A model of a miniaturised NDIR (Non Dispersive Infrared) gas analysis system, aiming to predict the final system specifications, is presented. It comprises the different elements of the NDIR detector, including a surface micromachined Fabry-Perot tuneable filter. These models have been used to estimate the response of the NDIR system to different gas mixtures. Multivariate regression methods like Partial Least Squares (PLS) allow the recovery of the true sample composition from the IR absorption spectra measured with the NDIR system, despite the limited selectivity of the filter. Combining model and data processing permits one to predict the effect on the final system specification of design parameters. Here we compare the effect of the technology used for the filter on the system errors.
Published in:
Sensors, 2002. Proceedings of IEEE
(Volume:2
)
Date of Conference: 2002