EEVblog #584 - What Effect Does Your Multimeter Input Impedance Have?

[EEVblog]

What effect does your multimeter input impedance have on the circuit you are measuring? Dave shows a practical example of how it can really screw things up if you aren't watching out for it.

[caroper]

Hi Dave,

]Given that you have a Precision High Impedance Voltage source I wondered why you did the Calibration the way you did rather than dialing in a Voltage on the Keathley and then adjusting the POT until the LED extinguished.

Would that have been less accurate or just a different approach to the same task?

Cheers
Chris

[EEVblog]

Would that have been less accurate or just a different approach to the same task?

Just a different approach to the same task.
In practice, few people would have access to a 6 digit voltage source like I have, their meter would generally be the best and most accurate thing in their arsenal, so that's what normally would get used in such a situation.

[caroper]

Thanks Dave,

I realised that the Video was ment to be, and indeed was instructional, I just wondered if there would have been anything wrong in taking the other approach under other circumstances.

Cheers
Chris

[kaushleshchandel]

Very nice!

I am building a Kelvin Varley Divider using precision resistances. But I have seen the readings on Agilent 6.5 Digit meter not show as per the resistance divider set.... Whats happening is that when I calibrate the lowest three KVD strings, its fine... but the moment I go the the fourth & fifth string, the readings start behaving very different than what I would expect.

Ill go and Check if this is due to Impedance setting on Agilent meter.

btw, here is how I am trying to build the KVD.  http://conradhoffman.com/mini_metro_lab.html

And I bought 100 pieces of this from ebay http://www.ebay.com/itm/1x-5K0000-Vishay-VSR-Series-Bulk-Metal-Foil-Precision-Resistors-0-01-5K-/121032384194?pt=LH_DefaultDomain_0&hash=item1c2e179ec2

I have grouped them into 0.0001% range using a Bridge.

[opablo]

I loved this one !... I too say to myself... 10Meg... that's a lot... and never consider the impedance of the circuit under meassurement ja...

You made my mind constantly think in of this:

http://en.wikipedia.org/wiki/Observer_effect_(physics)

In science, the term observer effect refers to changes that the act of observation will make on a phenomenon being observed. ......
The observer effect on a physical process can often be reduced to insignificance by using better instruments or observation techniques.
Historically, the observer effect has been confused with the uncertainty principle.

"The observer effect ... can often be reduced to insignificance by using better instruments." tell that to a quantum physicist and he will start to cry like a baby 

[free_electron]

And it's not only multimeters ! Scope probes have the same effect ! They are either 1 meg or 10 meg input plus they have a few pF capacitance. When looking at ac signals with sufficiently high frequency that comes into play. Probe impedance can collapse depending on frequency !
Good probe makers will give you the bandcurve.

Smart probes (active) have an eeprom on board with the curve in it. The scope reads it and adjusts.

[Dr. Frank]

Nice practical video.

Some additional formula / calculations on the Whiteboard would have made a nice Fundamental Friday video, also.

That would also have shown this well known quantitative rule of thumb:

The introduced error can be calculated by dividing the source resistance of the DUT by  the multimeters resistance, in this case 0.02%. (source resistance in this case is given by those five 10k resistors in parallel)

Et voilá, the reason, why such high precision measurements need high impedance DMMs, would have been quite obvious, also for "young players".

regards from an old geezer - Frank

[juani_c]

Noob question:
¿Is necessary to use a high precision voltage reference if you are going to tweak/calibrate the values whit pots?

[DrMag]

Just thinking... for those of us without a fancy multimeter with multiple input impedances, is it possible to externally accomplish something similar by using a high-impedance op amp in a follower configuration?

[edpalmer42]

I found this video frustrating for two reasons.

First, as has been mentioned, Dave measured the voltage across the pot instead of setting the input voltage and then adjusting the pot to extinguish the LED.  The input voltage is the important parameter so it should be the thing used to make the adjustment.  Doing it this way would automatically include any and all offset voltages and component tolerances.  You don't need a precision voltage source, you use the same meter to set the input voltage that you were going to use to set the pot wiper voltage.

Second, Dave didn't explain how to solve this problem.  If I did have to make the measurement on the pot wiper, I wouldn't use any active circuitry, I'd set up a bridge measurement with an external voltage source and a ten-turn pot to create an infinite-impedance null voltmeter.

Ed

[jippie]

Why does a manufacturer like Agilent provide a 10M input impedance, if they're able to provide a 2G+ input impedance? What's the catch that I'm overlooking?

[c4757p]

Noob question:
¿Is necessary to use a high precision voltage reference if you are going to tweak/calibrate the values whit pots?

