رسائل ماجيستير
English
2026
TRADITIONAL AND MACHINE-LEARNING EQUALIZATION TECHNIQUES FOR BANDWIDTH-LIMITED SHORT-REACH OPTICAL COMMUNICATION CHANNELS
Abdullah Khawatmi, Mohamed Atef Abdelaal
كلية الهندسة-جامعة الإمارات العربية المتحدة · الامارات
This thesis investigates equalization techniques for bandwidth-limited short-reach optical communication systems, with a focus on Visible Light Communication (VLC) and Step-Index Plastic Optical Fiber (SI-POF) links. Commercial light-emitting diodes and photodiode receivers impose severe bandwidth constraints, inter-symbol interference, and noise sensitivity, which fundamentally limit achievable data rates. The work addresses these impairments through systematic evaluation of traditional digital signal processing–based equalizers and modern machine-learning-based post-equalization methods. The primary aim of this thesis is to enhance the achievable data rate and reliability of commercial short-reach optical links while maintaining practical computational complexity. Specifically, the objectives are to (i) design and experimentally validate hybrid traditional equalization pipelines for real-time VLC systems, (ii) investigate machine-learning post- equalizers for VLC and compare them with conventional approaches, and (iii) perform a performance–complexity study of major neural-network families for SI-POF channel equalization. The research combines analytical modeling, experimental measurements, and data-driven learning. For VLC systems, zero-forcing pre-equalization and LMS adaptive post-equalization are implemented using a commercial LED-based testbed with real-time performance evaluation. Machine-learning post-equalizers, including artificial neural networks and a proposed Binary Neural Tree (BNT) architecture, are trained on experimentally acquired data. For SI-POF systems, an emulated channel model is employed, and ANN, CNN, and LSTM-based equalizers are evaluated under varying noise and signal-to-noise ratio conditions. Experimental results demonstrate that hybrid traditional equalization enables reliable high-speed VLC transmission with significant bandwidth extension compared to unequalized links. Machine-learning post-equalizers achieve improved bit-error-rate performance at higher data rates, with the proposed BNT equalizer providing comparable accuracy to conventional ANN models while requiring substantially lower computational cost. Simulation results on SI-POF channels show that neural-network-based equalizers exhibit different robustness and complexity trade-offs, particularly under low-SNR conditions. This thesis provides a unified experimental and simulation-based evaluation of traditional and machine-learning equalization techniques for commercial short-reach optical systems. A key contribution is the introduction and validation of the Binary Neural Tree equalizer, which achieves efficient performance–complexity trade-offs suitable for practical deployment. Additionally, the work establishes standardized performance and computational comparisons across multiple equalizer families. The research addresses the lack of experimentally validated, complexity-aware equalization studies using commercial LEDs and realistic short-reach optical channels. It bridges the gap between high-performance machine-learning equalizers and practical real-time implementation requirements for VLC and SI-POF communication systems.