MEMS-based photoacoustic spectroscopy
Thanks to its experience with MEMS microphones and through experimental processes, Infineon has succeeded in developing a new CO2 sensor based on photoacoustic spectroscopy (PAS) – a physical method that is suitable for detecting gas components in a mixture and, for example, determining the CO2 concentration in indoor air.
Photoacoustic spectroscopy utilises the fact that gas molecules only absorb light with a specific wavelength; in the case of carbon dioxide, this wavelength is 4.2µm. An infrared source with an optical filter supplies the gas with energy in a rapid succession of light pulses at precisely this wavelength. This leads to the rapid heating and cooling of a gas sample, which in turn leads to thermal expansion and contraction. The sound generated by this can be recorded with a microphone, evaluated and used to draw conclusions about the amount of CO2 in the gas. The higher the CO2 concentration, the stronger the signal. The use of a highly sensitive MEMS microphone as a detector allows for significant miniaturisation compared to NDIR-CO2 sensors.
Challenges in sensor development
The Infineon CO2 sensor integrates a photoacoustic transducer with detector, infrared source and optical filter on a printed circuit board. The sensor has a small microcontroller for on-board signal processing, sophisticated algorithms and a MOSFET for operating the infrared source. A modulated IR light source radiates onto the gas mixture in the sampling chamber. The CO2 present absorbs the IR light, heats up and increases the pressure in the sampling chamber and these pressure changes can be measured by a MEMS microphone.
A major challenge in developing a PAS-CO2 sensor was to push the performance of the microphone to its limits and minimise system noise, i.e. to isolate the MEMS detector from external noise so that only the pressure change originating from the CO2 molecules in the chamber is detected. Infineon modelled the MEMS microphone response before prototyping some units to validate the modelling results.