All-Electronic THz Direct Detection Wireless Transmission Enabled by a Neural-Network-Assisted Multi-Layer Cascaded Algorithm
Year: 2026
Type: article
Abstract: The Kramers-Kronig (KK) receiver is a promising architecture for high-speed terahertz (THz) direct detection (DD) systems, enabling high-order quadrature amplitude modulation (QAM). However, traditional KK algorithms suffer from severe linear and nonlinear impairments in THz links. This letter proposes a neural network (NN)-assisted multi-layer cascaded algorithm to achieve robust minimum-phase signal recovery. The proposed architecture strategically cascades a long-memory linear equalizer (LE) to capture a large observation window, an NN layer for high-dimensional nonlinear mapping and KK phase retrieval, and a subsequent LE layer for residual error correction. This hybrid design leverages the powerful nonlinear processing capability of NNs while utilizing the flanking LE layers to mitigate the exponential complexity surge typically associated with NNs when handling long-memory effects. Utilizing an all-electronic THz transmitter based on harmonic mixing and a DD receiver, the wireless transmission of 30 GBaud probabilistically shaped (PS)-32QAM signals with a line rate of 142.5 Gbps (net rate of 121.125 Gbps) is experimentally demonstrated. Furthermore, the proposed algorithm offers a 26.13% reduction in computational complexity compared to conventional methods at equivalent BER performance.
Institution: Peng Cheng Laboratory
Language: English
Topic: Optical Network Technologies
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $2,645
OpenAlex ID: W7221574410