Fast-OFDM-based intensity-modulation and direct-detection (IM/DD) has been proposed for deployment of cost-efficient optical access networks, due to its implementation simplicity and high spectral efficiency. In this article, the accuracy of the generalised memory polynomial (GMP) for the non-linear modelling of optical Fast-OFDM links is studied, including memory effects and considering different model parameters. After model validation using measured data of a 10 km single mode fibre link, the GMP is used for performance investigations of a distortion compensation approach to optical Fast-OFDM, for up to 16PAM modulation formats and different number of Fast-OFDM subcarriers. This study firstly reports the performance results of optical 16PAM-Fast-OFDM systems using either 2PAM- or 4PAM-based training signals for digital post-distortion and FFT-based channel estimation, and firstly investigates the influence of the zero padding (ZP) length on the performance of optical Fast-OFDM. Excellent performance improvements are achieved using the proposed distortion compensation scheme, relative to conventional system implementation.

中文翻译：

---

光学快速正交频分复用系统中的非线性建模和失真补偿

基于快速 OFDM 的强度调制和直接检测 (IM/DD) 因其实现简单且频谱效率高而被提议用于部署经济高效的光接入网络。在本文中，研究了用于光 Fast-OFDM 链路非线性建模的广义记忆多项式 (GMP) 的准确性，包括记忆效应和考虑不同的模型参数。使用 10 km 单模光纤链路的测量数据进行模型验证后，GMP 用于光学 Fast-OFDM 失真补偿方法的性能研究，适用于高达 16PAM 调制格式和不同数量的 Fast-OFDM 子载波。本研究首次报告了使用基于 2PAM 或 4PAM 的训练信号进行数字后失真和基于 FFT 的信道估计的光学 16PAM-Fast-OFDM 系统的性能结果，并首先研究了零填充 (ZP) 长度的影响光快速正交频分复用 (Fast-OFDM) 性能的研究 相对于传统的系统实现，使用所提出的失真补偿方案实现了出色的性能改进。