Input impedance in Fluke autoranging meters

[FlexibleMammoth]

Hi all,

I bought a Fluke 80K-40 HV probe recently because it was cheap and I wanted one for a long time, and had a close look at the manual. Being a regular voltage divider network, the HV probe depends on a 10MOhms input impedance of the meter. I have a Fluke 289 and had a close look at the impedances, and found that the impedance changes as a function of the range:

50mVDC: 10MOhms
500mVDC: 10MOhms
5VDC: 11MOhms
50VDC: 10,11MOhms
>= 500VDC: 10MOhms

Volts AC is AC-coupled, I didn't check these ranges yet. More important: The Fluke 289 manual specifies 10MOhms at all (!) ranges (page 77).

Input Characteristics
V          10 MOhm <100 pF
mV       10 MOhm <100 pF
V~        10 MOhm <100 pF (ac-coupled)

However, the manual of the HV probe states that this is normal:

Specifications
The 80K-40 will achieve rated accuracy when used with a
voltmeter (ac or dc) having an input impedance of 10 MOhm±1.0%.*
Specifications for the probe are as follows:
(...)
Input Resistance: 1000 MOhm
Division Ratio: 1000: 1 (1000X attenuator)
(...)
* The input impedance of Autoranging Fluke handheld digital
multimeters varies as a function of range. The only range that
deviates significantly from 10 MOhm is the 3V (Models 21, 23, 25,
27, 70, 73, 75, 77) or 4V (Models 10, 11, 12, 29, 79, 83, 85, 86,
87, 88) range where the impedance is 11.11 MOhm
. To enhance themeasurement accuracy when using this range, apply a correction
factor of 0.99, i.e. multiply the displayed reading by .99.

My question at this point (because google did not help):  Why is that the case?
- If it is a feature, I cannot see why anybody would want that. That 10% increase in input impedance does nothing but throw my readings off. If you give me high impedance, give me proper >1GOhms.
- If it is a bug, I do not understand how a company that has been building handheld multimeters for decades cannot overcome this problem.
- Saying "just multiply by 0.99 sometimes and you'll be fine lol" is a bit weak for a $1000 equipment combination in my opinion.

I would be happy if anybody has some insights to share.

Best regards,
Andreas

[fcb]

Interesting behaviour.

Here are some results from my Fluke 89IV (tested with a 34465A).

mV DC (manual) setting
50mV DCV range: 9.970 Mohm (took about 15 seconds to settle down)
500mV DCV range:  9.987 Mohm ("")
5000mV DCV range: 9.989 Mohm ("")

V DC (manual) setting
5V DCV range: 10.992 Mohm
50V DCV range: 10.090 Mohm
500V DCV range: 10.006 Mohm
1000V DCV range: 10.003 Mohm

On the last page of the manual, it also says 10 Mohm 'nominal'.
http://assets.fluke.com/manuals/87_89iv_umeng0200.pdf

Bit of an oversight from Fluke?

[ch_scr]

I have recently listened to this Amphour with a Fluke engineer: https://theamphour.com/180-an-interview-with-dave-taylor-multi-talented-meter-maker/
I am pretty sure the design restraints and considerations leading to this behavior where discussed there.

[FlexibleMammoth]

Thanks for you podcast suggestion featuring a former Fluke engineer. As far as I understood, he said a lot about how Fluke tried bringing (older) multimeters closer to 10MOhm, but he did not mention that (or why) modern autoranging meters have this weird impedance characteristic.

@mods:
Being a n00b in the eevblog forum, I am not sure if this topic even belongs here, or rather in the test gear section. I just assumed that there is a high probability the voltnuts would know something like that. If this is the wrong place to ask my question, please move the thread. 

[bd139]

This has been a problem for a number of years. It makes some things quite difficult for example high voltage probes and RF probes as they are calibrated to a constant meter impedance.

This was one reason I bought a bench meter from GW recently; that has constant impedance. You can configure the impedance on low ranges as well. My Fluke 77 got replaced with an old Fluke 8024B about 10 years ago for the same reason.

[The Soulman]

Here is data from a Fluke 175 (measured with another 175):

DC:

600 mV 3,875 M (d.u.t. reads OL, input could be clamped down)
6 V  11,11 M
60 V 10,11 M
600 V 10,01 M
1000V 10,01 M

Input resistance could vary with applied voltage, more investigation needed.

AC:

All ranges read OL i.e. larger than 60 M at DC, I should do a input impedance measurement over entire audio range. 

