4-wire Resistance and the Seebeck Effect

[raptor1956]

I've been playing around with my Rigol DM3058E bench multimeter and noticed something interesting.  Using one set of leads to a piece of wire about 750mm long I measured the resistance at 37 mOhm with it.  I then used another DMM to measure current and output a few different current values from my bench PS through the meter to the wire then read back the voltage drop using the 3058E in mV mode and the calculated resistance was 41 mOhm.  I varied the current from about 1mA to 250mA and it was right at 41 mOhm the whole time.  OK, that's one data point...

I was then poking a round using different meter leads and was getting different values for the resistance, from 33 mOhm to 41 mOhm.  So, there seemed to be an effect related to the leads I used with the Probemaster ones being gold plated and the other ones more of the Ni/Chrome kind of thing -- not sure what the metal plating is...

And now for the kicker...

While setup and reading 37 mOhm I put my hand on the negative connection and the resistance reading increased up beyond 43 mOhm.  I waited till it returned to 37 then put my hand on the positive connection and watched as the resistance decreased to about 31 mOhm.  I duplicated this several times then fired up my hot air gun and brought it near the connection without blowing the hot air on it and there was no change, but the second I started blowing the hot air on it the resistance value changed significantly and in the same direction as with my hand.  The response at both the negative and positive connections was the same as with my hand but the deviation was much greater.  I did not keep the hot air blowing long.

So, my best guess is that this is due to the Seebeck Effect and if so it highlights the difficulty in getting truly accurate resistance measurements with very low resistances where the voltage drop is often much less than 1mV and where even a small Seebeck Effect is then noticeable.  We have some guys that spend a fortune building special Kelvin clips to get a few uOhm more accurate but all that effort is swamped if you can't control for temperature variations of the connections.

I'll probably make a video of this and post it...


Brian

[Vgkid]

One way to think about is to realize that most meters will put out 1mA of current out. Multiply this by the 40microhm and youget a voltage reading of 40uV.now consider that the various thermocouple effects scross a grafient adding a few uV, or more. Ways to deal with this would be to use an ac excitation voltage, or a pulsed dc current. In which a voltage reading is taken with no current applied. Then the current is applied, and the two voltages are dubtracted. This gives an offset compensated ohms reading. A few dmm's actually give this feature.

[raptor1956]

Increasing the current to produce a larger voltage drop will reduce the effect at a rate proportional to the increase in current and so long as the current is low enough so as not to produce a heating effect that would tend to increase the measured resistance.  Going from 1mA to 100mA should lower the Seebeck Effect by a factor of 100.

I made a video and just completed editing and will upload to youtube -- might take a hour, we'll see...


Brian

[Dr. Frank]

For low Ohm values, i.e. low test current and low test voltages,you need an additinal function, called Offset Compensation. This removes thermo voltages, as you encounter.
Either you need a more sophisticated instrument, or you may also put a switch into the current line of your 4W setup, and null the reading, when switching the current off towards the DUT.
There are several videos, how to implement such a switch box.

Frank

[raptor1956]

For low Ohm values, i.e. low test current and low test voltages,you need an additinal function, called Offset Compensation. This removes thermo voltages, as you encounter.
Either you need a more sophisticated instrument, or you may also put a switch into the current line of your 4W setup, and null the reading, when switching the current off towards the DUT.
There are several videos, how to implement such a switch box.

Frank

Good info -- got a more specific pointer to those videos?

The use of a higher current that's pulsed to reduce any heating effect would also work, and if several different current values were pulsed the meter could normalize out the error using something like regression -- I think...


Brian

[raptor1956]

OK, videos up...

https://www.youtube.com/watch?v=jQC-wC-khXw&feature=youtu.be


Brian

[Dr. Frank]

Robrenz made several nice videos about 4W Kelvin clamps.. He's more a mechanical guy, but in this one, he builds that switch  box:

[ebclr]

http://www.ti.com/europe/downloads/2-%203-%204-Wire%20RTD%20Measurement.pdf

[ocw]

My way around the polarity problem for low resistance ohmmeter readings is to use an AC voltage to make measurements.  I do that in the form of an Analog Devices EVAL-CN0359-EB1Z board.  While the board was designed for conductivity measurements of fluids, it can also make regular four wire conductivity measurements using a test frequency between 100 Hz - 10 kHz.  Obviously resistance is the inverse of the displayed conductivity value.  While it's rated to have an accuracy of an better than 0.3% over a conductivity range of 0.1 ?S to 10 S (10 M? to 0.1 ?), I've found the accuracy to be much better than that.  Higher conductivity readings (lower resistance readings) are possible, although the accuracy there starts to suffer.  For more information see:
http://www.mouser.com/ds/2/609/CN0359-586926.pdf

The picture provided shows my CN0359 measuring a 0.05 ohm 0.5% resistor.  The 20.028 S reading displayed translates to 0.0499301 ohms (-0.14%).  The numbers which you see on the case are the Cell Constants required on my meter for more accurate measurements on the listed frequencies.  Obviously any reactance on the test frequencies will affect the readings.  It actually measures the impedance on its frequency range.

