4-wire electrode array resistance measurement

[leonhart88]

Hey All,

I'm working on a research project that involves an electrode array with 4 elements inside a flexible tube filled with conductive liquid (let's say water).  The idea is to run a constant current through the first electrode, with the last electrode being ground.  And then measure the resistance across the internal two electrodes.  Upon deformation of the tube, the cross sectional area will change, thus changing resistance.  Assuming a constant current with uniform current density, I would get a proportional change in voltage.

I believe this is typically how some multimeters measure resistance, and this is also apparently commonly done for measuring the resistance of geographical ground types.

My question is whether a constant DC current or AC current is preferable for this.  A constant DC current is easier to implement, and since we are only interested in the change in voltage, any DC biases would not really be a concern.  However, we would be able to AC-couple an AC signal and remove and DC offsets/biases but then I would need to consider the frequency and impedance of the electrodes.

What would be your suggestions?


Thanks!

[TimFox]

It's easier to measure low resistances using low-frequency AC, to avoid DC offsets.
For very low signal levels, you can use a low-noise differential preamplifier and a lock-in amplifier synchronized to the AC source.
At higher frequencies, you need to worry about the dielectric properties of the fluid.
Be careful of electrolysis of some fluids, especially water or salt solutions in water.

[Ian.M]

DC is almost certainly going to get you into trouble with electrochemical changes at your electrodes.  Even if the anode is sufficiently corrosion resistant (e.g.platinum), you are likely to get any non-alkali metal cations plating out at the cathode + evolution of hydrogen.   If the electrolyte isn't continuously agitated, its also highly likely you will develop concentration gradients and non-uniform conductivity in the tube.
 

[leonhart88]

It's easier to measure low resistances using low-frequency AC, to avoid DC offsets.
For very low signal levels, you can use a low-noise differential preamplifier and a lock-in amplifier synchronized to the AC source.
At higher frequencies, you need to worry about the dielectric properties of the fluid.
Be careful of electrolysis of some fluids, especially water or salt solutions in water.

What frequencies are we talking about?  100Hz?  1kHz?

I was thinking of using an AC-coupled preamplifier and then a differential amplifier.  However, I'll look into different methods (I know there any many amplification techniques) when I come to that point in time.  The voltage change that I'll measure is directly proportional to the amount of current I'm driving as well so I need to somehow determine an acceptable current (it needs to be large enough for good signal to noise ratio, but low enough so that it doesn't change the temperature of the media, otherwise resistivity willl change).

At higher frequencies I would also worry about skin effect affecting the current density uniformity.

[leonhart88]

DC is almost certainly going to get you into trouble with electrochemical changes at your electrodes.  Even if the anode is sufficiently corrosion resistant (e.g.platinum), you are likely to get any non-alkali metal cations plating out at the cathode + evolution of hydrogen.   If the electrolyte isn't continuously agitated, its also highly likely you will develop concentration gradients and non-uniform conductivity in the tube.

The electrodes will be platinum.  Can you explain further on how electrochemical changes will affect the system if DC is used and not AC?  Is this because corrosion will change the DC offset/bias over time?

If using AC, any suggestion on what kind of signal this should be?  Square wave, sine wave, etc.?

I should mention that this thing is meant to be placed in sterile saline only at the time of an experiment.  Once it is finished, the electrodes are removed from the solution.

[Vgkid]

I know that in cryogenic temperature bridges they use a square wave, as it is easier to measure. Though if you use a dac, you could use a a2d that uses the dac output as a reference.(I believe that is how the GR 1657-89 digibridge's work), I dont have a manual handy.

[SeanB]

You want to use AC, as pure a sine wave as you can generate ( typically you want low distortion, under 5% is a good point to aim for) at a low frequency, preferably a simple transformer derived AC mains waveform so as to reduce the effects of induced AC hum on the signal.  Use a low voltage transformer and use a low pass filter to get extra smoothing, and then you can use this with AC coupling on all the electrodes, using a large value film capacitor ( large so the reactance at the test resistance is lower by about an order of magnitude than the solution) to get all DC offsets away, so the electrodes do not corrode from electrolytic effects.

Then you can use a differential amplifier for the 2 sense terminals, and use a synchronous demodulator to only sample the AC waveform on the peaks, which reduces noise on the signal, and also gives a good low ripple DC signal voltage.

