# How to diagnose thermistor problem

I have a thermistor which I believe is a NTC 3950, and which I suspect does not work properly.

It is from a 3D printer extruder/hot end. The displayed temperatures and the ones I measure using a meat thermometer differ by 50% or more, and I can feel with my hand in the vicinity (not touching) that it is cooler than it used to be.

The normal operating temperature is around 200 deg C. The resistance of the thermistor at room temperature, measured by a multimeter, is 108k Ohm, which seems to be normal.

Any other things I can do to test the thermistor before replacing it? I do not have an electronics lab, only basic stuff. Can it even be the thermistor that causes the problem, ie is it normal/not unheard of for a thermistor to give normal values at room temperature and be off by 50% at higher temperatures?

• First, I think you need a better reference than a meat thermometer. Commented Sep 19, 2020 at 12:07
• @Barry I don't think the exact number of degrees is super important. There is an obvious difference from the way it was before. Commented Sep 19, 2020 at 12:12

If you replace the thermistor with one or more known fixed resistors (disconnecting the heater to avoid it possible overheating) you can see if the readings are correct. Use the table or equation for your thermistor (for example 25/50 $$\\beta\$$ = 3950. The two temperatures which are used to define $$\\beta\$$ are significant.

Note that if your circuit is (say) designed for a $$\\beta\$$= 4267 and your thermistor is $$\\beta\$$ = 3950 your controlled temperature will run high, perhaps by 20-30°C.

If it appears to be correct, swapping out the thermistor is an obvious and expedient thing to try.

But if you want to get to the bottom of it, a thermocouple or RTD temperature sensor with appropriate instrumentation would be a good way to do it. If you have a bead thermocouple that works with your multimeter on a °C range that is one possibility, though getting a good reading depends on being able to bury the bead within a hole or something like that, ideally to a depth of at least 5 - 10x the diameter of the bead. With care you should be able to get within 5°C or so (check the cold junction temperature and correct for that- that and nonlinearity are two major sources of error in cheap multimeters- you can't do much about the latter without a T/C calibrator but you can check the reading at room temperature).

A raw Pt1000 sensor can be used with an ordinary multimeter on resistance range and will be more accurate if you can attach it to the surface and insulate it, and wire it with thin wires. Accuracy will mostly be dependent on your skill in application- the sensors themselves are frequently good to a fraction of a degree and are not expensive (less than \$10 in single quantity, even cheaper for the lower accuracy variants).

In my experience, the thermistor is a good place to start. Replace it and see if that works. That’s the simple solution.

If you want to further test the thermistor, use a multimeter to to measure the resistance while wiggling the thermistor to test for breaks in the wires.

If you have the spec sheet on the thermistor, you can measure the resistance of the thermistor and see if it matches the spec sheet. Use ice water as a reference if necessary.

Just replacing it is probably the simplest though. You should keep extra thermistors on hand anyway for when it does go out.

• It's just odd that everything looks ok in the software. It shows 25 deg initially (true value 21-22, ie room temperature), and then when I set it to 200 (and the termistor value goes up to that), the meat thermometer shows around 130 eventually. If the termistor had been broken it should show negative values, or some error msg. If I disconnect the termistor I get an error indication in the sw. Commented Sep 19, 2020 at 13:22
• Anything is possible. Stuff fails all kinds of different ways. That’s why, for cheap components, it’s usually easiest to just replace it and see if it fixes it. Commented Sep 20, 2020 at 15:54
• Okay, that's one way to look at it. Commented Sep 20, 2020 at 18:25
• It probably isn’t software unless you were fiddling around with the code. So it is probably hardware. Well, which components are most vulnerable? The PCB is usually enclosed and protected. So most likely your issue is the thermistor itself or the cable running to it. Commented Sep 21, 2020 at 2:52
• I did not say the sw was the problem, I said the sw does not display thermistor error indications. Part of the issue is that I have already replaced the termistor, but not with an identical one (they may in fact be the same type, I am not sure), and it did not solve the problem. Commented Sep 21, 2020 at 10:01