The Role of FPGA Technology in the Design of Smart Transceivers for use in Optical Communications Systems

Abstract

This thesis looks at where the recent advances in Field Programmable Gate Array (FPGA) technology have been or could be used to complement and further the development of smarter optical devices in both the in-service and product development environments.

One approach to the development of cost-effective integrated components is the utilisation of low cost, low power circuitry, built as part of the module, which is capable of being repurposed from providing automated manufacturing orientated functions, such as characterisation and calibration, to operational control functions. Although these individual functions are well known, computationally efficient and low-cost implementations are required to enable competitive module pricing. This work examines the case of an optical Mach Zehnder Modulator (MZM) where the relationship between bias voltages and output power requires determining. With this aim, this study investigates the self-characterisation of this relationship with low cost and low power components which could be readily integrated into an optical module. In particular, the behaviour and capabilities required for automatic digital bias characterisation functionality, implemented in small gate count, low-cost FPGAs, are developed. The suitability of highly efficient implementations of DSP functions within the bias measurement function, such as digital filters, is tested, by investigating experimentally the use of a computationally efficient algorithm for computing a single component of a discrete Fourier transform, as a demonstration of the viability of using low-cost digital hardware to implement a circuit capable of monitoring the MZM transfer function. The results of this experiment are then used to investigate whether a simple Machine Learning model can be trained to extract characterisation data for the modulator.

Smarter optical modules, deployed in optical networks, are being developed as part of the solution to the increasing demand for bandwidth. There is a growing interest in augmenting existing communication systems, to carry greater bandwidth over the existing installed infrastructure, by adopting more parallel transmission techniques, such as Spatial Division Multiplexing (SDM), and by the application of advanced Digital Signal Processing (DSP) methods, e.g. compensating for non-linear impairments, allowing a higher transmit power and thus improving the Signal to Noise Ratio (SNR) which is required to support higher order modulation formats. Here, the role which can be played by FPGAs is investigated, through the first-time realization in FPGA technology of a real time, recurrent machine learning solution addressing optical channel non-linearities.

This thesis reports on three first-time achievements:
•the real time demonstration and the novel use of a filter based on the computationally efficient Goertzel algorithm to monitor an MZM transfer function.
•the implementation in FPGA technology of a biLSTM based equalizer for non-linear effects in optical communications systems. This is a novel use of the biLSTM architecture for this application.
•the use of a small size Neural Network to perform the Vπ and VOFFSET characterisation of an MZM optical component. In this work the novel application of ML, a NN with only 3 hidden layers of 8 neurons per layer and a small input layer of 82 inputs, is demonstrated as being able to determine the V and VOFFSET parameters of a real MZM.

The combination of the computationally efficient filter and the small NN capable of processing the output from the filter together show the ability of cost-optimised FPGAs to perform the analysis and processing of data required to carry out characterisation procedures on the photonic module, without the time overheads associated with large scale data transfers to a computer. Additionally, the implementation of a novel biSTLM equalizer demonstrates a practical design flow to realise the applications of NNs in FPGAs.

Date of Award	Dec 2023
Original language	English
Awarding Institution	Aston University
Supervisor	Wladek Forysiak (Supervisor), Paul Harper (Supervisor) & Richard Nock (Supervisor)