CO2 sensors
AOSong ACD10
The AOSong ACD10 is a CO2 sensor which works on the principle of non-dispersive infra-red (NDIR) absorbance. CO2 absorbs light at specific wavelengths, and an NDIR sensor involves an infra-red source and two detectors with filters in front of them, one specific to the wavelength that’s absorbed by CO2 and one that’s not. The concentration of CO2 can be deduced by measuring the outputs of the two detectors and using the Beer-Lambert law to calculate the concentration of CO2.
Testing with this sensor yielded usable results, but its output is quite noisy and any readings below 400ppm are clipped to 400ppm, which means that any filtering applied to its output will produce inaccurate results at lower concentrations of CO2.
Infineon PAS CO2
The Infineon PAS CO2 CO2 sensor uses photoacoustic spectroscopy, a technique where infra-red light at a wavelength absorbed by CO2 is used to heat the CO2 in the sensor’s chamber and a MEMS microphone is used to detect the resultant pressure change. In testing, this sensor yielded good results with little noise.
Senseair Sunlight CO2
We’ve just received a Senseair Sunlight CO2 sensor, and we’re going to run this alongside the two sensors above for a few days and will publish the results in due course.
As of 2025-02-13, we’re running the three sensors outdoors for 24 hours. The AOSong and Infineon sensors were calibrated to 420ppm CO2 at the start of the test, in line with the value being reported by the Senseair one. Live data at https://co2.dave.cy
2025-02-15: looks like both the ACD10 and the PASCO2 did auto-calibrations which, in the case of the ACD10, pinned its output at 400ppm. Updated code to turn off auto-calibration and will run a further test today.