TMP007: Body temperature sensor - random noise of object temperature

Hi,


We are now developing a core body temperature sensor using TMP007, which measures the temperature on the eardrum. We have the expectation that ear bud environment is ideal due to close proximity, closed area, and relatively small operating range, so the accuracy we expect is in the order of 0.1°C. 

However, what we observed from our prototype is a random peak point. I think it is a random noise, but in severe cases the noise is sometimes measured up to 2°C.

We haven't done a calibration that can possibly adjust the final reading to a normal body temperature. All the configurations are default, while conversion time is 4 sec.

What we now do not understand is:

1. What are the vulnerabilities to noise such as light or wind?

2. Despite the ideal experimental conditions such as closed area, is it possible to ensure the accuracy of 0.1°C?

The stabilization time according to the change of experimental environment was sufficient. When measuring in a closed area for a long time, the local temperature (Tdie) seems to be able to secure a stability of 0.1 degrees. What is the best way to maximize the stability of the object temperature (Tobj) in an ideal space where these lights and winds are blocked? Can the Through-Hole Layout be the solution?

Thank you.