Multimeter input impedance riddle

[alexwhittemore]

Here's a riddle for the masses (which I don't know the answer to). I have 3 meters, a HP 3478a bench meter, a U1253A handheld meter, and a Fluke 87-V. I'm curious to measure the input impedance of all of them (of course, I know that the agilent is 10Meg, the Fluke is 10Meg unless you select Hi-Z mode and the mV range, and the HP is 10Meg in the 30V and above ranges, but in the gigaohms in the lower ranges).

To measure A's impedance, I need to use meter B. But B sources some particular test current, which will present to A a voltage, thus potentially changing A's range automatically.

Let's start with the Agilent in ohms measuring the HP. The Agilent sources 1.85V in its highest ohms range, which leaves the HP in its 3V range, which is very high impedance (>500Mohm, which is all the agilent can tell me). However if I manually range up on the HP to the 30V range, I see the voltage sourced by the Agilent drops to .93 or so, and it shows 10Mohm on the nose. So far so good.

Let's swap. With the agilent on volts, the HP sources 1.14ish volts and reads 11odd Mohm. So far so good.

Now, the Agilent doesn't have a high-Z mode, but the Fluke does. If I plug them together, Agilent on ohms and flip Fluke to mV, I see 10Meg on the nose (11.1Meg in Volts mode). If I turn the fluke to mV while holding the Hz/% key (to put it in High-Z), both meters show overload (the Fluke because the Agilent is sourcing >600mV, the Agilent because the Fluke's impedance is >500Mohm). So again, so far so good.

Now the fun is when I try the same measurements Fluke vs HP.

With the Fluke in mV High-Z and the HP autoranging in Ohms, the Fluke says OL but the HP says 27Mohm. According to the Agilent (and anecdotes about what the real value is for the Fluke's high-z), the measured value should easily be overload, >500Mohm. What if I flip it around? Fluke in Ohms, HP in volts, HP shows 2.6 while the fluke shows 10Mohm [sidenote: this 2.6 is JUST below the hysteresis mark for the HP to range down from 30V to 3V, except as soon as it does, the impedance shoots up and thus the measured voltage with it, kicking it back to 30V range. This happens very quickly, and makes a chirping noise in the HP when it does, so if autoranging is on these two machines will just sit there flopping back and forth, chirping once or twice a second. Neat!]. 10Mohm on the 30V range is correct, so what if I range down to 3V? Suddenly the Fluke reads 27.3Mohm! That's not >Gohm at all!

In short, when measuring Agilent vs HP or vs Fluke, everything is completely normal. But when measuring Fluke v HP or vice versa, suddenly what should be many Gohms (well, overload ohms anyway) is appearing as 27Mohms, on both meters! What gives?! When the HP is in autoranging Ohms measuring the Fluke in mV or mV-HighZ, it shows 26Mohms as well! Oddly, with the Fluke mV (not high-z) range in overload, the HP seems to think it drops to around 3Mohms rather than 10M. Not sure what that is, either.

So the riddle is, why does impedance on the fluke seem to drop when the rang is in OL, and where is this magical 27Mohm coming from *on both meters* when I think I should be getting >Gohms?

[David Hess]

I think your anomalous input resistance measurements are coming from the input protection networks when you are outside the input range of the high input resistance range that you are on.

I usually check the 10M input resistance using an external 10M resistor which should divided the input reference voltage by half so the meter effectively measures its own input resistance.  High input resistance modes are just specified to be greater than some number so they need a more sophisticated and non-trivial test.

[Mr. Coffee]

...I usually check the 10M input resistance using an external 10M resistor which should divided the input reference voltage by half so the meter effectively measures its own input resistance...

That.

[alexwhittemore]

I think your anomalous input resistance measurements are coming from the input protection networks when you are outside the input range of the high input resistance range that you are on.