[glarsson]

The Fluke 8846A is specified to have 10Mohm on all DC ranges (can select >1Gohm on lower ranges).
The AC ranges are specified to have 1Mohm +/- 2% (!).
My meter fails the AC impedance -- it measures 10Mohm on all AC ranges. Nice failure mode. :-)

[HighVoltage]

I have always found that the Keysight bench DMM are pretty stable in their resistance over the voltage rages.
That is why I sometimes use these meters as one resistor in a divider to measure high voltages.

These two meters I just had on my desk, so I measured them with each other...

Keysight 34461A (Measured with a Keysight 34410A
100mV:  10.049,73 MOhm
1V:         10.049,74 MOhm
10V:       10.049,75 MOhm
100V:     10.049,77 MOhm
1000V:   10.049,90 MOhm


Keysight 34410A (Measured with a Keysight 34461A
100mV:  10.022,02 MOhm
1V:         10.022,02 MOhm
10V:       10.022,02 MOhm
100V:     10.024,01 MOhm
1000V:   10.024,01 MOhm

[Jay_Diddy_B]

Hi,
The non-constant input resistance has come up before in a discussion on some Caddock resistor networks in this thread:

https://www.eevblog.com/forum/projects/caddock-1776-c48-odd-resistor-dividor/

Here is a reply that I made in that thread:

Hi,

Those resistors values can be used to make divide by 10, 100, 1000 and 10000 with switches to ground:




The conventional divider requires floating switches.


So the Caddock 1776-C48 is probably easier to use in auto-ranging DMMs.

Jay_Diddy_B




You can see that the input impedance would not be constant with this resistor network.

The Fluke 289 probably has a similar network.


If you are using a HV probe like the Fluke 80K-40 you have to check your meter.

Regards,

Jay_Diddy_B

[e61_phil]

The Fluke 8846A is specified to have 10Mohm on all DC ranges (can select >1Gohm on lower ranges).
The AC ranges are specified to have 1Mohm +/- 2% (!).
My meter fails the AC impedance -- it measures 10Mohm on all AC ranges. Nice failure mode. :-)

Is 1Meg specified for DC? Perhaps they assumed a certain frequency and took the input capacity also into account.

I have always found that the Keysight bench DMM are pretty stable in their resistance over the voltage rages.
That is why I sometimes use these meters as one resistor in a divider to measure high voltages.

I also tried that last year (https://www.eevblog.com/forum/metrology/measuring-10kv-with-a-34401a/), but it doesn't work as I expected.


Most of the benchtop multimeters I know, do not switch the attenuator. There is often a 1:100 attenuator and in the 100V Range it is used togehter with the 10x amplifier (1V mode) and in the 1000V range it is used with 1x amplification (10V range). Therefore, the input impedance should be stable over the ranges.

[glarsson]

Is 1Meg specified for DC? Perhaps they assumed a certain frequency and took the input capacity also into account.

Fluke specifies 1Mohm +/- 2% (shunted by <100pF) between 3Hz and 300kHz.

[FlexibleMammoth]

Thanks to Jay_Diddy_B for the schematic suggestion. I will update my readings later with more precise values (Keithley 2015, calibrated), though I'm pretty sure 5VDC was 11MOhm spot on. Dont get me wrong, I love my 289. I would not even be mad if they were upfront about it. But stating information in datasheets and the manual which is imprecise at best and then telling me otherwise once I buy a $300 accessory to a $800 meter is kind of disappointing.

Note: apparently, it's even worse with Agilent meters, which have 5MOhm except for a reeeeally tiny range around zero: http://brianhoskins.uk/agilent-u1253b-input-impedance-problem/

I have always found that the Keysight bench DMM are pretty stable in their resistance over the voltage rages.
That is why I sometimes use these meters as one resistor in a divider to measure high voltages.

I also tried that last year (https://www.eevblog.com/forum/metrology/measuring-10kv-with-a-34401a/), but it doesn't work as I expected.


Most of the benchtop multimeters I know, do not switch the attenuator. There is often a 1:100 attenuator and in the 100V Range it is used togehter with the 10x amplifier (1V mode) and in the 1000V range it is used with 1x amplification (10V range). Therefore, the input impedance should be stable over the ranges.

Stupid question: Why is this not possible in handheld multimeters? As bd193 stated, there seems to be a GW multimeter that has constant input impedance.

[Jay_Diddy_B]

Hi FlexibleMammoth and the group,

Please don't miss-understand. The resistor network that I showed is not tied to a specific meter, Fluke or otherwise. It is an example of a resistor network produced by Caddock for use in some multi-range DVMs.

The measurements you made on your meter are probably correct.

Jay_Diddy_B

[e61_phil]

I took a Fluke 80K-40 and measured the output impedance without any source. As one could expect the resistance is 1.1Meg. Hence, if you use a meter with 11Meg instead of 1Meg the error in the divider ratio is only ~1%.

