RE: Blog #4: Characterising Thermistors – An Inconvenient Truth, Taking Things to the Fifth Degree

DAB — Sat, 06 Aug 2022 19:34:35 GMT

Good job Gough.

If you want precision measurements, you have to use precision parts.

Even 1% resistors may be too lose for your application.

When in doubt, user your meter to measure the true resistance, so you know where you stand.

RE: Blog #4: Characterising Thermistors – An Inconvenient Truth, Taking Things to the Fifth Degree

michaelkellett — Sat, 06 Aug 2022 12:20:05 GMT

Thanks for your efforts, interesting discussion.

I mentioned a few days ago that I have a current experiment running (over a temperature range from -10 to +50C). I'm using a 6.5 digit DMM in 4 wire mode to multipex and measure 4 thermistors. I can resolve 1mK which is useful for my purposes and I don't much care about absolute accuracy).

In my experiments I've found the spread between different parts from the same batch to be small but the absolute errors to be quite large.

The magenta trace is the chamber temperature and the other three are derived from three thermistors glued to test samples in an isolated inert atmosphere isothermal box. This internal enclosure gives a lot of thermal noise filtering. The temperature is derived by single beta resistance to temperature conversion. The thermistors are ATC Semitec 103JT050 parts. The large errors in absolute calibration that you discuss are clearly visible. (I trust the chamber's Pt sensor !)

It seems to me that apart from all the problems you describe it is also necesary to think very carefully about the electronic hardware as well. The thermistor in your data table (from an earlier post) has a more than 1000: dynamic range of resistance and you probably need better than 1% resolution - so that's either a logarithmic ADC or an 18 bit linear one !