| PhD Viva

Name of the Speaker: Mr. Sameer Ahmad Mir (EE19D418)
Guide: Prof. Deepa Venkitesh
Venue: ESB-244 (Seminar Hall)
Online meeting link: https://meet.google.com/ipv-hurq-ybz
Date/Time: 6th March 2025 (Thursday), 12:00 PM
Title: Low Complexity Algorithms and Methods for High-Capacity Coherent Optical Systems.

Abstract :

Capacity scaling is crucial in optical fiber communication to meet the rising demand for high-speed data transmission. One way to increase capacity is by using the higher-order modulation formats. However, as modulation order increases, the phase margin narrows, making the system more susceptible to impairments from transmission non-idealities. Common impairments in optical communication systems include laser phase noise, chromatic dispersion, polarization mixing, frequency offset, and transceiver IQ imbalance. Effectively addressing these impairments often requires joint equalization techniques, where multiple impairments are corrected using a single algorithm. This approach reduces computational complexity and minimizes the risk of error propagation between separate correction stages, improving system robustness and performance. In the first part of this thesis, a low-complexity digital signal processing algorithm that jointly corrects for laser phase noise while simultaneously tolerating the transmitter IQ imbalance is proposed. This is achieved by adapting reference constellation points through gradient descent adaptation. The algorithm is further modified to handle residual inter-symbol interference caused by the system's bandwidth limitations. For receiver-side IQ imbalance, a separate algorithm based on geometric parameter extraction is introduced. Both algorithms are validated through simulations on 16-QAM and 64-QAM signals and are experimentally verified on 16-QAM signals.

Probabilistic constellation shaping (PCS) has garnered significant interest for its ability to approach Shannon's capacity limit in optical communication systems. The second part of the thesis explores PCS and its advantages, focusing on various challenges and solutions. It examines the impact of transmitter IQ imbalance on probabilistically shaped signals, highlighting the distinct challenges compared to uniformly shaped signals. To address these issues, the thesis extends the use of the carrier phase recovery algorithm for the shaped signals, the performance of which is validated through simulations using 200 GBaud PM-64 QAM signals. While PCS is known for its tolerance to Kerr nonlinearity, resilience to distortions introduced by semiconductor optical amplifiers (SOAs) is also studied. In mQAM signals amplified by SOAs, outer constellation points typically suffer higher nonlinear distortion than inner points. However, in probabilistically shaped constellations using Maxwell-Boltzmann distributions, outer points occur less frequently, when compared to the inner ones, resulting in reduced overall distortion compared to unshaped signals. The thesis experimentally demonstrates low-distortion amplification of 32 GBaud 16QAM signals, analysing performance across various entropy levels. Results show that low-entropy signals are more tolerant to SOA-induced nonlinearity than high-entropy signals, further affirming the advantages of PCS in practical amplification scenarios.

To further scale the capacity achievable with traditional single-core fibers, multi-core fibers offer an effective solution by integrating multiple optical cores within a single fiber. This configuration enables the simultaneous transmission of independent data streams in each core, thereby substantially increasing the fiber’s overall transmission capacity. In such systems, inter-core crosstalk and polarization mixing present significant challenges, which can degrade signal quality. Accurately modelling these impairments is computationally complex. Techniques like coupled mode theory and coupled power theory are commonly employed to describe the propagation and accumulation of crosstalk along the fiber length. However, from a signal processing perspective, the focus is typically on determining the final crosstalk levels between cores at the output rather than tracking their evolution along the fiber length. In the final part of the thesis, two approaches are presented for modelling inter-core crosstalk and polarization mixing. The first approach uses the Kronecker product of two matrices to model crosstalk, with a mitigation strategy that leverages Kronecker product properties for low-complexity correction. The second approach employs random matrix generation with a single control parameter, adaptable to multi-core fibers with varying core counts. Lastly, transmission through a 4-core fiber is demonstrated over a length of 480 km using a recirculating loop.