I'd be more inclined to believe you if the mV range (non high-z) didn't show 10Mohm even in overload. HMM. Wait. Sometimes it's 10Mohm. But depending on what mode I was in previously, it might be 2.7odd Mohm. The manual does mention something about, while making DC offset measurements of AC signals, they'll be more accurate if you measure AC first then go to DC from that mode, disabling some input protection. Or something - maybe I have that backwards. But for sure, if I switch the meter on straight to mV, it's 10Mohm, then if I go to V, it's 11Mohm, then if I go back to mV, suddenly it's 3Mohm. Except not the time I did it just now, it's been consistently 10.

Anyway, the Agilent ohms range overloads the Fluke mV range just like the HP does, yet IT measures >>30Mohm. So I'm pretty sure I don't buy the input protection circuitry theory.

I usually check the 10M input resistance using an external 10M resistor which should divided the input reference voltage by half so the meter effectively measures its own input resistance.  High input resistance modes are just specified to be greater than some number so they need a more sophisticated and non-trivial test.

That's certainly interesting, I'll try it!

[robrenz]

AC impedance is another aspect as I posted in another thread:

I just measured my 8846A on DC ranges using my 87 and I get the expected 10M. I measure it with the 87 on AC ranges and get 10M but the specs say 1M on AC inputs. I measure the AC ranges using my LCR meter so it is a AC excitation instead of DC and I get the expected  1M.  AFIK this makes sense since the 8846A can measure AC and DC simultaneously.

[alexwhittemore]

AC impedance is another aspect as I posted in another thread:

I just measured my 8846A on DC ranges using my 87 and I get the expected 10M. I measure it with the 87 on AC ranges and get 10M but the specs say 1M on AC inputs. I measure the AC ranges using my LCR meter so it is a AC excitation instead of DC and I get the expected  1M.  AFIK this makes sense since the 8846A can measure AC and DC simultaneously.

Certainly interesting, though I'm talking exclusively about DC measurement and excitation.

[David Hess]

I think your anomalous input resistance measurements are coming from the input protection networks when you are outside the input range of the high input resistance range that you are on.

I'd be more inclined to believe you if the mV range (non high-z) didn't show 10Mohm even in overload. HMM. Wait. Sometimes it's 10Mohm. But depending on what mode I was in previously, it might be 2.7odd Mohm. The manual does mention something about, while making DC offset measurements of AC signals, they'll be more accurate if you measure AC first then go to DC from that mode, disabling some input protection. Or something - maybe I have that backwards. But for sure, if I switch the meter on straight to mV, it's 10Mohm, then if I go to V, it's 11Mohm, then if I go back to mV, suddenly it's 3Mohm. Except not the time I did it just now, it's been consistently 10.

Anyway, the Agilent ohms range overloads the Fluke mV range just like the HP does, yet IT measures >>30Mohm. So I'm pretty sure I don't buy the input protection circuitry theory.

I have never seen a meter that retained any state information between ranges.  I am inclined to think something is wrong if the input resistance is changing on the same range.

Can you try the tests in manual ranging mode?

[alexwhittemore]

I'd have agreed with you until I saw that bit in the manual. But now that I reread it, maybe it's not as explicit as I made it out to sound:

When measuring voltage, the Meter acts approximately like a 10 M? (10,000,000 ?) impedance in parallel with the circuit. This loading effect can cause measurement errors in high-impedance circuits. In most cases, the error is negligible (0.1% or less) if the circuit impedance is
10 k? (10,000 ?) or less.
For better accuracy when measuring the dc offset of an ac voltage, measure the ac voltage first. Note the ac voltage range, then manually select a dc voltage range equal to or higher than the ac range. This procedure improves the accuracy of the dc measurement by ensuring that the input protection circuits are not activated.

I'm not QUITE sure what they're getting at there.

As for testing manual range, I assume you mean on the Fluke as the target of the ohms measurement? In HighZ, it's already manual ranging only, and I can't change range when it's already in overload on the highest.

