measuring very low resistance (nano- to microOhms)

[aluck]

Please help to find the method of measuring very-very low resistance. The range is from 0.0001 up to 5 mOhms.

Actually, I need the difference between two resistors, each of them is 2-5 mOhms. And the decimation should be such that I would be able to measure with a precision about 100 nOhms.

I heard about some not-too-straightforward approach involving AC current, transformer and True RMS ADC, but couldn't find the detailed explanation of this method.

Any link/suggestion/book reference will be very appreciated!

[alm]

Keithley's low level measurement handbook (available for free from keithley.com after registration) may provide some useful information, don't remember how extensive its treatment of resistance measurement is. The advantage of AC is that you eliminate the Seebeck (thermocouple) effect.

Depending on what it is, one easy poor man's way is to just shove a lot of current through them. 10A through 1mOhm will drop 10mV. This assumes that the DUT can tolerate 10A and that heating isn't an issue. Offset compensation will likely be necessary if you use DC (constant current AC sources are fairly rare). The current has to be stable and known. One advantage is that at 10A, EMI won't be much of an issue.

[aluck]

Unfortunately, putting high current isn't an option here - the system is supposed to work in an explosive environment (oil refinery, etc.), and they possibly wouldn't allow us to use high current there.

[Zero999]

What's the current limit?

I don't see how this can be done without using a considerable current and 10A is quite low in the scheme of things.

[Kiriakos-GR]

And I was thinking so far that I am the lunatic of the forum, who has chasing mOhs   
(At the range of 1-100mOhm.)

You definably need a four wire instrument.
But I have no idea about suggestions.   

[Fox]

Maybe you should look at the Appnote Nr.98 from Linear Technologies  on page 11 is a milliohmmeter with an resolution of 1mOhm.
Could be a good startingpoint.

[Neilm]

If you are only limited to very low currents the chances are that thermal noise will swamp what you are trying to measure - I very much doubt you will get nano-ohms. You might be able to do something that will resolve to 1 micro-ohm.

What's the current limit?

I don't see how this can be done without using a considerable current and 10A is quite low in the scheme of things.

10A is low. However for a potentially explosive atmosphere the requirements are very fierce. I seem to recall that the case has to be sealed to prevent any sparks. Also if there is any chance of inductance this could lead to voltage rising in the event of the test lead falling off while the test is in progress causing sparks. Otherwise I would dig up a reference to a 600A AC tester I once played with.

Yours

Neil

[Bloch]

You definably need a four wire instrument.

I think about a Wheatstone bridge. They are good at check difference between two resistors. But not sure if it can be used in this case.

[Bloch]

A 8845A/8846A http://us.flukecal.com/products/data-acquisition-and-test-equipment/bench-multimeters/8845a8846a-65-digit-precision-multime

Wide ohms range from 10 ohm to 1 Gohm with up to 10 u ohm resolution


8508A 8.5 Digit Reference Multimeter http://us.flukecal.com/products/data-acquisition-and-test-equipment/bench-multimeters/8845a8846a-65-digit-precision-multime but that do not look that cheap 

[nyo]

Problem is, the smaller the resistance, the bigger the amperage. You would also need very precise current and voltage. Maybe you can use a power meter or measuring receiver if high ac freq or thermal converter for low freq and use it to measure the power output.

[aluck]

If you are only limited to very low currents the chances are that thermal noise will swamp what you are trying to measure - I very much doubt you will get nano-ohms. You might be able to do something that will resolve to 1 micro-ohm.

What's the current limit?

I don't see how this can be done without using a considerable current and 10A is quite low in the scheme of things.

10A is low. However for a potentially explosive atmosphere the requirements are very fierce. I seem to recall that the case has to be sealed to prevent any sparks. Also if there is any chance of inductance this could lead to voltage rising in the event of the test lead falling off while the test is in progress causing sparks. Otherwise I would dig up a reference to a 600A AC tester I once played with.