[Vgkid]

Sean responded faster than I could respond
see the 1689.3 for a detailed description, see section 4 for schematics
http://bama.edebris.com/download/gr/1689/1689%20Precision%20Digibridge.PDF

[SeanB]

Funny enough this is a common problem, as you get a lot of liquid level sensors that use electrodes, and they often have to run for decades without problems. AC mains derived sine wave is the best cheap solution.

[TimFox]

60 Hz AC (filtered as suggested above) should work.
For liquid-level sensing, capacitance is usually used, especially with non-conductive liquids such as fuels.
I'm not sure, but I believe the capacitor itself is coaxial, contained inside an outer cylinder.

[leonhart88]

I forgot to mention that the resistivity of the saline is around 712 ohms-mm at 25C.  The electrode pair is 20mm away from each other and the internal diameter of the tubing is 12.7mm.  This means the resistance between the electrodes is roughly 28.4 ohms when un-deformed.

You want to use AC, as pure a sine wave as you can generate ( typically you want low distortion, under 5% is a good point to aim for) at a low frequency, preferably a simple transformer derived AC mains waveform so as to reduce the effects of induced AC hum on the signal.  Use a low voltage transformer and use a low pass filter to get extra smoothing, and then you can use this with AC coupling on all the electrodes, using a large value film capacitor ( large so the reactance at the test resistance is lower by about an order of magnitude than the solution) to get all DC offsets away, so the electrodes do not corrode from electrolytic effects.

Then you can use a differential amplifier for the 2 sense terminals, and use a synchronous demodulator to only sample the AC waveform on the peaks, which reduces noise on the signal, and also gives a good low ripple DC signal voltage.

Can you explain why a sine wave is preferable, and why low frequency instead of something in the kHz range (other than induced AC hum)?  I haven't done a lot of AC work so my knowledge is a little lacking/rusty.

I did some more searching on electrolysis and from what I can gather, you want AC because the chemical reactions will effectively "cancel" each other out, whereas in the presence of DC, chemical reactions can cause the electrodes to corrode.

I was thinking of using a frequency generator we have at the lab to produce a sine wave as an input to a constant current driver.  We would use some sort of ADC system to capture the output from the sense electrodes.  I suppose to could also hook up the output from the generator and use that as a reference to sample only on the waveform peaks.  I can see that doing this will give a good DC signal voltage, but why is sampling at the peaks less noisy than sampling at other times?

[leonhart88]

Sean responded faster than I could respond
see the 1689.3 for a detailed description, see section 4 for schematics
http://bama.edebris.com/download/gr/1689/1689%20Precision%20Digibridge.PDF

Thanks for this Vgkid.  I'll give it a good read.

[leonhart88]

Also, wouldn't you ideally want to use a frequency other than 60Hz so you can filter out possible noise from 60Hz mains or use a band pass filter for the desired frequency range you're targeting?

[Vgkid]

yes,typically a low frequency
9.8 Hz, 11.6 Hz, 13.7 Hz (default), 16.2 Hz, or 18.2 Hz. really any non multiple of the mains frequency

[Chris Jones]

I believe that for measuring the conductivity of deionised water, AC at a few kHz is often used. Personally I cannot see any reason why a sine wave would be better than a square wave.

If you can digitise the waveform with a much higher sampling rate than the frequency of the AC excitation, then you should be able to suppress noise much better than would be feasible by sampling on the peaks, as you could make the noise bandwidth arbitrarily narrow (e.g. use a long window with a Discrete Fourier Trasform). You would not need any finely adjusted analogue filters. A crude anti-aliasing filter ahead of the ADC might be worthwhile, but the cutoff frequency of this only needs to be chosen based on the sampling rate and could therefore be much higher than the excitation frequency, so that the filter component tolerances have no effect on the measurement result.

As pointed out already, it would be best to make the excitation frequency different from the mains frequency (and ideally not even a harmonic of the mains frequency), so that mains interference has little effect on the result.

[TimFox]

If there be any frequency dependence to the fluid impedance, such as skin effect or dielectric effects, then the results from a square wave will be different than from a sine wave.  Assuming the overall problem is linear, the sine wave is the only input waveform that produces a sine wave output waveform (shifted in magnitude and phase), but the square wave will be distorted by different responses at the odd harmonics of the fundamental.