# NTC thermistor

#### Missio2468

Joined Mar 18, 2022
37
I am designing a temperature monitoring instrument using 10k NTC thermistor. I have made the Circuit design which is pretty much linear for certain range(Considering calibration ) . The issue that I face is that the ntc thermistor has 5 % resistance tolerance which can give approximately more than 1 degree offset.I want my sensor circuitry to be as accurate upto 0.3 degree celcius. As of now I am using voltage divider with 5.62 k ohm resitor in series and measuring voltage across 5.62k ohm resistor.

#### crutschow

Joined Mar 14, 2008
31,112
You need to calibrate the circuit at a known temperature.
You could add a pot in series with one of the resistors to allow adjustment.

If you post a circuit schematic, we may be able to suggest better solutions.
Does the output go to a microprocessor?

#### panic mode

Joined Oct 10, 2011
2,360
"...which can give approximately more than 1 degree offset.I want my sensor circuitry to be as accurate upto 0.3 degree celcius."

what is the range? do you understand the difference between accuracy and precision? when you say "as accurate" it suggests that you are ok if result is within the same 1 degree C.

#### Ian0

Joined Aug 7, 2020
6,674
I am designing a temperature monitoring instrument using 10k NTC thermistor. I have made the Circuit design which is pretty much linear for certain range(Considering calibration ) . The issue that I face is that the ntc thermistor has 5 % resistance tolerance which can give approximately more than 1 degree offset.I want my sensor circuitry to be as accurate upto 0.3 degree celcius. As of now I am using voltage divider with 5.62 k ohm resitor in series and measuring voltage across 5.62k ohm resistor.
There's only one answer, and it's simple: buy a better thermistor. Epcos B57861S0103F045 has 1% tolerance.

If you have plenty of thermistors, then you could put several in series or parallel. If the error is distributed randomly, then the accuracy will improve with the square root of the number of devices used.

• Kjeldgaard

#### MrChips

Joined Oct 2, 2009
27,670
The difference between 5% and 1% tolerance refers to the variability from one device to another.
If you are making more than one unit and don't want to have to calibrate each unit then get 1% devices or better.

If you are building only one instrument then it does not matter if you use 5% or 1% device. You still need to calibrate the instrument over the full operating temperature range. Thermistors are inherently non-linear devices.

• drjohsmith

#### ag-123

Joined Apr 28, 2017
263
you can try calibrating with Steinhart–Hart equation
https://en.wikipedia.org/wiki/Steinhart–Hart_equation

as for thermistors, I've got a 3d printer kit where the delivered deemed 100k NTC thermistor measures 50k at the specified reference temperature commonly deemed 25 deg C. try that for 'accuracy', > 50% wrong.
The only way to use it is to calibrate it against the equation.
this may be likely to happen if one buys on the online "flea" markets, they always state 3950 thermistor with no further specs.
and it is a gamble, some of they are good and literally measures 100K if you ordered 100K, then there are the exceptions.

#### Missio2468

Joined Mar 18, 2022
37
You need to calibrate the circuit at a known temperature.
You could add a pot in series with one of the resistors to allow adjustment.

If you post a circuit schematic, we may be able to suggest better solutions.
Does the output go to a microprocessor?
Yes my circuit goes to microcontroller.
Also the 10 k ntc thermistor is in series with 5.6K ohm resistor(due to 33-44 degree celcius). 5V is applied and analog signal is taken across 5.6 k ohm resistor.

#### Missio2468

Joined Mar 18, 2022
37
"...which can give approximately more than 1 degree offset.I want my sensor circuitry to be as accurate upto 0.3 degree celcius."

what is the range? do you understand the difference between accuracy and precision? when you say "as accurate" it suggests that you are ok if result is within the same 1 degree C.
Range is from 33 degree celcius to 44 degree celcius. I want my circuit to display temperature to precise to about 0.1 degree C

#### Missio2468

Joined Mar 18, 2022
37
There's only one answer, and it's simple: buy a better thermistor. Epcos B57861S0103F045 has 1% tolerance.

