Multi-Channel All-Optical Signal Processing Based on 3D Tensor Computing

Abstract

The capacity of single-mode optical fibre communication systems is approaching a
fundamental limit known as the Nonlinear Shannon Limit. Spatial division multiplexing
(SDM) leverages the transverse spatial dimensions of optical fibre to address the
capacity limitation through spatial parallelism. However, the dominant signal
impairments in SDM systems, namely the inter-channel crosstalk (IC-XT) and the
differential group delay (DGD), are detrimental to the SNR and necessitate the
expansion of multi-input multi-output (MIMO) signal processing from a 2D matrix to a
3D tensor operation. This expansion significantly raises the computational complexity,
power consumption, and cost of digital signal processors (DSP). As channel counts
increase, DSPs may become economically unviable and even technically impossible,
thus forming the primary bottleneck in current strategies for impairment equalisation.
Research efforts that focus on the management of SDM fibre characteristics mitigates
this challenge to some extent but does not fully solve the problem.
This project explores the intrinsic advantages of optics to overcome this bottleneck.
Firstly, this project establishes theoretical groundwork for a novel multi-channel alloptical signal processing methodology. For its physical implementation, an optical
MIMO processor comprising diffractive optical components is investigated, exploiting
the extensive spatial freedom of free space optics. It is demonstrated that a free space
optical MIMO processor can replicate all functions of an electronic MIMO processor
but may have scalability limitations and high insertion loss. Consequently, the focus
shifts to integrated optics. Benefiting from optimised building blocks, various versions
of integrated optical MIMO processors based on analogy optical filters and shown to
be able to restore the SNR to good levels have been proposed. However, in scenarios
with large channel counts, the integrated optics schemes pose fabrication challenges
due to the difficulty of monolithically integrating hundreds of components including
delay lines on a chip. To mitigate such challenges, an electro-optic accelerator has been
developed that implements core MAC functions of a digital MIMO equaliser. The
electro-optic complementary approach may relieve the need for large optical chips,
while still offering benefits including lower power consumption and processing latency.

Date of Award	3 Jul 2024
Original language	English
Awarding Institution	University of Bristol
Supervisor	Siyuan Yu (Supervisor), Martin J Cryan (Supervisor) & George Kanellos (Supervisor)