Yes, because he designed it so the pots only adjust over a tiny range. The voltage reference has to start off somewhere inside that range.

[robrenz]

Also higher precision reference will usually carry a higher stability.  No point in precisely tweaking a value only to have it drift to a value out of tolerance a week later.

[edpalmer42]

Why does a manufacturer like Agilent provide a 10M input impedance, if they're able to provide a 2G+ input impedance? What's the catch that I'm overlooking?

As Dave mentioned in the video, the high impedance is only on the lower voltage ranges.  Each meter's design is different so you have to check the specs to see where the impedance changes from high to 10M.  Also, if you are using a voltage divider probe to measure voltages higher than the meter's maximum, you want to have an industry standard 10M impedance so that the ratio works out and you definitely don't want the impedance changing if the meter happens to autorange.

Ed

[casinada]

Well, Ideal voltmeters have infinite internal resistance and ideal ampmeters have zero resistance but real devices don't that's why Dave designed a microcurrent device to overcome the burden problems at low currents. Multimeters and scopes have defined input impedances so when you plug probes designed for them they still show the right values. For example High voltage probes for multimeters or scopes. This is a great simplification. The point is to know what you're doing with your test equipment and know the limitations.

[EEVblog]

Also, if you are using a voltage divider probe to measure voltages higher than the meter's maximum, you want to have an industry standard 10M impedance so that the ratio works out and you definitely don't want the impedance changing if the meter happens to autorange.

Yes, and some meter that say they are 10M, can be oddball values like 11M or something that can change with the range, that 10M might just be nominal value. A trap if using external divider probes.

[EEVblog]

¿Is necessary to use a high precision voltage reference if you are going to tweak/calibrate the values whit pots?

No. But higher precision references usually (not always) have a lower tempco. In this case I'm using this reference on another board, so I just re-used it in this design.

[EEVblog]

First, as has been mentioned, Dave measured the voltage across the pot instead of setting the input voltage and then adjusting the pot to extinguish the LED.  The input voltage is the important parameter so it should be the thing used to make the adjustment. Doing it this way would automatically include any and all offset voltages and component tolerances.  You don't need a precision voltage source, you use the same meter to set the input voltage that you were going to use to set the pot wiper voltage.

No, you don't need a precision voltage source, but you do need to rig up suitable small range high resolution adjustment pot or other source on the input to do that. That's extra work. In that case it's just easier to use the high impedance meter directly as I did. Oh, and if it's a high impedance source you've got driving it, beware of any input current...

Second, Dave didn't explain how to solve this problem.  If I did have to make the measurement on the pot wiper, I wouldn't use any active circuitry, I'd set up a bridge measurement with an external voltage source and a ten-turn pot to create an infinite-impedance null voltmeter.

Why? A high impedance voltmeter does the job just fine.

The only point of the video was to show the effect of input impedance on a circuit. But if seems a few people are nit picking that I used a bad example, or should have expanded the video to include different ways to do this adjustment in this instance. But that wasn't the point. Oh well.

[edpalmer42]

The only point of the video was to show the effect of input impedance on a circuit. But if seems a few people are nit picking that I used a bad example, or should have expanded the video to include different ways to do this adjustment in this instance. But that wasn't the point. Oh well.

I hear you Dave.  It's kinda like a different form of 'feature creep'.  Is this a big enough topic to be a 'Fundamental Friday' video?  Maybe expand it to include scope probes and ammeters and cover both the problem and various solutions.

Ed

[AlphZeta]

Of course, I suppose you could always use an extra OPA2376 as voltage followers after the voltage divider so that the impedance of the meter does not mater any more. Especially, OPA2376 is relatively inexpensive. 

[Zucca]

Confused young player here... (It's funny how a master of science can be turned in nothing here at the EEVBlog, experience is everything and I don't have it...)

Second, Dave didn't explain how to solve this problem.  If I did have to make the measurement on the pot wiper, I wouldn't use any active circuitry, I'd set up a bridge measurement with an external voltage source and a ten-turn pot to create an infinite-impedance null voltmeter.

Ed

Man, I want to understand this. Are you talking about a Wheatstone bridge or something similar? If not can someone point me please in the right direction?

Moreover the low impedance in parallel to the pot for the stability jazz is because in every pot the total resistance (across the non wiper terminals) is changing a little by moving the wiper?

[edpalmer42]

Confused young player here... (It's funny how a master of science can be turned in nothing here at the EEVBlog, experience is everything and I don't have it...)

