Super-resolution an optical frequency comb sampling via multi-dimensional signal separation

Electromagnetic sources show wide distribution, broad frequency coverage, and numerous quantities, posing challenges for traditional sensing techniques to achieve ultra-wideband, large-scale detection and localization. The “electromagnetic eye” imaging technique, inspired by the human eye, utilizes a Luneberg lens and a wideband optoelectronic sensing array as the electromagnetic “lens” and “retina,” respectively. This technique utilizes femtosecond optical pulse sampling reception to down-convert wideband signals, facilitating rapid, large range, and wideband sensing of multiple targets in complex electromagnetic environments. However, the limited aperture of the Luneberg lens results in diffraction-limited blurring, and optical down-conversion may lead to spectral aliasing, causing time-frequency-space overlap and reduced system resolution. In this paper, the frequency variation of the point spread function (PSF) in the wideband degraded images is analyzed, and a multi-dimensional joint super-resolution algorithm is proposed, which involves joint time-frequency-space diagonalization of eigenmatrices based on convolutional mixing array model. The concept is demonstrated through a four-sources imaging simulation achieving 2° resolution, breaking the Rayleigh limit 7.25 times. Furthermore, experimental results show 4-10 GHz imaging breaks the Rayleigh limit 4.5 times.