Design and Application of a Controlled-Environment Calibration System for Low-Cost CO2 Sensing Devices

This study presents the design, development, and experimental validation of a portable CO 2 calibration system capable of generating reproducible concentration profiles under actively controlled air temperature and relative humidity. The platform integrates a modified environmental chamber, an automated gas injection system, and a centralized control and data acquisition architecture, enabling both fixed set-point and dynamic concentration profiles representative of real indoor environments. The performance of the calibration system was first evaluated independently to assess its stability, repeatability, and linearity. Repeatability tests conducted under fixed and dynamic operating conditions demonstrated a good level of reproducibility, with expanded uncertainties in the range of ±14–17 ppm and linear behavior across the 400–2000 ppm concentration range, confirming the suitability of the system for multi-point sensor calibration. Following chamber validation, the platform was applied to the calibration of a commercial low-cost CO 2 sensor as a case study. The sensor response was characterized against a reference instrument using a stepwise multi-point exposure protocol. Results revealed a systematic bias of 154 ppm and a sensitivity coefficient of 1.2 (R 2 = 0.99). After applying a linear correction derived from the calibration procedure, the sensor achieved an expanded uncertainty of ±26 ppm under the tested conditions. A subsequent 48-hour in-situ deployment confirmed its practical applicability, as the chamber-calibrated sensor reliably tracked real-world dynamic ambient fluctuations (R 2 = 0.96) with a RMSE equal to 66 ppm. Compared with conventional laboratory-grade calibration facilities, the proposed system offers a portable, cost-effective, and reproducible alternative that bridges the gap between high-cost laboratory systems and in-situ calibration approaches and supports the deployment of dense low-cost sensor networks.