Thermistor measuring

Question

I need to acquire the temperature of a box thanks to a thermistor. I am wondering how to do that with a really high precision. Here are the caracteristics of the thermistor:

T° Thermistance

-55°C 500K

-30°C 135K

50°C 3.9K

And here are the wanted specification:

T° Max error

-40°C 0.8°C

0°C 0.5°C

40°C 0.3°C

I already though about a divider entering an OPamp and then using an active filter but I think it will not be precise enough due to the other resistance error of the divider and of the error on the alimentation of the divider.

Does anyone have an idea to measure the thermistor with this precision?

Precision I need to convert the thermistor value into a 0-10V voltage.

Edit: It will actually be directly read with a DAQ, then it is okay now I do not have to do the schema anymore. Thank you all

Answer 1

Let Vs be a low voltage precision bandgap reference voltage regulator. Let R3 = 10K or whatever the room temperature spec for the thermistor is given. Let R1=R2= precision resistors. The instrument Amp configuration will need bipolar supplies or a reference level offset in the output stage.

Modern thermal sensors however are more accurate in one chip such as; The INA330 offers excellent long-term stability, and very low 1/f noise throughout the life of the product. The low offset results in a 0.009°C temperature error from –40°C to +85°C. (not including sensor error)

There are many other low cost temp sensors in 1 chip that are accurate to 1 deg and much cheaper.

Answer 2

How are you planning to measure the resistance? Using a DVM or panel meter? With an A/D converter on a computer?

You can purchase thermistors with accuracies specified to within 0.1C, but besides that, here are some things to consider.

• To avoid reference voltage error, either measure the reference before taking a thermistor measurement, or use the same reference for your A/D converter
• For small numbers of units, you can measure the actual resistance in the other leg of the divider and use that instead of just its "face value"
• Limit the amount of current through the thermistor to reduce the amount of self heating
• Instead of a piecewise-linear or 2-point approximation, use the Steinhart-Hart equation to convert resistance to temperature

Our instruments use incubators that are temperature controlled to within a few tenths of a degree and this requires measuring the temperature to a precision of much less than that, so it's certainly possible. Semiconductor sensors have better linearity, but I don't remember seeing any that advertise an absolute accuracy of +/- 0.1C