Space–time domain equalization algorithm based on complex-valued neural network in a long-haul photonic-aided MIMO THz system

Abstract

The urgent demand for high-bandwidth wireless services in enhanced mobile broadband networks needs innovative solutions for mobile front-haul systems. The terahertz (THz) band offers a promising candidate for ultrahigh-capacity data transmission. This study investigates the integration of photonics-aided THz signal generation with MIMO and PDM technologies. We proposed a novel, to the best of our knowledge, space–time domain equalization algorithm based on MIMO-complex-valued neural networks (CVNN), which can preserve the signal phase and the relation between the X- and Y-polarization. We experimentally demonstrate the transmission of 60-GBaud PDM-QPSK and 30-GBaud PDM-16QAM signals over a 100-m 2 × 2 wireless MIMO link at 320 GHz with BER below 3.8 × 10−3 and 1.56 × 10−2 for QPSK and 16QAM signals, respectively. Compared with the MIMO-Volterra, our MIMO-CVNN has an advantage in terms of calculation complexity and decision accuracy due to its effective handling of phase information and inter-polarization relationships simultaneously.