Thermal noise doesn't matter at all! As I said before, the goal is actually a difference between two resistances. And one of those resistances is not supposed to change at all during a lifetime of this device. It is sealed, so any fluctuations in it's resistance is due to thermal changes. The second resistance is placed as near as possible, so any change in temperature would show up in both of them almost simultaneously.

By measuring a change in the first resistance I will be (hopefully) able to cancel out all the thermal noise.

[aluck]

Maybe you should look at the Appnote Nr.98 from Linear Technologies  on page 11 is a milliohmmeter with an resolution of 1mOhm.
Could be a good startingpoint.

Thanks for the suggestion! Nice appnote!

[aluck]

You definably need a four wire instrument.

I think about a Wheatstone bridge. They are good at check difference between two resistors. But not sure if it can be used in this case.

More likely it would be a Thomson's (Kelvin's) Bridge. Yes, that was the first thing I thought about. Actually, I am going to take a straightforward approach and use a cascade with three op-amps to get 15000 amplification. Not sure if it will work, but going to try it out anyway.

But still, I couldn't stop thinking about that another method with AC current and transformer involved. Seems like the mentioned above app. note from Linear has something to do with it.

[nyo]

You need a 1689 rlc digibridge
http://www.ietlabs.com/pdf/Datasheets/1689.pdf

[Zero999]

Thermal noise doesn't matter at all!

Measuring a low resistance value with a low test current means measuring a very low voltage so thermal noise would be a very large problem.

As I said before, the goal is actually a difference between two resistances. And one of those resistances is not supposed to change at all during a lifetime of this device. It is sealed, so any fluctuations in it's resistance is due to thermal changes. The second resistance is placed as near as possible, so any change in temperature would show up in both of them almost simultaneously.

Is the value of one of the resistances known to a high accuracy?

If not you need to measure it before you can get the difference so it's not much help.

Otherwise a wheatstone bridge can be used, as already mentioned above but the current still needs to be high enough to create a voltage high enough to swamp the thermal noise.

[aluck]

Is the value of one of the resistances known to a high accuracy?

Yes! It's a high-precision resistance. Sorry, I forgot to mention that.

[aluck]

You need a 1689 rlc digibridge
http://www.ietlabs.com/pdf/Datasheets/1689.pdf

It's a device. What I need is a method to measure it in my own device.

[Mechatrommer]

Is the value of one of the resistances known to a high accuracy?

Yes! It's a high-precision resistance. Sorry, I forgot to mention that.

i think he meant... Is the value of one of the resistances known to a high accuracy?

[aluck]

Is the value of one of the resistances known to a high accuracy?

Yes! It's a high-precision resistance. Sorry, I forgot to mention that.

i think he meant... Is the value of one of the resistances known to a high accuracy?

And - yes again, it is known beforehand to a high accuracy.

[tekfan]

Or one of these:
http://amplifier.cd/Test_Equipment/other/Burster-2302.html

Although I don't know if it can measure resistances low enough.

[alm]

A 8845A/8846A http://us.flukecal.com/products/data-acquisition-and-test-equipment/bench-multimeters/8845a8846a-65-digit-precision-multime

Wide ohms range from 10 ohm to 1 Gohm with up to 10 u ohm resolution


8508A 8.5 Digit Reference Multimeter http://us.flukecal.com/products/data-acquisition-and-test-equipment/bench-multimeters/8845a8846a-65-digit-precision-multime but that do not look that cheap 

The display may be able to indicate 10uOhm, but good luck trying to actually use this resolution. At a typical 1mA test current, you're talking about 10nV. Noise and offset voltages are going to swamp this unless you're extremely careful. Seebeck effect tends to be in the order of 1-100uV/K, even in the best case a 10mK difference in temperature will make 10nV resolution impossible. Even mOhm resolution is tricky, let alone uOhm.

