Advanced digital signal processing for next-generation flexible optical networks

To keep pace with the rapid expansion in data-exchange traffic around the world, optical networks are anticipated to provide flexibility to maximize utilization of the deployed optical fiber resources. On the other hand, digital signal processing (DSP) has been employed in coherent optical systems to enable 100G and beyond optical fiber networks. The goal of the thesis is to develop advanced DSP techniques for the flexible optical networks. With the reconfigured modulation formats in the systems, modulation classification (MC) is essential in the DSP to facilitate the sequential compensation modules which are modulation format-dependent. Based on the cumulative distribution function (CDF) of received signal's amplitude, an MC algorithm for M-ary quadrature amplitude modulation (M-QAM) formats with M = 4, 8, 16, 32, and 64 is proposed. Results show that the proposed algorithm achieves accurate classification at optical signal-to-noise ratio (OSNR) of interest and is robust to frequency offset and laser phase noise. Relying on the CDF of received signal's amplitude, a non-data-aided (NDA) OSNR estimation algorithm is developed for coherent optical systems employing multilevel constellations. It outperforms the state-of-the-art NDA algorithm in terms of performance and complexity. Furthermore, a joint OSNR estimation and MC algorithm enabled by support vector machine is designed. Compared to deep neural network-based joint estimation approach, the proposed algorithm achieves better performance with comparable complexity. In addition, a low-complexity two-stage carrier phase estimation algorithm is proposed for coherent optical systems with 16-QAM format. The proposed algorithm exploits the second power operation instead of the conventional fourth power to remove the modulation phase, which is enabled by constellation partition and rotation. Optical back-to-back experiments and numerical simulations are carried out to evaluate the performance of the algorithm. Results show that, compared with the conventional fourth power-based CPE algorithm, the proposed algorithm provides comparable tolerance to the carrier phase noise, with reduced complexity. Lastly, a novel transmission scheme is investigated for the open and disaggregated metro coherent optical networks, which impose the requirements for multiple user connectivities on the limited orthogonal frequency resources. Thus, it is desirable to provide connections simultaneously to various users in a non-orthogonal way. A transmission scheme based on the non-orthogonal sparse code multiple access in a digital subcarrier multiplexing is proposed. Compared to power domain-based counterpart, the proposed scheme supports more than 2 users without user pairing and clustering. The feasibility of the proposed scheme is verified through numerical simulations. Three scenarios with 2, 4, and 6 users over 1, 2, and 4 subcarriers, respectively, are considered. Performance evaluations show that in all scenarios, the proposed scheme attains bit error ratio lower than the forward error correction limits with the transmission ranges of interest in metro applications