable they used. Thus, using the proper cable is a challenge for high ohmic resistor measurements. Here is a quote of the presentation:

• in 1999 unknown systematic error detected
• apparently, a solution had been found
• Re measurement in 2012 showed:
the error is back again
• Looking at the result, the possible discrepancy in the CCEM K2 can be attributed to excess dielectric absorption in the wiring used.
• A change of the wiring reduces the necessary waiting time for the measurement.

-branadic-

[RYcal]

Thanks all

I have those exact cables that you mentioned bdunham7 I bought them for this purpose.

When I output 109.9 Mohms from the 5522A the best measurement I can get on the 8588A is 110.0305 Mohms the spec for the 5522A is 0.05795 Mohms, according to my calculations using the absolute 2 year Tcal +/- 5degC specs the accuracy of the 8588A is 0.0296721 Mohms. This gives me a combined uncertainty/spec of 0.0876221 Mohms.

Unfortunately I dont have access to another decent meter or some fixed accurate resistors at this time. Could anyone recommend a good set to look at buying without having to sell a kidney.

Thaks 

[bdunham7]

I more or less agree with your uncertainty numbers, at least what I got using 1year/Tcal+/-5 absolute is pretty close.  As long as you are using the 'normal' mode on the 8588A and not Lo-I, HV or Tru Ohm, the test current should be 100nA which is within the spec for the 5522A @100M.

So how noisy is it?  And have you tried other values such as 50M, 100M, maybe 150M?

As far as a standard resistor goes, I think you probably just need to get a stable low-TC resistor and get someone to measure it for you and send it to you if necessary.  300ppm is what you're aiming for, more or less.  Standards at that level will cost you both kidneys.

As it is, your equipment isn't that far out of whack--and high resistance is one of the most difficult to be precise in.

EDIT:  I forgot to add--since your measurements aren't really all that far off, are you able to measure your local EM/EF fields?  How noisy is your environment?

[bdunham7]

Twisting the cables is likely not a good idea - this helps against magnetic coupling, but this is not a problem with higher impedance.  Twisting may increase the leakage.

I'll defer to you as to metrology practices, but at my level twisted wires has really helped reduce noise, especially in a noisy environment with things moving around.  I suspect that having a more constant capacitance between the leads helps, as well as the actual capacitance being a bit of a filter. I thought of leakage, but I think that uncompromised PVC insulation at 10V would probably be very, very low.  Of course at G-ohms and with ppm accuracy expectations, perhaps 'very, very low' isn't low enough. 

[RYcal]

I more or less agree with your uncertainty numbers, at least what I got using 1year/Tcal+/-5 absolute is pretty close.  As long as you are using the 'normal' mode on the 8588A and not Lo-I, HV or Tru Ohm, the test current should be 100nA which is within the spec for the 5522A @100M.

So how noisy is it?  And have you tried other values such as 50M, 100M, maybe 150M?

As far as a standard resistor goes, I think you probably just need to get a stable low-TC resistor and get someone to measure it for you and send it to you if necessary.  300ppm is what you're aiming for, more or less.  Standards at that level will cost you both kidneys.

As it is, your equipment isn't that far out of whack--and high resistance is one of the most difficult to be precise in.

EDIT:  I forgot to add--since your measurements aren't really all that far off, are you able to measure your local EM/EF fields?  How noisy is your environment?

I'm not to sure if "Noise" is the correct term but it certainly isnt very stable and takes for ever to settle to an acceptable stability, It starts off way high then very slowly the readings come down, this is at 109.9M, I haven't tried any other values within this range but the other resistance ranges seem to be ok all be it very slow to reach there target values.

Why would it be taking so long to "settle"?.

Cheers 

[bdunham7]

Why would it be taking so long to "settle"?.
Cheers

Dielectric absorption?  Try this--set up your resistance measurement in a way that allows you to short the output for a zero-ish ohm reading (don't worry about making it accurate) and set your 8588A to the manual 100M range and the 5522A to your 109.9M setting.  Then remove the short and see how it behaves.  If this was an actual, normal resistor, I would expect it to now read low and settle upwards due to capacitances in the system.  OTOH, if the synthesized resistance circuit in the 5522A is causing this, the behavior may not change.

[RYcal]

I remember a last years talk at Metrology Days in Kassel by PTB, stating that they too had issues in their K2 comparison, which turned out to be a problem of the PTFE cable they used. Thus, using the proper cable is a challenge for high ohmic resistor measurements. Here is a quote of the presentation:

• in 1999 unknown systematic error detected
• apparently, a solution had been found
• Re measurement in 2012 showed:
the error is back again
• Looking at the result, the possible discrepancy in the CCEM K2 can be attributed to excess dielectric absorption in the wiring used.
• A change of the wiring reduces the necessary waiting time for the measurement.

-branadic-

Thanks branadic. I tried to download the article but I dont have permission. Do you have access to it? 

[alm]

I was thinking just trying to measure a decent quality 10 MOhm or 100 MOhm resistor with the 8588A to get a feel for the stability and noise without the 5522A resistance simulator in the picture. For that you don't need a standards-grade $$$ resistor. If 10 MOhm is also problematic with the 5522A, you could even use a handheld meter in DCV mode as a 10 MOhm resistor in a pinch. Use 10 of them in series for 100 MOhm

Another thing, what mode are you using for measuring resistance on the 8588A? Does changing between normal and lo current change anything? Obviously the lo current mode will be even more sensitive to noise.

Looking at the specs for the 8588A and 5522A specs side by side, they are actually cutting it close for the current range. For the 8588A measuring 110 MOhm - 200 MOhm on the 5522A in its 100 MOhm range (which goes up to 200 MOhm), the current would be outside the 5522A's specifications. I think it should be fine for the points you are using, but double check that you're staying inside the limits:
The 8588A measurement current:


The 5522A allowable current for the high resistance ranges:

[RYcal]

Well as it turns out if I put the 8588A into low current mode it trips the 5522A out.

Looking at the specs they match up well at 109.9Mohms the 8588A will put 100nA stimulus and the 5522A will work from 25 to 180nA so its right in the sweet spot.