Abstract: In order to realize high sensitivity detection of low concentration NO, a tunable diode laser absorption spectroscopy (TDLAS) detection system is designed by using a quantum cascade laser (QCL) with a central wavelength of 5.18 µm and a single optical path gas with an optical path of 25 cm. In order to reduce the influence of system noise so as to improve the sensitivity of NO detection and reduce the detection limit, the empirical mode decomposition (EMD) algorithm is used to filter the second harmonic (2 f ) signal of wavelength modulation spectroscopy (WMS) to identify and remove the high-frequency noise and optical fringes hidden in the detected signal, and then the noise reduction effect of (2f) signal of EMD algorithm is compared with that of the other common filtering methods. Based on this, NO continuous monitoring experiment is carried out, and Allan variance analysis is used to compare the stability and detection limit of the system before and after EMD filtering. The results show that the response linearity of the detection system reaches 0.999 after EMD noise reduction. The fluctuation range of NO detection concentration at 30 × 10-6 is reduced from 29:424 × 10-6 ∼ 33:184 × 10-6 to 29:585 × 10-6 ∼ 31:273 × 10-6, and the relative error range is reduced from 0:17% ∼ 10:61% to 0∼4.24%. In the average time of 1.5 s, the detection limit of the system is reduced from 653 × 10-9 to 442 × 10-9. Under the optimal average time, the detection limit of the system is reduced from 272:7 × 10-9 to 185:5 × 10-9, which proves that this method can effectively improve the detection accuracy and flexibility of the detection system and further reduce the detection limit.

Key words: spectroscopy, filtering and noise reduction, empirical mode decomposition, low NO concentration, Allan variance