The global communications network has become a pervasive and critical item of everyday life, spawning and enabling countless worldwide services that went from nonexistent to must-have in less than a decade. Its implementation makes considerable use of optical transmission systems, which are the physical medium of choice for most non-wireless links, being capable of high data rates over long distances. However, the potential of optics is still underexploited, and can help a smarter network meet the simultaneous challenges of ever-higher data rates, network switching, and the "last-mile" access network. <p> Very high data rates were achieved in optical transmissions in the late 1990s especially through wavelength-division multiplexing (WDM) over the C and later the L spectral bands. For some time, the way to increase data rates was forecast to be higher symbol rates per wavelength, for which optical-to-electronic (O-E) conversions are a speed bottleneck. This required all-optical functionalities, especially to process optical time-domain multiplexed signals. In that line, I contributed to ultrafast clock recovery using opto-electronic phase-locked loops. <p> However, the recent comeback of coherent optical communications points to easier ways to increase the data rate by pushing towards higher spectral efficiencies, closer to the optical channel's Shannon capacity in the presence of certain physical impairments. Notably, some of my recent results suggest that polarization-dependent loss can be handled close to the limit thanks to a combination of space-time codes and more conventional error-correcting codes. <p> Switching is another bottleneck: the Internet's great versatility results in part from its packet-switching paradigm, but current optical networks are essentially circuit-switched using wavelength granularity. Packet-switching functionality is implemented purely in electronics, incurring numerous energy-inefficient O-E conversions and ballooning energy costs. <p> My work on all-optical functionalities included an all-optical label-processing scheme for switching nodes, though this approach would be subject to scaling problems in practice. More recently, my concern has shifted to hybrid switching nodes using electronic buffers to supplement an optical switching matrix. My current studies show great improvements of their sustainable load compared to all-optical switches at a given packet-loss probability. <p> Access network is the last stronghold where optical transmissions are not quite dominant yet. The focus there is on cost effectiveness and resource sharing, especially in passive optical networks (PONs). In order to bring WDM to PONs, I contributed to a pulsed continuum optical source that could have provided optical channels to multiple users simultaneously. More recently, I also oversaw work on reflective semiconductor optical amplifiers designed for colorless optical network units. <p> Finally, the challenge goes on for a better match between network functionalities and the untapped potential of optics. My focus is currently shifting towards cross-layer optical networking, requiring novel network architectures to break free from the electronic-centric layered-network model, and finally meeting the energy consumption problem square-on.
| Identifer | oai:union.ndltd.org:CCSD/oai:tel.archives-ouvertes.fr:tel-00983948 |
| Date | 26 November 2013 |
| Creators | Ware, Cédric |
| Publisher | Telecom ParisTech |
| Source Sets | CCSD theses-EN-ligne, France |
| Language | English |
| Detected Language | English |
| Type | habilitation ࠤiriger des recherches |
Page generated in 0.0021 seconds