MHz-rate scanned-wavelength direct absorption spectroscopy using a distributed feedback diode laser at 2.3 µm

MHz laser absorption spectroscopy via diplexed RF modulation for pressure, temperature, and species in rotating detonation rocket flows

“…However, given that commercial laser drivers are designed and built to minimize the risk of instability over a wide range of loads, the phase error resulting from the components used to achieve this stability results in a lower bandwidth than might be achieved by the laser. For quantum cascade lasers, injection current controllers with current range on the order of 1 Amp have typical modulation bandwidths on the order of 10 5 Hz [29], lower than can be achieved with diode laser controllers [30].…”

Adaptively Optimized Gas Analysis Model with Deep Learning for Near-Infrared Methane Sensors

“…As seen in Figure a, the impact of different processing stages on the experimental results is clear: the methane sensor with the enhancement of AOGAM achieved the highest coefficient of determination ( R 2 = 0.99989) and is followed by the methane sensor assisted only by the NSF ( R 2 = 0.95122), while the methane sensor without any optimizing process was the most unreliable ( R 2 = 0.85738); the differences are statistically significant at P < 0.005. Unlike the AOGAM, the other two methods must convert the measured transmission spectra into absorbance spectra, then subsequently perform the Voigt profile fitting on the absorbance spectra, and finally calculate the concentration corresponding to the peak, which will inevitably introduce interferential errors. , The experimental results demonstrate that the method of extracting only the maximum absorbance value brings greater fluctuation and unreliability to the measured methane concentrations, especially in the dynamic measurements with time-varying noise. In contrast, the NCR that extracts information from the whole spectrum (dimensionally reduced spectrum) constructs afresh the mapping function between the full spectrum and its corresponding concentration, thus avoiding distortions introduced by processing with prior knowledge.…”

“…Unlike the AOGAM, the other two methods must convert the measured transmission spectra into absorbance spectra, then subsequently perform the Voigt profile fitting on the absorbance spectra, and finally calculate the concentration corresponding to the peak, which will inevitably introduce interferential errors. 18,40 The experimental results demonstrate that the method of extracting only the maximum absorbance value brings greater fluctuation and unreliability to the measured methane concentrations, especially in the dynamic measurements with time-varying noise. In contrast, the NCR that extracts information from the whole spectrum (dimensionally reduced spectrum) constructs afresh the mapping function between the full spectrum and its corresponding concentration, thus avoiding distortions introduced by processing with prior knowledge.…”

“…In contrast, the tunable diode laser absorption spectroscopy (TDLAS) 17 technique has higher robustness to various natural factors. The TDLAS technique can be divided into direct absorption spectroscopy (DAS) 18 and wavelength modulation spectroscopy (WMS). 19 The WMS is often used in trace gas detection since by increasing the detection frequency it is possible to greatly reduce lowfrequency noise (1/f noise) and improve detection sensitivity.…”

“…However, it has been proven that only the Voigt profile fitting is still insufficient, and the Dicke narrowing and speed-dependent effects should also be considered . In addition, more studies have demonstrated that taking the global spectrum into account can achieve more accurate concentration retrieval than referring only to the absorption peak. , Therefore, it is necessary to develop data-driven methods that more accurately derive the concentration. Deep learning enables computers to learn from experiences without actually modeling the physical rules that govern the spectral absorption process of gas molecules, which provides us a new solution for nonlinear inverse problems when relations between dependent and independent variables are not clear.…”