Sub-THz SC-FDE MIMO Systems with Deep Neural Network Equalization for 100 Gbps of 6G

Sub-THz has received considerable attention for 6G wireless communications due to its large available bandwidth. Sub-THz multiple-input multiple-output (MIMO) system is expected to implement 6G data rates with more than 100 Gbps. Nonlinear distortion has always been a challenge for wireless communications due to the nonlinear characteristics of the RF chain especially for power amplifiers (PAs). The highest data rates cannot be obtained without effective compensation of nonlinear characteristics of the whole transmission channel. In this paper, we demonstrate and experimentally validate a deep neural network (DNN) nonlinear equalizer (NLE) to effectively compensate for nonlinearities of PAs on the SC-FDE MIMO system. Using the proposed DNN-NLE, we successfully transmit a 256QAM-3/4 signal with 107.52 Gbps for 2×2 MIMO in a chip rate of 10.24 GHz at a carrier frequency of 96 GHz over 36 m.