High speed optical communications in silicon photonics modulators
|Abstract:||Optical communications based on silicon photonics (SiP) have become a focus of the recent research for future high speed optical network technologies. In this thesis, we investigate digital signal processing (DSP) approaches to combat the physical limits of SiP Mach-Zehnder modulators (MZM) driven at high baud rates and exploiting advanced modulation formats with coherent detection. In the first section, we present a novel adaptive pre-compensation method known as gain based iterative learning control (G-ILC, previously used in RF amplifier linearization) to overcome nonlinear distortions. We experimentally evaluate the G-ILC technique. Adaptation of the G-ILC, in combination with linear digital pre-compensation, is accomplished with a quasireal- time hardware-in-the-loop procedure. We examine various orders of quadrature amplitude modulation (QAM), i.e., 16QAM to 256QAM, and symbol rates, i.e., 20 to 60 Gbaud. Furthermore, we exploit joint digital and optical linear pre-compensation to overcome the bandwidth limitation of the system in the higher baud rate regime. In the second section, inspired by lower symbol rate G-ILC results, we push the baud rate beyond the bandwidth limit of the SiP system. For the first time, we experimentally report record-breaking 16QAM and 32QAM at 100 Gbaud in dual polarization back-to-back transmission. The optimization is performed on both MZM operating point and DSP. The G-ILC performance is improved by employing linear multiple input multiple output (MIMO) equalization during the adaptation. We combine our innovative nonlinear pre-compensation with post-compensation as well. Via dual polarization emulation, we achieve a net rate of 833 Gb/s with 32QAM at the forward error correction (FEC) threshold for 20% overhead and 747 Gb/s with 16QAM (7% FEC overhead). In the third section, we experimentally present a digital pre-compensation algorithm based on a one-dimensional lookup table (LUT) to compensate the nonlinearity introduced at the transmitter, e.g., nonlinear frequency response of the SiP MZM, digital to analog converter and RF amplifier. The evaluation is performed on higher order QAM, i.e., 128QAM and 256QAM. We examine reduction of LUT complexity and its impact on performance. Finally, we examine the generalization of the proposed pre-compensation method to data sets other than the original training set for the LUT.|
|Document Type:||Thèse de doctorat|
|Open Access Date:||13 February 2020|
|Collection:||Thèses et mémoires|
All documents in CorpusUL are protected by Copyright Act of Canada.