Thermistor interface design

[ernestmyname]

I will be using a thermistor to measure temperature with a microcontroller. I plan to use the controllers 10 bit ADC. I want to display the temperature to tenths of degrees so I need the resolution to be as high as possible.

The chip will be powered by 5 volts. Without any other circuitry, by using a common voltage divider, I can get a 2.4 volt difference between the min and max temperature. The voltage range with a voltage divider circuit with maximum voltage swing is 2.01-4.41 volts measured at the ADC.

Would it be better to use a current source circuit rather than using the 5V from the uController? I will only be checking the voltage every two seconds to prevent excessive thermal heat from the temperature readings. Or what would be the best way to scale 2.01V-4.41V to 0-5V? I would like to measure multiple sensors which will have different voltage ranges. So I would rather not change Vref of the controller for each one. I considered using the thermistor in an opamp circuit to vary the gain. I am eager to hear how you guys would approach this problem.

Thanks in advance!

I may use this thread to post my progress. I will be using SPICE modeling to simulate before I build anything.

 

[alexan_e]

I don't think it is possible to get the accuracy that you want with a thermistor, for the most accurate results you should probably use a sensor like LM35 Temperature sensor tutorial - Using the TMP36 / LM35 or a digital sensor like DS18B20

For thermistor you can check **broken link removed**

Alex

 

[kak111]

I plan to use the controllers 10 bit ADC. I want to display the temperature to tenths of degrees so I need the resolution to be as high as possible.

for accuracy i recommend you to use TMP36 or some equivalent

Look picture you see the reason...................

View attachment TMP36_versus_NTC_01.pdf


Datasheet......................
https://www.analog.com/static/imported-files/data_sheets/TMP35_36_37.pdf


Regards
KAK

 

[ernestmyname]

I have looked at temperature IC's like the TMP35 but the datasheet claims +/- 1 degree C. But apparently that is without calibration. The accuracy graph looks very promising.

If I were to use one of the IC's listed, I would go with the TMP37. It has the largest voltage swing over the temperature range I am interested in (70-100C). This seems like a very good starting point for my project. I have instruments to calibrate my system with that are good to +/- .01 degrees C so I should get a good idea of how precise these IC's are.

Thanks for the input! This is going to be a fairly large project. I will be interfacing several 4x 7-segment displays and some SSR's to control some 240V heating elements. I will also be coming up with some type of PID type control scheme, so I'll be back looking for more ideas soon!

 

[kak111]

For more resolution....
Have you check for example NXP`s SE95
it`s 13bit sensor with I²C - bus

**broken link removed**

KAK

Ps. Accuracy needs calibration

 

[alexan_e]

The datasheet says that the temperature resolution is 0.03125°C but the actual accuracy as it says is +- 1°C (-25°C to 100°C)
In comparison DS18B20 is 12Bit so the resolution is lower (0.0625°C) but the accuracy is +-0.5°C (-10°C to 85°C) in a slightly lower range

Alex

 

[ernestmyname]

Of those options I am still thinking the TMP37 makes the most sense for my scenario. If you look at the error chart on the datasheet it has, at max, .2 degrees C error. Since I can calibrate it to .01 degrees C, it should be plenty for my application. I may build more of these circuits in the near future but for now I will appreciate the simplicity of interfacing with the TMP37.

[edit] I just reviewed the chart and I am referring to "typical" error. The max and min accuracy error is much larger. But I will probably use the TMP37 for now since it is very cheap and it will be enough to get my controls tuned.

 

[alexan_e]

According to the datasheet the TMP37 provide typical accuracies of ±1°C at +25°C and ±2°C over the −40°C to +125°C temperature range.
Each way has its own disadvantages, if you decide to use the ADC then you will have the ADC accuracy error, if you go with the digital sensors then the the result is transfered digitally so you have only the sensor error.
I'm not sure if the digital sensor error is persistent, if it is then after the calibration you can have a very high resolution.

Alex

 

[ernestmyname]

I did some more research and decided to go with a RTD. I ordered a cheap sensor that I can use for testing. I am going to order a 22 bit serial ADC since the resistance varies so little.

The range I am most concerned with varies only from 127 to 133 ohm.

The ADC is the MCP3551. I am using app note AN1154 from Microchip as a guide.

**broken link removed**

Any ideas on how I can get more resolution from my limited resistance change?