Public Lab Research note

Conductivity Sensing: Open Questions

by donblair | January 07, 2016 17:55 07 Jan 17:55 | #12569 | #12569

Design Goals

A lot of people are interested in measuring conductivity in liquids -- because it provides a useful picture of how much stuff has dissolved in the liquid, it has applications in monitoring contaminants in water, as well as a host of other applications. I've been finding through readings online that are many subtleties to measuring conductivity, and typical commercial instruments employ lots of little tricks to get the most accurate and precise measurements possible. The goal of the Riffle project has been to make effective water monitoring as accessible as possible; so, if some of these tricks involve e.g. platinum electrodes, we might consider a tradeoff between e.g. sensor reliability and accessibility -- maybe stainless steel screws provide good enough reliability for most of what we want to do. One complication is that we don't yet know what sort of instrument quality we're aiming for; so, at this point, the goal has been "as good & accessible as possible".

So, in that spirit, here are some questions that have arisen. Some of them are already fairly well settled.

Probe material

There are a lot of options here, but we were encouraged to see that some expensive commercial instruments use stainless steel screws for this purpose, so we were encouraged that we can simply go with that. So: probably 'solved', for now.

Probe geometry and calibration container size

This is interesting. The trick here is that electric field lines between the two probes "blossom out". In a small container, this means that the field lines can interact with the container walls [REF NEEDED], affecting the measurement. This isn't a problem for field measurements, necessarily (unless the device is inside another container, like PVC pipe) -- but it makes calibration in small containers a challenge -- we have found that the size of the small container, and how deep the probe is situated within it, affect the measurement significantly. How 'small' is a small container, and how 'significant' is the effect? We haven't really answered these questions yet, but they'll be important if we want to characterize the accuracy of our sensor. Some ideas:

  • We can try to assess the effect that container size has on our measurement. Likely the effect drops off asymptotically, and we can specify a particular minimal container size to get the error under X%. Issue: large quantities of standard conductivity solution might be a pain to source / create.
  • We can come up with a standard container geometry (i.e. a large yogurt container) and probe depth for making calibration measurements. Then we can see how this changes in the field, and attempt to the relate the two measurements.
  • We can see what effect probe spacing and size has on the 'stray field' effect. There are indications online [REF NEEDED] that the field effect is maximized (i.e., worst) when the probe spacing (the space between the screws) is comparable to the probe size (the head of the screws).
  • There are distinct advantages to using four probes, in lieu of two probes -- in terms of field effects, but also in terms of avoiding issues of corrosion. This would require redesigning the circuit completely; but there are some examples of simple four probe circuit geometries out there to follow.

Voltage level

Right now we're using a 555 timer running at 3.3V to make our measurement, which means that we're putting 3.3 Volts into the water when making the measurement. All indications in my readings are that this is a Bad Idea, because electrolysis occurs above about 1.2 volts. Most of the instrument schematics I've seen try to drop the voltage down at least below .5 volts. Ideas:

  • Try to run the 555 timer at < 1.2 volts. There is, in fact, a variant of the 555 that runs as low as 1V. This isn't great, but it's better than > 1.2 volts. This approach would mean that we'd need to power the device via a voltage divider circuit (and an op-amp voltage follower, potentially?); then we'd need to amplify (via an op-amp, presumably) the output of the 555 back up to 3.3V, so that the Atmel328, running at 3.3V, could register the 'on / off' states of the 555 output as the full [0, 3.3V], and count the pulses appropriately.
  • Use a different circuit for measuring conductivity. There are several that I've found out there, and some of them seem fairly straightforward, would be lower-power than the 555, and have behavior that is easier to troubleshoot (the innards of the 555 are nicely understandable as far as ICs go, but they're still a bit complex to understand compared to e.g. a few resistors and a capacitor).