[David Hess]

I'd have agreed with you until I saw that bit in the manual. But now that I reread it, maybe it's not as explicit as I made it out to sound:

When measuring voltage, the Meter acts approximately like a 10 M? (10,000,000 ?) impedance in parallel with the circuit. This loading effect can cause measurement errors in high-impedance circuits. In most cases, the error is negligible (0.1% or less) if the circuit impedance is
10 k? (10,000 ?) or less.
For better accuracy when measuring the dc offset of an ac voltage, measure the ac voltage first. Note the ac voltage range, then manually select a dc voltage range equal to or higher than the ac range. This procedure improves the accuracy of the dc measurement by ensuring that the input protection circuits are not activated.

I'm not QUITE sure what they're getting at there.

That sounds like precharge of either the AC coupling capacitor or the bootstrap voltage of the input protection circuits.  Oscilloscopes vertical inputs often include a "precharge" setting for their AC input coupling capacitor.

For high input resistance multimeters, the input protection circuits are a special problem because they add significant leakage to the input which would not be a problem at 10 megohms.  The solution is to bootstrap them (and often the input amplifier as well) so they follow the input within the safe input voltage range and then clamp outside of it.

[KedasProbe]

I'd have agreed with you until I saw that bit in the manual. But now that I reread it, maybe it's not as explicit as I made it out to sound:

When measuring voltage, the Meter acts approximately like a 10 M? (10,000,000 ?) impedance in parallel with the circuit. This loading effect can cause measurement errors in high-impedance circuits. In most cases, the error is negligible (0.1% or less) if the circuit impedance is
10 k? (10,000 ?) or less.
For better accuracy when measuring the dc offset of an ac voltage, measure the ac voltage first. Note the ac voltage range, then manually select a dc voltage range equal to or higher than the ac range. This procedure improves the accuracy of the dc measurement by ensuring that the input protection circuits are not activated.

I'm not QUITE sure what they're getting at there.

The first part is just telling you that if you measure you influence what you measure.
The second part is telling you to choose the DC range wisely, the whole signal (AC+DC) must fit in the ADC range to get a correct measurement, if it clips on the peaks your measurement will be wrong, hence the advice to check AC first not for the meter to act differently but to inform you about the signal you measure.

There is no point of trying to measure resistance of a meter in overload, it will just clamp and crew up the measurement, even if it doesn't show overload it is possible it is partially clipped and gives you a wrong value. So put it on your scope see how big the signal is and then decide what the minimum voltage range is you can measure. (just putting the device in series with a resistor with a DC source makes it much easier but again don't overload)

[alexwhittemore]

There is no point of trying to measure resistance of a meter in overload, it will just clamp and crew up the measurement, even if it doesn't show overload it is possible it is partially clipped and gives you a wrong value.

I mean, it can't clamp. If my meter clamped hundreds of volts of input to the 600mV full scale of the mV highest range, presumably it'd, you know, explode. Of course, I don't know enough about input protection circuitry to know if there'd be a measurable change in input impedance in overload vs not. On the one hand, it seems reasonable there would be. On the other hand, for the input impedance to DROP SUBSTANTIALLY in overload would be a bit psychotic, no?

So put it on your scope see how big the signal is and then decide what the minimum voltage range is you can measure. (just putting the device in series with a resistor with a DC source makes it much easier but again don't overload)

Again, I'm not personally measuring AC anything. I only brought up that quote in the manual because it alluded to some interaction between different ranges, but your explanation seems good: it's not that there's saved state between ranges, the point is that if *part* of your waveform is getting clipped by input protection circuitry, your DC offset measurement might still be incorrect even though the DC value is within the range.

[SeanB]

Typically a DMM input impedance is 10M or so but in overload it drops to whatever the input protection resistance is ( normally a 1k resistor plus a PTC thermistor to limit current) with a voltage clamped to either a zener diode or the battery voltage plus a diode drop or two. Some it will be 6V or 8V, but the current is a lot more than what would normally flow, but it's exact value is not well defined, it depends on the overload voltage and time as the PTC heats up.

