I've been doing some reading on here about measuring temperature with NTC thermistors and using the Steinhart-Hart Equation.

However I am still a bit confused about the selection of a resistor to put in a voltage divider with the thermistor. Should I select a resistor value that is the same as the thermistor's resistance at the centre of my range of interest (200F) to maximise the sensitivity?

Also do I need an opamp or inamp to remove any noise common to both thermistor leads as the thermistor probe will be in an espresso machine boiler, so I expect some mains and switching noise ?

If I use a precision voltage reference for the thermistor, my current understanding is that I must use that same voltage for the ADC, is this correct?

  • \$\begingroup\$ Are you sure a thermistor is the proper choice for measuring 200 C? Can you provide a data sheet? \$\endgroup\$ Dec 7, 2016 at 4:30
  • \$\begingroup\$ @WhatRoughBeast 200°F ~= 93°C. Not so bad. \$\endgroup\$ Dec 7, 2016 at 4:40
  • \$\begingroup\$ @SpehroPefhany - Oops. Damn, I hate it when that happens. \$\endgroup\$ Dec 7, 2016 at 13:16
  • \$\begingroup\$ There is an easy way to calculate the resistor value here \$\endgroup\$
    – Yasindu
    May 7, 2020 at 12:20

1 Answer 1


Probably you should use the mid-range-interest resistance. The exception might be if you were willing to give up some actual performance to make the temperature display look better at room temperature, with a low-resolution ADC.

Noise can be mitigated with a simple RC filter - temperature is a slow variable.

Reference voltage is preferably the same as the ADC but really a thermistor typically changes around -5%/degree C so any "reasonably stable" reference will give you acceptable performance for your particular application.

  • \$\begingroup\$ What would be a reasonable time constant for a low pass filter for a thermistor in a boiler? I was planning on taking say 10 measurements (per second?), rejecting the outliers (?) and averaging the rest of the values to improve the resolution. I've read in these forums that a moving average is also a type of low pass filter. If thats the case would it be sufficient to just use a 10uF electrolytic and a 0.1uF (ceramic?) in parallel to remove any non DC from the signal? \$\endgroup\$ Dec 8, 2016 at 18:55
  • \$\begingroup\$ Time constant maybe 0.1~1 second. The commercial cappuccino machine I designed the controls for used a flash boiler so it used thermocouples and ran much hotter. I think there was a regulatory (safety) reason. If you have an input that can work with, say, 100K you can add some series resistance and then a capacitor such as a 1uF-10uF ceramic to ground. \$\endgroup\$ Dec 8, 2016 at 19:44
  • \$\begingroup\$ I'm sorry, I don't quite understand your statement "If you have an input that can work with, say, 100K you can add some series resistance" Could you explain further please? \$\endgroup\$ Dec 9, 2016 at 1:28
  • \$\begingroup\$ Whatever you are connecting the thermistor + resistor to will allow some maximum source resistance before the error gets unacceptable. If that number is low you probably can't add much resistance in series so the cap has to be large value. An op-amp buffer would easily allow 100K or more depending on the type so you can avoid using an electrolytic cap. \$\endgroup\$ Dec 9, 2016 at 1:38
  • 2
    \$\begingroup\$ I'm now using a RC filter with a 100K resistor and a 1uF capacitor giving a time constant of 0.1 seconds, then a voltage follower using a MCP6001 op-amp. I'm getting nice stable results. Thank you for your help @SpehroPefhany \$\endgroup\$ Dec 25, 2016 at 15:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.