Characterization of Broadband Microwave Signals Using a Photonic Lightweight Speckle Pattern Spectrometer

The operating wavelength and modulation frequencies are the critical components of every communication system. Here, we present a speckle pattern (SP) fiber-based spectrometer using a nominally fixed 852-nm laser to measure and calibrate the wavelength and modulation frequencies of the SP produced in multimode fiber (MMF). The laser wavelength is finely tuned within the 100 GHz range and at a resolution of 2 pm (picometers) by appropriately varying the laser injection currents. The wavelength-dependent SP is calibrated by varying the laser current over a preset range, and a series of wavelength-dependent SP data are recorded. The laser is tuned to the middle of the recorded wavelength range for the modulation frequency calibration, and a standard lithium niobate (LiNbO $_{3})$ electrooptic modulator Mach–Zehnder modulator (MZM) is applied. With ten meters of MMF, the incoming radio frequency (RF) is modulated in 5 GHz steps up to 40 GHz at a constant wavelength of 852.555 nm, and the generated SPs are stored separately for spectral processing and training. The wavelength-dependent SPs are trained using artificial intelligence (AI), with a reported prediction accuracy of 98.7% at 2-pm wavelength resolution. To the best of our knowledge, this is the first proof-of-concept of a high-resolution, low-cost SP AI-based spectrometer (SPAIS) that has been experimentally reported to predict the exact values of modulated frequencies from the modulated SPs.