I have started working with some BME680 sensors to evaluate indoor air quality.
I had two boards, sourced from Adafruit, operating on two identical setups, each also including an SCD-41 CO2 sensor to see how the readings compared with each other over time. I have had these boards running for over a week now. The sensors are powered off of a USB supply through the ESP32, with added filtration on that rail by a 10 µF electrolytic, and a 0.14 µF ceramic capacitor. USB supply voltage is 4.97 V. The Adafruit boards then have an onboard AP2112K linear regulator supplying 3.3 V with their own 10 µF and 0.1 µF filter caps, and a level shifter.
At first, the results seemed good, with the value from the BME680 roughly but predictably following the value from the SCD-41. Notable exceptions being the sensitivity of the BME680 to VOC containing products which then caused an understandably higher reading.
Two days ago, the values reported by both BME680s jumped from just under 700 ppm to 3600 and 4400 ppm, while the actual CO2 levels in the environment did not change. Over the last two days these values have drifted down somewhat, but are still reading thousands of ppm in an environment that is sitting under 600 according to the SCD-41 on the same hardware, and a separate SCD-30 based sensor.
To determine if there was some unknown source of airborne contamination, I fired up a 3rd (new) BME680, as well as an SGP30 VOC sensor, also deployed on the same ESP32-S3 platform.
The new BME680 reading roughly tracks with the other sensors, the SGP30 reports 400ppm eCO2, and only 9 ppb of total VOCs. The SCD-41s continue to report as they always have.
What could have caused this sudden increase in readings from the two BME680s? It has been several days so I would have thought that the background calibration would have leveled things out by now, but it has not.
The attached graph shows the original two BME680s in blue and yellow, tracing along well with the value reported by the SCD-41. Friday night those two values spiked and although they are slightly trending down have not recovered. Coming in with the reddish/orange color is the newly opened BME680 which reads alongside the actual CO2 value as I would expect. The second spike in the yellow monitor was caused by a little bit of alcohol vapor on my fingertip as I tested the SGP30, as its initial reading of less than 10 ppb seemed suspect, but it does seem to be working properly.
Any insight would be greatly appreciated.
