How To Temperature Compensate Electrical Conductivity Measurement

Hi all,

Hopefully someone has done this before…
I have several conductivity sensors from here (http://webpages.charter.net/tdsmeter/products.html).
My probes have a thermistor attached so I can get the temperature of the water being tested, allowing me to compensate (In theory) Has anyone had success calculating this?

Thanks.

As Mr. Google will be glad to demonstrate, a lot of information is available on line regarding the temperature dependence of the electrical conductivity of water solutions. The formulas for the empirical corrections are quite simple and easy to apply.

As you will see, both the conductivity and temperature coefficient thereof depend very strongly on what is dissolved in the water, so what solutes at what concentrations do you expect to encounter?

3 sensors; Domestic tap water, Reverse Osmosis water and remineralised RO water. (The conductivity value is used in a mixing container to bring RO water up to 250 micro/S)

I know that the RO water is linear with conductivity rising by 4.55% for every degree centigrade. Tapwater is a difficult. The mixed RO water is mineralised using calcium, magnesium, potassium, and sodium powers so also not easy. 0.019% appears to be a rough number for tap water. I guess I could use a different calculation for the tap water, RO and re-mineralised.

It seems the most common method is to calculate to 25C: EC = electrical conductivity (222) T = Temperature (18 °C) a = 0.019 °C EC25 = EC / 1 + a (T - 25) EC25 = 222/ 1 + 0.019 * (18 - 25)

Code: float EC25 = x / (1 + (0.019 * (temp - 25)));

That is a standard empirical correction, but the “a” factor can change by a factor of two or more for even quite dilute solutes. A summary from this guide http://www.tcd.ie/Biochemistry/courses/js_conductivity.pdf? is as follows. (a is given in percent/degree C)

Acids: 1.0 - 1.6%/°C
Bases: 1.8 - 2.2%/°C
Salts: 2.2 - 3.0%/°C
Drinking water: 2.0%/°C
Ultrapure water: 5.2%/°C

Depending on what you want to accomplish, picking a value like 0.019/C seems like a reasonable compromise.

I am thinking that you will be measuring the actual conductivity, regardless of the temperature. Where the temperature will come into it, is working backwards from the conductivity that your apparatus actually measured, back to the implied ionic concentration that this corresponds to.

What do you think about heating the sample water to 25C and comparing against the recorded value (at say 5C)? I could then adjust the "a" factor to suit the fluid.

The fluid is pure water, mixed with an exact mineral amount. I may be able to find out the composition of the substance.

of the three samples tested, the pure water and mixed water should be fairly straight forward, the source (tap) could be more difficult, unfortunately this is also the sample that's temperature varies greatly (5 - 20C)

You could certainly determine the "a" factor yourself. The correction is nonlinear and the usual empirical correction is only an approximation. To determine "a" you would need to measure the conductivity of the unknown at several temperatures and then fit the empirical curve to the result. Can be done with Excel, on-line curve fitting tools or any of lots of different graphing programs.

How accurate are those sensors? Do you routinely calibrate them against a standard? I have some inexpensive sensors from ebay and they work well, but I have to calibrate them before every use. I use 0.1000 M KCl in ultrapure water @25 C as the standard.

The correction is nonlinear and the usual empirical correction is only an approximation.

Check multimap for non linear mapping - http://playground.arduino.cc/Main/MultiMap -