## Measuring temperatures with TMP03/04 (Part 3)

Here are some measurements performed by two TMP03 sensors.

One is attached to a dissipator the other one is not.

Acquisition parameters are: 1 cycle per second, 2 channels, no averaging, resolution 0.05°C

We can make two major observations from these plots: The top one (free standing sensor) shows a higher temperature: this is due to the self heating of the sensor. It is also trivial that this sensor is very sensitive to fast temperature changes compared to the other one. Let’s now see how much information we can extract from here. Applying some smoothing gives a cleaner view on temperature profile: this plot has been obtained after applying a box car filter type

This other one has been obtained after applying a gaussian filter type. The weighting factor allows a more accurate fit to fast changes.

Now, lets find a trend from the raw temperature profile. We will do that using a linear polynomial regression (third degree)

Even more explicit is the plot of residuals. They represent the distribution of data points around the “mean”, which is equivalent to noise.

From the noise plot, we can calculate the probability match function (pmf), often improperly nammed density spectrum

The distribution of matches in this plot is a perfect illustration of the expected thermal distribution.