Second, Dave didn't explain how to solve this problem.  If I did have to make the measurement on the pot wiper, I wouldn't use any active circuitry, I'd set up a bridge measurement with an external voltage source and a ten-turn pot to create an infinite-impedance null voltmeter.

Ed

Man, I want to understand this. Are you talking about a Wheatstone bridge or something similar? If not can someone point me please in the right direction?

Moreover the low impedance in parallel to the pot for the stability jazz is because in every pot the total resistance (across the non wiper terminals) is changing a little by moving the wiper?

I stole this diagram from http://www.allaboutcircuits.com/vol_6/chpt_3/13.html



The null meter shown could be any ua or na meter, or it could be any DVM that has 'only' a 10M input impedance.  When you adjust the pot (which can be almost any value) to zero the reading on the meter, you are at a point where no current is flowing, i.e. a point of infinite impedance.  The reading on the voltmeter on the right is now identical to the value at the midpoint of the resistor string but there is no loading on that midpoint.  Do the calculation and you'll find that a simple 10Mohm, 200mv, 3.5 digit meter has a least significant digit that represents a load of only 10 pa.  For Dave's 1V measurement, that is equivalent to 100 Gohms.  And yes, noise and a host of other low-level effects can easily become an issue in measurements like this.

This is a very, very old idea.  I don't know how old, but probably over 100 years.  It dates from the days before vacuum tubes (valves) when the meters had very low sensitivity and tricks like this were needed to make ANY measurements.  Today, it still has value when you're making measurements in high impedance circuits.  Even a '2 Gohm' meter might need some help if the voltage is high enough to exceed the level where the impedance switches from high to low.  Dave's Agilent meter has high impedance on the 10V scale and below, so if he was trying to measure 25V, he might need something like this.

Ed

[Zucca]

Thanks edpalmer42, now it is crystal clear. To me it looks the principle behind that idea is still the Wheatstone bridge, which indeed is much older than, let's say, the NE555.

[T3sl4co1l]

What does mine do?

DMM: Hi-Z (haven't measured typical; 100s M+?) on low range (< 400mV), 10M above
VTVM: 11M, all ranges (1.5 to 1500V in 1.5-5-15-etc. steps)

Since the DMM has autoranging, it can oscillate on a high impedance source.  And it's not just that it looks like it's flipping shit, the oscillation is really there.  A nice demonstration is an electrolytic cap charged to a volt or a few, then clipped to the meter (and nothing else).  It slowly discharges, then recharges due to absorption, and so on.

And yes, VTVM.  And I use it regularly too.

Tim

[vk6zgo]

Another situation where the DMM's 10M input impedance gave a misleading result was in the following case:

A 24 v battery supply was isolated from an external circuit by a reverse biased power diode in the positive lead.
A colleague measured between  the outer end of the diode & the negative terminal,confidently expecting a reading of zero volts.
He was astounded to see a reading of 11 volts!

A good reminder that power diodes aren't ideal,& that a reverse resistance of  around 11.8M is quite possible.

That was a very different scenario,but again confirms that the best Test Instrument is the one between your ears---closely followed by a Scientific Calculator.

[Rerouter]

Hi Dave, Given your precision current source and precision voltage standard and infinite impedance Agilent Multimeter ( I have the same model, BTW), I was wondering if you can make an accurate measurement of some 13 gig-ohm ( thats right 13 x10^9 ohms) resistors I have. See attached image. Perhaps this is an odd request, but you seem rather clever and I imagine you or someone here might  have some ideas other than applying 1000 volts to it and trying to measure 77 nano-amps. I can send you some if you promise to do a short video on it. Years ago Jim Williams at LTC measured them with some contraption he had in his lab and confirmed the value but he never disclosed how he did it ( I can't call him today....for obvious reasons) or with what.   Thanks and wet soggy cheers from Oregon --Bob

Have a look into industrial megger testers used by electricians, for 6.3KV and higher capable units gigaohm is the standard unit, with the mid end meggers for 3 phase 1000V can generally measure close to the gigaohm value your chasing,

[Rufus]

What is this "wouldn't put a low value pot in there because the wiper resistance is going to have an effect on the stability" at about 2:09?

Pots usually have poor end to end tolerance and maybe not very good end to end temperature stability so if used as part of a divider swamping them with a better fixed value resistor may be a good idea. I don't see what wiper resistance has to do with it - what if anything am I missing?

[SeanB]

13G and with a half watt power rating. Even in SF6 at 400Bar those resistors will flash over before they get to a half watt power dissipation. No Idea how those were measured, but I remember there were a few appnotes from him about measuring ultra high impedance opamp input currents, might be a good starting point.

