My Ongoing Battle with Condensation (Part 2)

After finding moisture inside the Night Sky Pi dome, the next step was measurement. Two additional BME280 sensors were ordered to join the one already on hand. The hope was that ordering two together might land sensors from the same batch, which could read similarly enough to compare cleanly.

Bench check

Before deployment, the sensors were tested side by side on the same board.

One sensor’s address was changed by cutting the addr+1 pads so both could sit on the same I²C bus. Two copies of the environment checker were cloned, each pointed at a different sensor address and a different output file. A cron job ran both every 15 minutes for a day.

The graphs showed the occasional odd reading, but nothing more than about 0.5 °C out. Close enough for the job. The goal was not lab-grade calibration, just making sure there were no big differences between sensors.

Deployment

• One sensor went inside the Night Sky Pi dome, next to the camera.
• The second went outside in a Stevenson screen to protect it from the elements. A printable enclosure is available on Printables.
• The environment checker scripts were installed again and scheduled with cron every 15 minutes.

A third copy of the environment checker was set up to pull the same metrics from the OpenWeather API. The idea is to compare API data against the outside sensor and the dome sensor. If the API matches the outside sensor closely enough, the external hardware could be dropped to reduce complexity.

Data collection plan

The sensors and scripts have been live since 1/5/2024 and will run for 30 days to capture a mix of weather. Once there is a solid baseline, the results will guide the moisture fix inside the dome.

Progress for this series is grouped under the moisture battle tag.

Thanks for reading. If you have a clean way to compare low-cost sensor drift over time without a lab reference, share it so others can benefit.