[nyo]

You need a 1689 rlc digibridge
http://www.ietlabs.com/pdf/Datasheets/1689.pdf

It's a device. What I need is a method to measure it in my own device.

Its the method behind it, it's 4 wire and with 1 khz freq yous should be able to measure the low ohms

[aluck]

Its the method behind it, it's 4 wire and with 1 khz freq yous should be able to measure the low ohms

is there some detailed description of the method behind it?

[Fox]

Keithley's low level measurement handbook (available for free from keithley.com after registration) may provide some useful information, don't remember how extensive its treatment of resistance measurement is. The advantage of AC is that you eliminate the Seebeck (thermocouple) effect.

The methods of measurement are not that extensive, but there is an extensive section about the unwanted effects while measure low resistances.
So everyone who wants measure in the micro and nanoohms range should read it, and as alm said its free so get it.

Edit: Lock-In Amplifier is a method of measuring small signal within the noisefloor

[aluck]

Thanks a lot, guys! The book is spectacular.

[Ronnie]

Please help to find the method of measuring very-very low resistance. The range is from 0.0001 up to 5 mOhms.

Actually, I need the difference between two resistors, each of them is 2-5 mOhms. And the decimation should be such that I would be able to measure with a precision about 100 nOhms.

The Keithley models 622X (AC and DC current source) and 2182A nanovoltmeter combination can measure down to 10 n? with low power and low noise http://www.keithley.com/products/dcac/sensitive/acdc/?mn=62212182A   

[ElektroQuark]

Keithley's low level measurement handbook (available for free from keithley.com after registration)

Any link for download without registration?
Thanks.

[ejeffrey]

You can look at cryogenic resistance bridges.  There are several schematics published for them, and many are capable of measuring resistances with 7 digit accuracy and femtowatts of power dissipation.

The secret is in two parts: the first was mentioned: AC excitation.  The second important point is to do synchronous detection (as a lock-in amplifier).   You can do this with a couple of analog switches/multiplexors.  You just create a constance current source of 10 milliamp, and use some analog switches to turn it into a square wave that reverses the current polarity at ~18 Hz.  Then you amplify the voltage just a bit with a differential output amplifier (one that has normal and inverted outputs).  Use a second analog switch to alternately connect the two amplifier outputs to charge a capacitor.

10 milliamp through 1 microohm generates a voltage of 10 nanovolt.  A low noise amplifier can have 1 nV/sqrt(Hz), and if you average for 100 seconds, you can get your noise floor down to 10 nanoohms (in principle).  By passing the current through both resistors and using a difference amplifier, you could directly measure the resistance difference with some improvement in the needed dynamic range.

Of course, every part of the signal path is quite important.  The analog switches must have very low (or very balanced) charge injection, the first stage amplifier must be very low noise, and have enough gain that noise introduced later doesn't affect the result, but not so much that it clips due to interfering signals.  The switching frequency should be chosen so you get good rejection at 50/60 Hz -- either a sub-harmonic of the local line frequency, or faster than the line frequency and far from any harmonic.

The ultimate limit is given by the johnson noise of the resistor, but this circuit will not approach that even closely.  The voltage noise of transistors is just too high compared to such a low source impedance.  You could get 100 times or more improvement in the resolution by using a high ratio signal transformer to boost the voltage before the input amplifier.

[Ronnie]

Any link for download without registration?
Thanks.

This is the 6th Edition.

[aluck]

ejeffrey, thanks a lot!

Now all I have to do is just to make this thing work and go enjoy my well-deserved vacation somewhere in the Mediterranean

[ElektroQuark]

Any link for download without registration?
Thanks.

This is the 6th Edition.

Thank you Ronnie.
It seems some others have downloaded from your link too.

[Ronnie]

I have a hardcopy of the handbook given to me for free by Keithley representative in the Philippines sometime May 2005. The handbook is really nice to have as it helps me solve low level analog circuits problem especially leakage current in the sub pA at work. It is highly recommended