I wondered if the 80K-40 is good enough to care about 1%. Therefore, I took my DIY 1000:1 divider (which should be much better than 100ppm accurate) and hooked it up to a calibrated 3458A. In parallel I connected a Fluke 80K-40 and hooked it to a calibrated Keysight 34470A. As voltage Source I used a FuG high voltage supply which is specified with 0,2% at 12.5kV. I made a measurement every 100V.

As you can see in the attachment the Fluke 80K-40 is much better than I expected.

[e61_phil]

And a cheaper Testec TT-HVP 40 one..

[bd139]

Thanks for the TT-HVP40 one. I just bought one of them

[FlexibleMammoth]

Just compared the Testec to the Fluke HV probe. They seem pretty identical, with a slightly worse spec on the Testec. Yours performs WAY above spec 

Regarding "only 1% deviation": Fluke tells you in the manual to multiply by 0.99 for the 5V range, which is supposed to correct that. Of course I could just ignore that, however, it would double measurement error.

anyone with a 289 or 287: Can you verify impedance in V/AC+DC vs. mV/AC+DC mode? The first is fine, but In the latter, my HV probe reads significantly low...

[David Hess]

I have a Fluke 289 and had a close look at the impedances, and found that the impedance changes as a function of the range:

This is typical of cheap digital multimeters.

My Tektronix DMM916 is 10 megohms on all of the ranges except the 4 volt range which is 11 megohms.  My Beckman RMS225 is inexplicably 11 megohms on all ranges but its specifications say 10 megohms; it is accurate so I doubt it is damaged.

When using a high voltage probe, then this is just something to to be aware of.

Note that measurements of the input resistance can sometimes be corrupted by charge injection.

My question at this point (because google did not help):  Why is that the case?

They do this because it simplifies automatic range switching of the input divider as Jay_Diddy_B showed.  Maybe a constant input resistance should be considered as an advantage of manual ranging multimeters.

- If it is a bug, I do not understand how a company that has been building handheld multimeters for decades cannot overcome this problem.

The only bug here is Fluke attempting to conceal what they are doing.  If I was looking to buy another multimeter, then a constant 10 megohm input resistance is one of the features I would consider but manufacturers make this difficult:

Nessus: Antimatter can disintegrate a General Products hull entirely.
Louis Wu: I didn't know that.
Nessus: It is not widely advertised.

- Saying "just multiply by 0.99 sometimes and you'll be fine lol" is a bit weak for a $1000 equipment combination in my opinion.

But everybody is doing it!

I wondered if the 80K-40 is good enough to care about 1%. Therefore, I took my DIY 1000:1 divider (which should be much better than 100ppm accurate) and hooked it up to a calibrated 3458A. In parallel I connected a Fluke 80K-40 and hooked it to a calibrated Keysight 34470A. As voltage Source I used a FuG high voltage supply which is specified with 0,2% at 12.5kV. I made a measurement every 100V.

As you can see in the attachment the Fluke 80K-40 is much better than I expected.

I did the same thing with my Fluke 80K-6 with similar results.  I ended up calibrating it against a specific meter and range immediately before making the unknown high voltage measurement.

[e61_phil]

I did the same thing with my Fluke 80K-6 with similar results.  I ended up calibrating it against a specific meter and range immediately before making the unknown high voltage measurement.

We use such probes for HV "tests" only and not for accurate measurements. But they seem to be better than I expected. Perhaps one should measure a few more of these probes to get some statistics.

[David Hess]

I did the same thing with my Fluke 80K-6 with similar results.  I ended up calibrating it against a specific meter and range immediately before making the unknown high voltage measurement.

We use such probes for HV "tests" only and not for accurate measurements. But they seem to be better than I expected. Perhaps one should measure a few more of these probes to get some statistics.

You have to cut the coat to fit the cloth.  If I wanted better precision than the probe would provide, I would do something else.  Part of doing the calibration was just to make sure the probe was functioning correctly.

[David Hess]

This discussion prompted me to do a detailed search for meters which I might be interested in.  This includes only the B&K and Amprobe models which have the 3/4" spacing banana plug arrangement that I like.

All of B&K's autoranging multimeters have three, or in one case two, DC input resistances depending on input voltage range.  Their manual ranging meters are 10 megohms on every range.

Amprobe has two autoranging meters with multiple input resistances.  All of their other autoranging meters and their manual ranging meters are specified to have a 10 megohm input resistance except for one really low end manual meter which is 1 megohm.  I am a little dubious that most of Amprobe's automatic ranging meters have a uniform 10 megohm input resistance so this is something I would want to check if I bought one.

