- February 27, 2021
By treating sensor sampling as a statistical challenge one can reason about how many samples to take from each sensor given a limited time interval for sampling (eg. 1s). It turns out that 30 samples per sensor is fine, and indeed, even one sample per sensor each second works. But life isn't only about usefulness, read on to follow my messy thoughts and perhaps learn a thing or two along the way.
- January 29, 2021
We set out to measure ozone but ended up getting better results measuring NO2. In the absence of a reference sensor, we correlated our measurements with the closest city air quality monitoring station. We used the correlation as a metric for how well the sensor was performing. Finally, we also combined this with studies about the indoor/outdoor ratio of NO2 to estimate indoor NO2 concentration.