Digital Multimeters (DMMs) and Precision DC Sources

cancel
Showing results for 
Search instead for 
Did you mean: 

Measuring mOhm resistances with a SMU (e.g. PXIe-4139, PXIe-4141)

I'd like to measure resistances in the mOhm range (~10 mOhm) with an accuracy of ~ 1 %. Currently I use a 4-wire setup with the PXIe-4139 or PXIe-4141, employing a delta method with a high current of +100 mA and a low current of +10 mA or -100 mA. (I tried both, similar results.) The measured voltage difference is ~1 mV. However, I see an offset in the measured voltage (or computed resistance), i.e. all voltages (measured on various calibrated resistors from 2 to 20 mOhm) are too high by anywhere from 0.0 mV to 0.2 mV. This results in an error of up to 20 %, which is unacceptable.

I assumed that the delta method should eliminate these offsets?

Is it possible that the lead wire resistance plays a role here (despite the 4-wire setup)? The PXIe-4141 spec sheet mentions "Add 0.1% of LO lead drop to voltage accuracy specification". Does that mean if the leads have 5 Ohm resistance, I have to expect an error of 5 mOhm in my measurement?

Thanks!

0 Kudos
Message 1 of 13
(874 Views)

1) Are you running 4-wire all the way out to the device?

2) What are algorithm are you using to measure the resistance?  It seems like you are forcing DC current and measuring DC voltage.  Can you confirm?

3) What ranges are you setting the instrument to?

4) At very low voltages, the EMF caused by relay contacts can come into play.  Based on your numbers 10mOhms*10mA*1% = 1uV, I think that is OK, but it you want to compensate, also measure the voltage with no current (the emf induced voltage) and then subtract that from your voltage measurement before calculating the resistance.

0 Kudos
Message 2 of 13
(849 Views)

ad 1) Yes, 4 wire all the way to the device.

ad 2) Yes, I apply DC current and measure DC voltage:
session.output_function = nidcpower.OutputFunction.DC_CURRENT
session.sense = nidcpower.Sense.REMOTE

ad 3) I usually use 'session.voltage_level_autorange = True'.

ad 4) I would measure the EMF voltage by applying a current of 0 mA?

0 Kudos
Message 3 of 13
(845 Views)

3) Normally I force the ranges - that way I know for sure what it is using.  If the SMU is choosing too high of a range, that would kill your resolution/accuracy.

4) Yes, you measure the voltage with the relay closed and no current flowing.  The EMF is like a very small battery that is induced by the (mechanical) relay contacts.  I have had to compensate for it for very small measurements but I believe it is normally under 1uV.

0 Kudos
Message 4 of 13
(840 Views)

ad 3) If I understand the documentation correctly, the PXIe-4141 has only one voltage range (10 V). The current I apply seems to be accurate, i.e. the reported current is very close to the set current (within 0.1 %  or so). Also, the high current of the delta method is the max the card can supply, so I would have to force it to the highest range, so it cannot pick any worse than than if in auto mode. For the low current I could force it to use one step lower range indeed.

ad 4) I will measure EMF and report back.

0 Kudos
Message 5 of 13
(836 Views)

I was looking at the 4139 (I use 4130, 4138, 4139).

I just looked up the 4141 and it appears that you are correct and that resolution and accuracy may not be enough for what you want to do.

 

markshancock_0-1652226046609.png

 

0 Kudos
Message 6 of 13
(833 Views)

To compute the resistance I would recommend using Measure Multiple API which returns both measured voltage and current as the actual current might slightly vary from the requested(set) current.

Santhosh
Soliton Technologies

New to the forum? Please read community guidelines and how to ask smart questions

Only two ways to appreciate someone who spent their free time to reply/answer your question - give them Kudos or mark their reply as the answer/solution.

Finding it hard to source NI hardware? Try NI Trading Post
0 Kudos
Message 7 of 13
(821 Views)

@santo_13 wrote:

To compute the resistance I would recommend using Measure Multiple API which returns both measured voltage and current as the actual current might slightly vary from the requested(set) current.


Good point. I do record the set current and the applied current and use the applied current for the resistance calculation (averaging over ~ 40 points, more points don't make a difference). However, the discrepancy between the applied current and the set current is very small (~0.1 %), so this is not the source of error.

0 Kudos
Message 8 of 13
(810 Views)

@markshancock wrote:

I was looking at the 4139 (I use 4130, 4138, 4139).

I just looked up the 4141 and it appears that you are correct and that resolution and accuracy may not be enough for what you want to do.

 

markshancock_0-1652226046609.png

 


I will repeat the measurement with a 4138/9 and use a higher current (though having 4 channels like on the 4141 would be nice). My calculations showed that with the PXIe-4141 and 100 mA current, I expect an error of ~ 50 %. Using the same current, the error would be 4 to 8 % for the 4138/9. At 3 A current, the error should be < 0.5 % for the 4138/9. I should have done these simple calculations before. What strikes me is that there is very little noise in the measurements I did with the 4141. All the numbers are very reproducible and they agree between various resistors from 2 mOhm to 20 mOhm, only they have an offset (up to 20 % of the voltage at 20 mOhm, so that kind of agrees with the 50 % I just calculated) that I cannot get rid off. If I was hitting the resolution limit of the card, I would expect there to be noise instead of a reproducible offset?

0 Kudos
Message 9 of 13
(807 Views)

Most of the offset spec on our SMUs is just that – offset.  It should go away if you're using the ∆V/∆I method of computing resistance.  It is quite possible that the error you're seeing is in fact from the error source covered by the "Add 0.1% of LO lead drop to voltage accuracy specification" line in the specs you mentioned earlier.  Can you reduce your lead resistance to see if that improves accuracy?  Or make it worse to see if the accuracy gets worse?

0 Kudos
Message 10 of 13
(798 Views)