If you have plenty of thermistors, then you could put several in series or parallel. If the error is distributed randomly, then the accuracy will improve with the square root of the number of devices used.
Will think on that

#### Missio2468

Joined Mar 18, 2022
37
you can try calibrating with Steinhart–Hart equation
https://en.wikipedia.org/wiki/Steinhart–Hart_equation

as for thermistors, I've got a 3d printer kit where the delivered deemed 100k NTC thermistor measures 50k at the specified reference temperature commonly deemed 25 deg C. try that for 'accuracy', > 50% wrong.
The only way to use it is to calibrate it against the equation.
this may be likely to happen if one buys on the online "flea" markets, they always state 3950 thermistor with no further specs.
and it is a gamble, some of they are good and literally measures 100K if you ordered 100K, then there are the exceptions.
had tried out using stein-hart equation but I do not have any reference which can tell me different temperatures. Also the digital temperatures are not accurate as they seem for calculation purpose.

#### Ian0

Joined Aug 7, 2020
6,674
had tried out using stein-hart equation but I do not have any reference which can tell me different temperatures. Also the digital temperatures are not accurate as they seem for calculation purpose.
You need three temperatures for the Steinhart-Hart equation to work.
Do you have a datasheet for your thermistor?
If so, look up the resistance values for 35°C, 40°C and 45°C and put them in the calculator.

#### MrChips

Joined Oct 2, 2009
27,670
The transfer function will be very close to linear if you keep the temperature range small enough.

How have you determined that your setup is not accurate enough for your purposes?

I can only assume that you are comparing your results against a known reference, e.g. a digital thermometer?
Hence you only need two pairs of calibration points, one at a low temperature and another at a higher temperature.
Post your calibration data, i.e. known reference temperature and ADC counts (not the temperature your unit is indicating).

#### BobTPH

Joined Jun 5, 2013
6,072
There are two temperatures you can easily produce for calibration. Ice water at 0 and boiling water at 100.

Bob

#### panic mode

Joined Oct 10, 2011
2,360
true but both points are out of range and in that case some form of linearization would be needed.

#### ag-123

Joined Apr 28, 2017
263
had tried out using stein-hart equation but I do not have any reference which can tell me different temperatures. Also the digital temperatures are not accurate as they seem for calculation purpose.
to calibrate stein-hart equation
https://en.wikipedia.org/wiki/Steinhart–Hart_equation
you would need a reference i.e. an accurate thermometer and to measure the temperatures and voltages and after that fit the equation against stein-hart equation to get the parameters A, B, C

there is an example, online utility as like mentioed by Ian0

if one is 'lazy' to do that
quite commonly the manufacturer provides a B value for it e.g. 3950
https://en.wikipedia.org/wiki/Thermistor#B_or_β_parameter_equation
one way is to use the manufacturer prescribed B value work the equation to give temperatures.
then in the same way you would still need to calibrate it again by measuring temperatures with a separate thermometer and verifying its accuracy.
if you skip this calibration/verification step, then you are left with taking the manufacturer's spec for precision/tolerance e.g. 5% as far as the measurements goes.

note that while measuring say using an ADC, there can be wide band noise at the ADC input which is often RF (radio waves) signals, these can be as high as more than 10 mV which are often times represents more than a degree difference, especially if small voltages e.g. 3v range is used.
to deal with this it may be necessary to connect a capacitor so that it works as a RC low pass filter. this would reduce but not eliminate the RF noise problem.

Last edited:

#### Missio2468

Joined Mar 18, 2022
37
to calibrate stein-hart equation
https://en.wikipedia.org/wiki/Steinhart–Hart_equation
you would need a reference i.e. an accurate thermometer and to measure the temperatures and voltages and after that fit the equation against stein-hart equation to get the parameters A, B, C

there is an example, online utility as like mentioed by Ian0

if one is 'lazy' to do that
quite commonly the manufacturer provides a B value for it e.g. 3950
https://en.wikipedia.org/wiki/Thermistor#B_or_β_parameter_equation
one way is to use the manufacturer prescribed B value work the equation to give temperatures.
then in the same way you would still need to calibrate it again by measuring temperatures with a separate thermometer and verifying its accuracy.
if you skip this calibration/verification step, then you are left with taking the manufacturer's spec for precision/tolerance e.g. 5% as far as the measurements goes.

note that while measuring say using an ADC, there can be wide band noise at the ADC input which is often RF (radio waves) signals, these can be as high as more than 10 mV which are often times represents more than a degree difference, especially if small voltages e.g. 3v range is used.
to deal with this it may be necessary to connect a capacitor so that it works as a RC low pass filter. this would reduce but not eliminate the RF noise problem.
Thank You But have tried everything that is available on Internet. I don't want to calibrate everytime I change or replace my thermistor.

#### Missio2468

Joined Mar 18, 2022
37

#### Attachments

• 27.3 KB Views: 4