Here I'm going to throw in a bunch of useful references I've found recently (with help from Yagiz)

Great general references: <-- really good

great thesis from the 60's:

description of the wein bridge oscillator

nice thesis on salinity

nice guide to conductivity measurements

great guide to AC current

great guide to bridge circuits -- symmetrical bridge measures unknown capacitor

"Addition of low-pass filter to “twin-T” feeds pure DC to measurement indicator", at this link: twin-t circuit for differential capacitance:

more good explanation of bridge circuits for measuring AC:

(same discussion as above link)

further explanation of twin-T filters

self-calibrating twin-t circuit:

twin-t circuit in seawater analysis:

good description of twin-t resistivity measurement circuit

soil moisture measurements -- great description:

history of impedance measurements -- fascinating

explanation of four probe measurement

this paper uses both methods (including 555 timer): and uses this chip:

four probe measurement circuit:

good references on the four probe measurements

appropedia four probe measurements

great reference circuit, with precautions taken by instrument **** study this

Fantastic reference!!


Hah, just saw that @mathew posted a link I was reading this morning, but forgot to include:

Perhaps we can drop the voltage of the 555 timer even lower than 1V (out of spec), and still get acceptable results. Worth trying!

Reply to this comment...

large yogurt container

Love this idea for standardization; used yogurt containers pile up quickly in my house. Or a mason jar! I'd guess material can have some effect too -- i.e. don't use a metal saucepan or something? And it'd have to be in the center of the container, so i guess somehow sticking it through a hole at the center of the lid or something?

I agree that empirical testing is the way to go to determine these effects. You don't want to be spending time optimizing unnecessarily. Maybe a "grid" experiment where you have a row of vessels, and a series of probe geometries, and you test each probe in each vessel, like we did with the Infragram filter/webcam tests back at an early LEAFFEST?

leaffest tests

Maybe this list of references could be copied into the bottom of the Coqui page so it's easier to find, as has been done on the Oil Testing Kit Literature page.

Is this a question? Click here to post it to the Questions page.

Reply to this comment...

What will be the long term corrosive effects of current and salt water on our stainless steel screws? I think we've seen stainless steel probes on some commercially available models - is that what we're basing this off of? I'm talking to folks on #arduino on freenode who had that question.

Is this a question? Click here to post it to the Questions page.

Reply to this comment...

I wish I had brands to list, but yes, we've seen commercial probes using small (0-80) screws as their probes.

Reply to this comment...

I can help with probe material: Regarding Stainless Steel, general purpose grade is 304 and marine grade is 316. Stainless steel, particularly 304 grade, will "tea stain" however this surface corrosion can be rubbed off. 304 grade stainless steel most common in hardware stores. I'm not sure of conductivity variation between the two grades [304 and 316]. I'd recommend 316 grade as a standard, for probes, especially for salt water/coastal conditions. Other option is zinc plated screws but there may be variations in quality of the base metal, which is usually mild steel, and often poor quality these days judging by the way my screw driver tears them up. Also the zinc plating disguises the base metal so you may have bronze base metal without knowing.

Reply to this comment...

I've built one of your fine Coqui circuits. It works very well with the LDR (photo resistor). Is there a recommended distance between the metal probes for measuring conductivity? Are there acceptable limits for variation in distance? Apologies for missing any obvious references.

Is this a question? Click here to post it to the Questions page.

Reply to this comment...

I have never met a conductivity meter that doesn't have some amount of drift.

Having said that, the commercial conductivity meter at work was a low level a.c. Generator ( don't remember the freq, but I think it was high audio) feeding a wheatstone bridge. The conductivity cell was one "leg" of the bridge. An a.c. Voltmeter read the voltage differences between each leg of the Whitestone bridge. The resistance of the other leg of the bridge ( the one without the cell) was then adjusted until the voltage was zero. From here, the water conductivity could be calculated.

Because the voltage was a.c. and low, there were no electrolysis problems. Drift problems... Yes still some of those.

Maybe this is the circuit you are talking about?

Is this a question? Click here to post it to the Questions page.

Reply to this comment...

Login to comment.