With the advancement of mathematical tools, computational spectrometers have gained new vitality by shifting the challenge of fine spectral detection to computing power and algorithms. This approach addresses the limitations of traditional spectrometers, including the need for precise dispersion elements, complex optical paths, and challenges in miniaturization. The traditional strategy of using static filter arrays with fixed spectral channels cannot flexibly or dynamically select spectral ranges based on the spectral characteristics of different targets. In this study, we present a computational spectrometer based on a mid-infrared photodetector integrated with a MEMS-FP tunable filter for dynamic encoding, enhancing reconstruction accuracy and flexibility compared with conventional static encoding approaches. In addition, we conducted in-depth research on reconstruction algorithms. Using the Enhanced Inverse Spectral Reconstruction algorithm, we optimize the selection of encoding channel numbers and the full width at half maximum of encoding curves. This balance ensures high reconstruction fidelity while preventing performance saturation, as an excessive number of encoding channels offers diminishing returns on reconstruction quality. Our system achieves a spectral fidelity as high as 99.8% with an encoding efficiency of 200%, ensuring a spectral resolution of 5 nm. Through our experiments, we successfully reconstructed the absorption spectra of aspirin and acetaminophen aqueous solutions within the 3–5 μm range, underscoring the spectrometer's potential in the pharmaceutical industry.
Li et al. (Tue,) studied this question.