PiicoDev Atmospheric Sensor BME280 - Raspberry Pi Guide

Hi @Zach
Thanks for that info. I will give the boards a dry bake followed by an ambient rehydration when the “heating equipment” is free. Since they are not working I guess there is nothing to lose.

Hi Fractal,

Just make sure there isn’t an exposed element radiating directly onto it, and that there is forced convection.

Without both there is a high chance the plastic begins to melt

1 Like

Thanks @Liam, @Fractal, @Zach, and @Liam120347

Some really great info there! @Fractal’s comments got me thinking about my own setup. I’ve been recording data from the BME280 and a precision temperature TMP117 every 2 seconds for over a year and I figured it would be worth comparing the data. I aggregated the data back to 1 minute averages over 8 months (February to October: ~300,000 paired data points) and had a look at the discrepancy (Both sensors were connected to the same RPi4B, daisy chained through I2C - so roughly 10 cm from each other and the pi).

The results were interesting! It seems like the two sensors give fairly similar results (Fig 1A), with the TMP117 being consistently warmer (BM280 had a median discrepancy of -0.529 deg C; Fig 1B). However the discrepancy of BME280 wasn’t distributed evenly across the temperature range of the TMP117. It seems that the difference between the sensors is highest at cooler temperatures, becoming almost identical as they both approach 35 to 40 deg C (Fig 1C). Looking at the discrepancy as a percentage of the TMP117, it seems the difference is at its worst around 10 to 15 deg C, where the BME280 approaches an error of 15% cooler (Fig 1D).


Figure 1 Temperature discrepancies between BME280 and TMP117

I also checked how the sensor discrepancy changed over time, to see if there was any degradation of the sensors. I couldn’t detect any change over the 8 month period that looked like sensor decay (Fig 2). The noise at the very beginning is interesting though. I suspect that is do with Perth WA having its driest ever 8 month period (October 2023 to end of April 2024). I’m not implying the sensors are being effected differently by drought, rather I suspect the reticulation system being used ~5m from the sensors during that period probably caused all sorts of fluctuations in temperature and humidity (Trying to save the garden!).


Figure 2 Changes to temperature discrepancy over time

Now obviously there’s a tonne of caveats to this kind of analysis, particularly in the way the sensors were setup. Both were openly exposed in a covered but outdoor area, and no care was taken to ensure the boards had adequate airflow, were out of direct sunlight, and weren’t receiving radiant/reflectant heat.

All in all, I’m not sure which board is more correct, or even if either are. Certainly I think the components likely have very different rates of heat output and thermal conductivity, and considering the BME280 is only rated to +/- 1 deg C, I’m pretty happy that they record very similarly. I’d love to compare these results to a more accurate sensor, and get a better feel for their overall accuracy.

Hope that helps,

Cam