Hi,
In the TMP102 data sheet there is stated that the accuracy is 0.5 C. It is said in the cover sheet. Later, there is a table where is stated that Accuracy (Temperature error) is TYP=0.5 and MAX=0.2. What those values exactly mean?
Second question. How good is stability over time? i.e what is your recommendation of calibration period. How often we should verify our devices if the accuracy requirement is +/- 1 C?
Kind Regards, Ari Pihlajamäki
Hello Ari,
The "typical" accuracy spec of ±0.5°C for the TMP102 is simply the average, or mean, value for the device. The "max" accuracy spec of ±2°C is the mean value ±3 (or more) standard deviations, also known as mean ±3-sigma. Usually the test limits are set to the mean ±4-sigma or mean ±5-sigma to improve yields, since ±3-sigma includes 99.73% of devices, but ±4-sigma includes 99.99% of devices and ±5-sigma is nearly 100% yield. See this article for a statistical discussion about the three-sigma rule: http://en.wikipedia.org/wiki/68-95-99.7_rule
As for the stability over time, this is not characterized. Our IC temp sensors use the ΔVBE method of an internal diode to measure temperature, so the stability of the temperature measurement is dependent on the stability of the current sources used to measure the diode's VBE per the equation below.
It's a safe assumption that the device is stable enough to not cause variations greater than or equal to 1LSB (0.0625°C). If you're performing a calibration, I think it's more important to have an accurate reference temperature measurement (RTD or similar) and allow enough time for the system to thermally stabilize before taking a measurement. Depending on the thermal properties of your system this can vary, but usually several minutes are needed. What are the set points you're using to calibrate the system? How well is the TMP102 thermally coupled to the PCB?
Best regards,
Ian Williams