While not meant as a precision meter I've found the CN0359 useful.  I have another completely homebrew meter that uses transformer isolated 60 Hz up to 35 VAC,  up to one amp to make other low resistance measurements.

[zlymex]

I use 1A current to pass thru the wire or low ohm materials to be tested, then I use two probes or alligator clips to measure the voltage drop that I'm interested. Also, cut off the 1A current and measure the same spots(and subtracted) will cancel the thermal EMF. In this way, I can measure down to micro ohms from most of 6.5 DMM that resolve 0.1uV in the best voltage range.

Photo below is an 600A switch that I measured for resistance at different sections.
The resistances are in micro ohm and marked as green. For instance, resistance between point E and F is 23 micro ohm.

[raptor1956]

I took another shot at the 4-wire resistance taking into account the lessons on 'offset compensation' and including errors related to resistance heating when using higher current.  I demonstrate 3 different methods for doing 4-wire resistance tests with 2 of the tests only requiring basic DMM's and a power supply.

There is one more thing I did not include in the video as I'm still trying to get a handle on it.  Suffice to say that there is some weirdness with the DM3058E and 4-wire resistance mode.  If you watch the video I think you can see that the piece of wire I used as the resistance was right at 41 mOhm but the 3058E was reading about 36 mOhm before compensation and after compensation it read 30 mOhm.  I played around with the leads but the values kept coming up at 36 mOhm and 30 mOhm compensated.  I then power cycled the 3058E and went back to ohms mode and 4-wire mode and now it was reading about 46 mOhm and when compensated it dropped down and was bouncing between 40 mOhm and 41 mOhm -- exactly where I would have expected it to be.  So, I'm not sure why power cycling did this but this is the second time this has happened.  I need to duplicate one more time and maybe record a video of it.

Anyway, if anyone is interested or wishes to comment on the video here it is...




Brian

[raptor1956]

I use 1A current to pass thru the wire or low ohm materials to be tested, then I use two probes or alligator clips to measure the voltage drop that I'm interested. Also, cut off the 1A current and measure the same spots(and subtracted) will cancel the thermal EMF. In this way, I can measure down to micro ohms from most of 6.5 DMM that resolve 0.1uV in the best voltage range.

Photo below is an 600A switch that I measured for resistance at different sections.
The resistances are in micro ohm and marked as green. For instance, resistance between point E and F is 23 micro ohm.

Interesting, and without upping the current to at least 100mA you would not have been able to see the resistance at all.  The limitation of the built in 4-wire systems in bench DMM's using just 1mA would require an 8.5 digit DMM to see anything a 1mA and even then just barely and with errors imposed by quantization limits.  Increasing the current to 1A made it possible and given the extremely low resistance the heating effect would have been truly insignificant -- uW's.

In the video I mentioned above I tested a piece of wire about 750mm long which had a resistance of about 41 mOhm and with my 5.5 digit bench DMM outputting 1mA for the 4-wire resistance test the voltage readback would only have been about 41mV and given the range and resolution of my meter that's pushing it for that low a resistance.  I then tested using a two-meter method with an external PS and varied the current from about 1mA up to 3000mA.  I found in this case the best readings were in the 30mA to 300mA range.  With currents less than 30mA there were errors do to offsets and at current above 300mA there was i^2r heating that increased the resistance. 


Brian

[nctnico]

So, my best guess is that this is due to the Seebeck Effect and if so it highlights the difficulty in getting truly accurate resistance measurements with very low resistances where the voltage drop is often much less than 1mV and where even a small Seebeck Effect is then noticeable.  We have some guys that spend a fortune building special Kelvin clips to get a few uOhm more accurate but all that effort is swamped if you can't control for temperature variations of the connections.

I recently ran into this effect with current shunt resistors where one side would cool faster than the other side. This caused a 'phantom' reading of current which wasn't there.