Vishay Thermistor Reading Issues

Question

I am using an NCD relay control with 8 channel ADC. I have thermometer that is reading temp of my room to be 70F but the Thermistor Vishay NTCLE100E3 10k is reading 62.26F. The relay board has a built in temperature that reads 62.507. I have been using the Steinhart and hart equation as provided by vishay. I have seen a tutorial on youtube with someone using the same thermistor i have plugged to an arduino and his thermistor seems to be responding better than mine. I have measured the voltage between both resistors and i get 3.27v. I use Rthermisor = R10k*(5.0/Vo)-1 where Vo is the voltage between thermistor and resistor. These are the values i use for the Steinhart eq. A1= 3.354016E-03 B1=2.569850E-04 C1=2.620131E-06 D1=6.383091E-08 my beta value is 3977.

At this point I am not sure what I am doing wrong or why the readings are so off. Should I change the resistor value? I was reading something about self heating but didnt quite understand.

^{simulate this circuit – Schematic created using CircuitLab}

Answer 1

R25 error is defined by the p/n suffix after the value 103.

This is likely the cause of both errors.

Replace in SAP by (fine print in spec)

J=5%,H=3%,G=2%.

Adjust offset for R25 error and B tolerance errors as required by your unstated system design spec.

Or get G tolerance part.

Or recompute worst case errors and decide how to compensate with series R (select on test jumpers) to thermistor or other methods.

But always create an acceptance criteria 1st before part selection and then do a tolerance stack up of all parts.

Answer 2

Your error is obviously way too large to be accounted for by tolerances (will typically be within a fraction of a degree C).

It is reading low, which implies that the voltage at the input is lower than expected. Since the ADC input impedance is given (not clearly, but the number is mentioned) as 10K, this is to be expected.

Either correct your equation for the loading or add a buffer amplifier.

---

For future reference, you can easily diagnose problems like this yourself if you measure the voltages at the input to the "black box" system with a voltmeter and do some back-of-envelope calculations. Always try to find a "test point" somewhere in the system and divide it. A single point problem can be located in \$\log_2(n)\$ tests