[e61_phil]

anyone with a 289 or 287: Can you verify impedance in V/AC+DC vs. mV/AC+DC mode? The first is fine, but In the latter, my HV probe reads significantly low...

I measured the input impedance of a Fluke 289. To not run into OL I used a 20Meg precision resistor in series and applied a voltage from the calibrator.

50mV  Range: 10.010 Meg
500mV Range: 10.009 Meg
50mV AC+DC: 10.907 Meg
500mV AC+DC: 10.020 Meg

The last digit isn't very stable. Therefore, I think the ranges without AC+DC have the same impedance. All measurements were taken at FS of the range.

[e61_phil]

Just compared the Testec to the Fluke HV probe. They seem pretty identical, with a slightly worse spec on the Testec. Yours performs WAY above spec 

I measured another two Testec Probes and they showed the same performance.

[e61_phil]

I found another two Fluke probes and measured them for fun...

[HighVoltage]

Hello e61_phil

Thanks for these nice plots.
Can you show us a picture of your test setup?

[e61_phil]

Sorry, I'm not allowed to post photos.

The setup isn't very special. A 3458A, a 34470A and a FuG high voltage supply (HCP series with high stability and low noise options). Everything is connected with coaxial high voltage cables and 10kV-SHV Plugs. The HV probes are simply connected to a wire which is soldered to a 10kV-SHV jack.

The DIY 1000:1 divider is built up of 20 selected Caddock USF 371 resistors and a 100k Vishay S102 (or Z201?)+some Rhopoint PWW for fine tunning. Worst case uncertainty calculation gives about 35ppm at 10kV, but it seems to be much better.

[FlexibleMammoth]

I'm genuinely amazed by the precision of the Testec probes. With fluke, it seems some kind of lottery - you may or may not get lucky...

Thank you so much for the 50mv/500mV measurements. I do not have the means to do these at home, probably need to take my meter to the lab at university. Now I have something to compare to! 

[HighVoltage]

The DIY 1000:1 divider is built up of 20 selected Caddock USF 371 resistors.

Wow... they are 30 Euro a piece !
Is that correct?

[FlexibleMammoth]

The DIY 1000:1 divider is built up of 20 selected Caddock USF 371 resistors.

Wow... they are 30 Euro a piece !
Is that correct?

Well as he can't post pictures, this is a cal lab/workplace/whatever. I doubt they care if they can afford both a 3458A and a 34470A   

[HighVoltage]

Wow... they are 30 Euro a piece !
Is that correct?

Well as he can't post pictures, this is a cal lab/workplace/whatever. I doubt they care if they can afford both a 3458A and a 34470A   
[/quote]
Well, I also have a 3458A and a 34470A and know about expensive equipment. Somehow I just did not expect this high price for this resistor. But I guess one has to spend the money to get this precision divider.

I have a few Tektronix high voltage probes P6015A and a couple Fluke 80K-40.
It would be nice to have such a precision divider to check the calibration of the probes.

 

[FlexibleMammoth]

Well, I also have a 3458A and a 34470A and know about expensive equipment. Somehow I just did not expect this high price for this resistor. But I guess one has to spend the money to get this precision divider.

Mark me down as impressed and envious :-)

It would be nice to have such a precision divider to check the calibration of the probes.

You're not the only one there!
Is 30 bucks for factory-preselected resistors? If not, that setup might have cost even more...

[fcb]

Sorry, I'm not allowed to post photos.

The setup isn't very special. A 3458A, a 34470A and a FuG high voltage supply (HCP series with high stability and low noise options). Everything is connected with coaxial high voltage cables and 10kV-SHV Plugs. The HV probes are simply connected to a wire which is soldered to a 10kV-SHV jack.

The DIY 1000:1 divider is built up of 20 selected Caddock USF 371 resistors and a 100k Vishay S102 (or Z201?)+some Rhopoint PWW for fine tunning. Worst case uncertainty calculation gives about 35ppm at 10kV, but it seems to be much better.

Such a shame about the pix.

I looked up FuG - their gear looks amazing. Very cool that they can build a 100+ person company making that stuff.

[HighVoltage]

I looked up FuG - their gear looks amazing. Very cool that they can build a 100+ person company making that stuff.

Yes, FUG is an amazingly good company. And when I repaired my 12.5 kV unit, I got all the support I needed, although this instrument is out of production for a long time.

[emg]

thanks for your answer.
My HIOKI clamp meter and UNI-T DDMs seems to use similar divider therefore having variable input resistance. 

[EC8010]

I just checked my Fluke 89 IV direct voltage ranges and got 10.991M, 10.092M, 10.002M, 10.002M.