Thermistor Calibration - Steinhart-Hart ?

[dkossman]

I bought a 10k thermistor (Pentair 520272 non-oem) on Amazon, (https://www.amazon.com/dp/B07Q8RHK7D?psc=1&ref=ppx_yo2ov_dt_b_product_details) and have this hooked up to an ADS1115 on a Pi, and did a quick test using a C program which uses the Steinhart-Hart model. I measured Vs and Rs with a pretty good multimeter. However the temperature measurement is off by almost 2 degrees (F).

I found a calculator online at https://www.thinksrs.com/downloads/programs/Therm%20Calc/NTCCalibrator/NTCcalculator.htm which calculates the coefficients for you, based on 3 temperature/resistance readings. I was planning to use this to adjust the coefficients, but:

Looks like REM use interpolation of resistance/temperature values in thermistorTable.txt rather than Steinhart-Hart. However, updating this table to calibrate against a particular thermistor would be pretty difficult as you'd need to take a large number of resistance readings in the temperature range of interest, which would be tricky e.g. using a water bath.

Any suggestions?

thanks!

`

define Rs 9930 // series resistor

define mx 32767 // adc range max

define Vr 2.048 // adc voltage measurement range

define Vs 3.28 // adc supply voltage

/ steinhart-hart coefficents for 10K Ohm resistor, -55 C to 150 C /

define sA 1.129241E-03

define sB 2.341077E-04

define sC 8.775468E-08

float temperature(float r) { / if r is the thermistor resistance we can use the Steinhart-Hart equation to calculate the temperature t in degrees kelvin: t = 1 / (sA + sB ln(r) + sC ln(r) ln(r) ln(r)) / float k, c; k = 1 / (sA + sB log(r) + sC log(r) log(r) log(r)); c = k - 273.15; // convert to Celcius return(c * 9 / 5 + 32); // and then return Fahrenheit }`

[tagyoureit]

REM includes a calibration offset.

[dkossman]

thanks for pointing this out - hopefully the offset will do the trick. I am waiting for my Pi 4b to arrive and then will breadboard this to check it out. Currently testing on an old Pi Zero W which i don't think has the horsepower to run NJSPC, REM, Nixie and DashPanel...

If i get really ambitious and you think it would be helpful, I may look at contributing a Steinhart-Hart option to REM for thermistors, but my javascript skills are C- at best.

[tagyoureit]

You can. But I'd just keep in mind that precision is never really a goal with pool management. IE if you change your diverters to use the floor drains instead of the skimmers then your temperature may drop 7+ degrees. So, philosophically, what is "the temperature" of your pool? If you can get the Steinhart-Hart to be 50% more accurate and the "true" readings are calibrated to be only 1 degree off (or even perfectly accurate) instead of 2 degrees, will that really be a huge gain over just setting an offset in the UI? Just food for thought to keep everything in perspective.

[rstrouse]

If you pull REM it will now allow you to set the calculation type to Steinhart-Hart. It uses the coefficients for 0 - 50 - 100C. I believe the table was also generated from these same values,

[dkossman]

thanks!