Scholarlink

<p>This work demonstrates a hybrid dual-polarization autoencoder (DPAE) framework for long-haul coherent optical fiber communication (OFC) by jointly optimizing a geometrically shaped <i>M</i>-ary quadrature amplitude modulation (QAM) constellation and a convolutional neural receiver over a physically scaled Manakov split-step Fourier method (SSFM) channel model. The long-haul optical link is implemented to emulate a realistic system by incorporating chromatic dispersion (CD), Kerr nonlinearity, attenuation, span-wise amplified spontaneous emission (ASE) noise, and laser phase noise into a fully differentiable channel layer, ensuring physics-consistent optimization. Furthermore, a deterministic electronic dispersion compensation (EDC) block is integrated before the neural decoder to bound the optimization state-space and mimic a practical coherent receiver, allowing the convolutional neural network (CNN) to focus exclusively on residual nonlinear and phase-noise impairments. The proposed framework is systematically evaluated over a 32&#xa0;GBd, 1000–4000&#xa0;km optical link across a wide operational grid of launch powers and effective signal-to-noise ratios (SNRs) by observing the pre-forward-error-correction (pre-FEC) bit-error rate (BER), symbol accuracy, generalized mutual information (GMI), and computation complexity. Performance analysis demonstrates that the hybrid transceiver can be optimized end-to-end at the bit level under realistic physical constraints, achieving minimum pre-FEC BERs of approximately <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(4\times 10^{-3}\)</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>4</mn> <mo>×</mo> <msup> <mn>10</mn> <mrow> <mo>-</mo> <mn>3</mn> </mrow> </msup> </mrow> </math></EquationSource> </InlineEquation> and symbol accuracies above 99% for 16/32/64-QAM setups. Furthermore, the hybrid architecture provides a balanced complexity–performance trade-off against baseline models, presenting a structured architectural blueprint for long-haul coherent optical transmission systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Hybrid Model Driven Dual Polarization Autoencoder for End-to-End Learning in Long-Haul Coherent Optical Systems

  • 期刊论文

  • 发表日期：

    2026年7月9日

查看全文
  • Waqas A. Imtiaz,
  • Abid Iqbal

摘要

This work demonstrates a hybrid dual-polarization autoencoder (DPAE) framework for long-haul coherent optical fiber communication (OFC) by jointly optimizing a geometrically shaped M-ary quadrature amplitude modulation (QAM) constellation and a convolutional neural receiver over a physically scaled Manakov split-step Fourier method (SSFM) channel model. The long-haul optical link is implemented to emulate a realistic system by incorporating chromatic dispersion (CD), Kerr nonlinearity, attenuation, span-wise amplified spontaneous emission (ASE) noise, and laser phase noise into a fully differentiable channel layer, ensuring physics-consistent optimization. Furthermore, a deterministic electronic dispersion compensation (EDC) block is integrated before the neural decoder to bound the optimization state-space and mimic a practical coherent receiver, allowing the convolutional neural network (CNN) to focus exclusively on residual nonlinear and phase-noise impairments. The proposed framework is systematically evaluated over a 32 GBd, 1000–4000 km optical link across a wide operational grid of launch powers and effective signal-to-noise ratios (SNRs) by observing the pre-forward-error-correction (pre-FEC) bit-error rate (BER), symbol accuracy, generalized mutual information (GMI), and computation complexity. Performance analysis demonstrates that the hybrid transceiver can be optimized end-to-end at the bit level under realistic physical constraints, achieving minimum pre-FEC BERs of approximately \(4\times 10^{-3}\) 4 × 10 - 3 and symbol accuracies above 99% for 16/32/64-QAM setups. Furthermore, the hybrid architecture provides a balanced complexity–performance trade-off against baseline models, presenting a structured architectural blueprint for long-haul coherent optical transmission systems.