Temp sensor EBK temperature differences

Version 1
    Question: Why are the four sensors showing different temperatures in temp sensor EBK?




    1. Thermocouple measures the air temperature while thermistor and RTD measure board temperature. The air temperature is slightly lower than the board temperature. Move the thermocouple junction near the thermistor or RTD on board and you’ll see an increase in temperature
    2. None of the sensors on board are calibrated. They can’t be calibrated because the EBK has no place to store calibration constants.
    3. Each sensor has various sources of errors. For instance, the RTD has the following sources of error (reproduced from AN70698 – Temperature Measurement with RTDs)

          Error Source


          Error value at 150 °C (0.1% reference resistor, class B RTD)


          Error Value at 150 °C (Both reference resistor and RTD calibrated)


          Offset Error/drift


          0 °C


          0 °C


          Gain Error/drift


          0 °C


          0 °C


          ADC INL*


          0.79 °C


          0.79 °C


          Error due to reference resistor (Ambient Temperature = 25 °C)


          0.43 °C


          limited only by calibration accuracy and Reference resistor temperature coefficient(very accurate)


          Error due to RTD interchangeability (Class B RTD)


          1.5 °C


          Limited only by calibration accuracy (very accurate)


          Polynomial fit error (5th order polynomial)


          0.0003 °C


          0.0003 °C


          RTD self heating error


          < 0.01 °C


          <0.01 °C




      The above table shows that the maximum error of an uncalibrated RTD can be 2° C (ignoring the ADC INL which is unpredictable) at 150 °C. This number will be lower at 25 °C (probably around 0.5 °C). But this number will be much lower for a calibrated RTD as shown in the table. This is not PSoC limitation but RTD limitation, which can be calibrated. Similarly, the current firmware doesn’t use a fifth order polynomial, but uses a first order polynomial which can be inaccurate up to 0.4 °C at 25 °C. This firmware will be corrected in the next firmware update to include the component which provides the fifth order polynomial.
       We can construct a table similar to this for all the sensors for any temperature. In other words the error limits at any temperature for any sensor can be readily calculated. 
       The above reasons account for the differences in temperature shown by each sensor. Note that the prime source of error in all the cases will be the sensor tolerence/interchangeability error.