[Pillager]

This video takes me back 20 years, when I was learning electronics.

We were shown how you can have a voltage or current error, depending on the setup, when measuring both values at the same time. Of course, we were using analog meters then, so it was even more important, especially when switching ranges...

[crusader66]

I recently discovered that my old Fluke 8810A has input resistance >= 1Gohm on it's 200mV, 2V and 20V ranges.  Probably would have been a good meter to use here.

[David Hess]

Hi Dave, Given your precision current source and precision voltage standard and infinite impedance Agilent Multimeter ( I have the same model, BTW), I was wondering if you can make an accurate measurement of some 13 gig-ohm ( thats right 13 x10^9 ohms) resistors I have. See attached image. Perhaps this is an odd request, but you seem rather clever and I imagine you or someone here might  have some ideas other than applying 1000 volts to it and trying to measure 77 nano-amps. I can send you some if you promise to do a short video on it. Years ago Jim Williams at LTC measured them with some contraption he had in his lab and confirmed the value but he never disclosed how he did it ( I can't call him today....for obvious reasons) or with what.   Thanks and wet soggy cheers from Oregon --Bob

Nothing fancy is needed.  Use the shunt resistance of your best 10 megohm input voltmeter as a voltage divider with the 13 gigohm resistor and a 10 volt source and you should get a reading of 7.69 millivolts.  Any good voltmeter will have an input bias current low enough to keep the error far below the 10% tolerance of the resistor.

[EEVblog]

Hi Dave, Given your precision current source and precision voltage standard and infinite impedance Agilent Multimeter ( I have the same model, BTW), I was wondering if you can make an accurate measurement of some 13 gig-ohm ( thats right 13 x10^9 ohms) resistors I have. See attached image. Perhaps this is an odd request, but you seem rather clever and I imagine you or someone here might  have some ideas other than applying 1000 volts to it and trying to measure 77 nano-amps. I can send you some if you promise to do a short video on it. Years ago Jim Williams at LTC measured them with some contraption he had in his lab and confirmed the value but he never disclosed how he did it ( I can't call him today....for obvious reasons) or with what.   Thanks and wet soggy cheers from Oregon --Bob

Nothing fancy is needed.  Use the shunt resistance of your best 10 megohm input voltmeter as a voltage divider with the 13 gigohm resistor and a 10 volt source and you should get a reading of 7.69 millivolts.  Any good voltmeter will have an input bias current low enough to keep the error far below the 10% tolerance of the resistor.

That's one way. The other way is to use a known measured lower value parallel resistor, to bring the value down into the region of the multimeter.
I haven't done the math on these values to work out what tolerance you'd be able to measure to, but probably good enough for a 13G resistor.
Or with a 10V source and my 480 picoameter I could measure that to 0.5% or better easily.
Or I could use my Keithley 515A megohm bridge

[plesa]

Hi Dave, Given your precision current source and precision voltage standard and infinite impedance Agilent Multimeter ( I have the same model, BTW), I was wondering if you can make an accurate measurement of some 13 gig-ohm ( thats right 13 x10^9 ohms) resistors I have. See attached image. Perhaps this is an odd request, but you seem rather clever and I imagine you or someone here might  have some ideas other than applying 1000 volts to it and trying to measure 77 nano-amps. I can send you some if you promise to do a short video on it. Years ago Jim Williams at LTC measured them with some contraption he had in his lab and confirmed the value but he never disclosed how he did it ( I can't call him today....for obvious reasons) or with what.   Thanks and wet soggy cheers from Oregon --Bob

Nothing fancy is needed.  Use the shunt resistance of your best 10 megohm input voltmeter as a voltage divider with the 13 gigohm resistor and a 10 volt source and you should get a reading of 7.69 millivolts.  Any good voltmeter will have an input bias current low enough to keep the error far below the 10% tolerance of the resistor.

That's one way. The other way is to use a known measured lower value parallel resistor, to bring the value down into the region of the multimeter.
I haven't done the math on these values to work out what tolerance you'd be able to measure to, but probably good enough for a 13G resistor.
Or with a 10V source and my 480 picoameter I could measure that to 0.5% or better easily.
Or I could use my Keithley 515A megohm bridge

I thing that picoammeter or electrometer is only possible solution. I'm using Keithley 6485 it has integrated math function, which automatically recalculate input  current to specified units (Volts, Ohms...etc). Quite handy for high impedance measurement or voltage measurment with extra high input impedance (protection circuit  above 500V is reguired).
I do not think that such a resistor will survive more than 500V. And with high voltage the voltage coefficient will degrade the measurement precision.