[alexwhittemore]

Oh right: except the other problem I have with the "input protection decreases impedance in OL" theory is that my Agilent on ohms overloads the Fluke, yet still measures OL itself (i.e. >500Mohm). Which is to say, at least by the amount my Agilent OLs the Fluke, the impedance is still >> 30Mohm.

Specifically, compare the attached three photos.

In the first, we see the agilent measuring the volts range on 83||87, forcing .663 volts and measuring 5.55 Mohm (11Mohm||11Mohm)
In the second, we see the agilent measuring the volts on the 83 (11Mohm) || overloaded mV on the 87 (high z) to be 11.096.
In the third, we see the agilent measuring volts on the 83 (11Mohm), but we've removed the 87. Note A. that the 87 is still overloaded, because in highZ the charge doesn't bleed off the input device quickly, implying that the actual system impedance is still pretty high, and B. that the measurement on the Agilent only dropped by .002Mohm. This implies that the 87 in highZ, even overloaded, is 61.5Gohm, which seems about right though in fact that's within the accuracy noise of the Agilent (I think it's ±2ct, anyway...)

[alexwhittemore]

Redoing the measurement above with the HP instead of the Agilent, the first picture is nominally the same: .539V test voltage and 5.266Mohm measured. When I switch the 87 to high-z mV, it overloads, but the HP measures 11.114Mohm, nominally the same to the second photo (albeit with a test voltage of 1.14ish). Remove the 87, I see 11.113Mohm, again, nominally the same as the third photo.

The oddness actually only strikes when I manually range the HP. If I manually downrange from 30Mohm FS to 3Mohm FS, the measurement suddenly becomes:
83: 2.86V, 87 OL mV (highZ), HP: 2.81Mohm. If I then remove the 87 by unplugging, the HP suddenly sees OL Mohm at 4V test (again, recall the HP is in the 3Mohm range, and the 83 is presenting 11ish Mohm, so this is right).

In this last measurement, it's almost as if the Fluke IS clamping, sinking at least some amount of current. And indeed, it is: when I hook up the agilent between the 87 and the HP, in EITHER high-Z OR normal mode, in OL mV, I measure it sinking .8µA of current at 2.8V test, which implies a resistance on the 87 of 3.5ish Mohm. If I manually range up to 30Mohm FS, the measured resistance simply increases by a factor of 10, and the current into the fluke decreases by a factor of 10 (that is, without the 83 in the loop, the 87 current goes from .1µA to .01µA).

So it looks like the conclusion is: the HP measures odd and incorrect values near the top of any given range because it uses such a high test voltage at the top of its range that the fluke clamps such that the measurement is always JUST below the top of the HP range.

Now my question is, if the 87 actually is clamping the input, how on earth could it safely survive 1000V on the mV range?

[David Hess]

I did some quick measurements with the multimeters I had laying around and noticed that despite a specified and measured 10 megohm input resistance for all of them (I sometimes use a x1000 high voltage probe so it is important that I know my multimeters have accurate 10 megohm input resistances.), some combinations reported about 11 megohms with various levels of drift and some were right on at 10 megohms.  I think you may be seeing the effect of input charge pumping from the ADC in the digital multimeters corrupting the ohms measurements.  My older digital multimeters which use a JFET input buffer, manually switched attenuator, and voltage-to-frequency based analog to digital converter did not display this behavior.

I tried finding links to "charge pumping" and "charge injection" in the context of digital multimeters but came up empty so this may be another case of James Burke's "if it is not on the internet, then it does not exist".

[alexwhittemore]

Sounds interesting!

Surely there's some analog engineer from Fluke or Agilent or Keithley on the forum who can shed some light on what aspects of the input circuitry might react oddly to certain test conditions, no?