Open dataset initiative for machine learning-based linearization in analog radio over fiber systems

Autor(es)

Pereira, L. A. M. ; Mendes, L. L. ; Cox, Mitchell A. ; Cerqueira Jr, S. Arismar

• Tipo de documento: Artigo Científico
• Fonte: Optics Communications; v. 590, p. 8-pg., 2025-10-01.

Resumo

In analog radio over fiber (A-RoF) systems, linearization is essential to mitigate nonlinear distortions. While many studies have explored both machine learning (ML)-based and traditional linearization schemes for A-RoF, they often do not provide the dataset used to train the ML models. We present a mathematical model for representing A-RoF systems and provide a comprehensive real dataset, which we make available via GitHub. We utilize a software-defined radio (SDR) approach to generate digital samples of the generalized frequency division multiplexing (GFDM) waveform, which are then applied to a real A-RoF system. This dataset, consisting of transmitted and received GFDM waveform samples, enables the training of both traditional and ML-based models, thereby supporting the development of robust linearization schemes. Furthermore, we conduct a performance evaluation of a ML-based digital pre-distortion (DPD) scheme, trained on both synthetic and real datasets. The linearization performance was assessed using key metrics, including the root mean square error vector magnitude (EVMRMS), normalized mean square error (NMSE), and adjacent channel leakage ratio (ACLR). The results show that the DPD scheme successfully reduced the EVMRMS from approximately 6% to 2%, along with improvements in NMSE and ACLR. Additionally, the dataset, in conjunction with the memoryless polynomial model presented in this paper, provides a robust framework for characterizing A-RoF systems. Researchers can leverage this publicly available dataset to model A-RoF systems and design novel linearization schemes, thereby minimizing nonlinear distortions and enhancing signal fidelity. (AU)

• Processo FAPESP: 20/05127-2 - SAMURAI: núcleo 5G inteligente e integração de múltiplas redes de acesso
• Beneficiário: Aldebaro Barreto da Rocha Klautau Junior
• Modalidade de apoio: Auxílio à Pesquisa - Temático
• Processo FAPESP: 22/09319-9 - Centro de Ciência para o Desenvolvimento em Agricultura Digital - CCD-AD/SemeAr
• Beneficiário: Silvia Maria Fonseca Silveira Massruhá
• Modalidade de apoio: Auxílio à Pesquisa - Centros de Ciência para o Desenvolvimento
• Processo FAPESP: 21/06569-1 - Tecnologias estratégicas para internet de alta velocidade
• Beneficiário: Evandro Conforti
• Modalidade de apoio: Auxílio à Pesquisa - Temático