[David Hess]

I thing that picoammeter or electrometer is only possible solution. I'm using Keuthley 6485 it has intergrated math function, which automatically recalculate input  current to specified units (Volts, Ohms...etc). Quite handy for high impedance measurement or voltage measurment with extra high input impedance (protection circuit  above 500V is reguired).
I do not think that such a resistor will survive more than 500V. And with high voltage the voltage coefficient will degrade the measurement precision.

Challenge accepted.

Here is my Tektronix DM502 which is old enough to drink (This particular meter was produced in 1975 making it 39 years old.) measuring a 999 megohm resistor with 10 volts of excitation.  You can see the top of the resistor, a Caddock MG815-3, directly attached to the negative terminal.  I am not exactly sure about the tolerance of the Caddock resistor other than it being 1% or better.  Unfortunately I do not have any larger value resistors handy at the moment but this will serve.

99.5 millivolts across the 10 megohm internal resistance of the voltmeter indicates 9.95 nanoamps of current.  9.99 volts - 99.5 millivolts = 9.89 volts across the unknown resistance.  9.89 volts / 9.95 nanoamps = 994 megohms which is well within the various tolerances including that of the 1% or better Caddock resistor.  I could make the same measurement with any of my other digital voltmeters that have a 10 megohm input resistance and get the same although more accurate results.

As you point out, a meter with integrated math would make this more convenient.

Inside every good digital voltmeter is an electrometer struggling to get out.

[plesa]

It will work for 1G resistors pretty well and results will be OK, but on >10G you must apply lot off averaging or increase the bias voltage. And it make the picoammeters more usefull in this measurement region.

[David Hess]

It will work for 1G resistors pretty well and results will be OK, but on >10G you must apply lot off averaging or increase the bias voltage. And it make the picoammeters more usefull in this measurement region.

The only change with the 10G (or 13G resistor) is the loss of one significant digit because of the smaller signal which still leaves about 1 part in 50 including noise for this 200mV 3.5 digit voltmeter.  The resistance seen by the voltmeter is practically identical and the noise will be the same.  1G in parallel with the internal 10M shunt resistor is 9.9M.  10G in parallel with the internal 10M shunt resistor is 9.99M.  The meter bias current is already lower than 9 picoamps (*) just to meet its specifications with 11.1M of input resistance (**) and a resolution of 100 microvolts.

Now you have got me wondering how to design a chopper stabilized input stage with a bias current lower than that.

(*) One nice thing about old test equipment is that in this case, full service documentation is available.  The input bias current for the integrated buffer is specified to typically be 4pA at 25C and 40pA at 70C.  Actual use shows that it is better than that.

(**) The high impedance buffer actually sees 11.1M maximum because there are additional 1M and 100K resistors in series with its input after the 10M decade divider for overload protection.

[kedwards22]

I haven't done the calculations on this but could you use capacitive discharge to measure 11gig. Get a high quality, low self discharge cap, charge it up to a known voltage, leave it for a few hours (?) completely open circuit, then re-measure it's voltage. You'd have to do the measurement quickly so the meter doesn't discharge it too much. This would give a value for it's self discharge internal resistance, then repeat with the 11 gig resistor attached, and then a simple calculation gets you the resistor value. To get an accurate measurement you might need peak detect on the meter and a fairly constant temp room. Haven't done the numbers so I might be way off with this.

[David Hess]

I haven't done the calculations on this but could you use capacitive discharge to measure 11gig. Get a high quality, low self discharge cap, charge it up to a known voltage, leave it for a few hours (?) completely open circuit, then re-measure it's voltage. You'd have to do the measurement quickly so the meter doesn't discharge it too much. This would give a value for it's self discharge internal resistance, then repeat with the 11 gig resistor attached, and then a simple calculation gets you the resistor value. To get an accurate measurement you might need peak detect on the meter and a fairly constant temp room. Haven't done the numbers so I might be way off with this.

It would be about 5400 picofarads per minute so this is certainly feasible.  At low voltages a JFET or MOSFET could be used to buffer the signal so it may be continuously monitored.

[ConKbot]

I know this trap! I was building a lipo battery dissipative balancer circuit with a micropower reference/comparator/amp IC, and since the battery voltage divider was a high impedance node I ended up using a FET input opamp as a unity gain buffer. 

I have its power switch toggle between +/- 4.5v  and 0-9V  (powered off a 9V battery)  I found some shrouded meter sockets, and some male-male shrouded DMM test cable leads, that way I can still keep using my normal DMM probes, and put it all in a housing. The offset voltage may be an issue on a 6 digit DMM, but its fine on my 87V