| dc.description.abstract | Elastic optical networking (EON) is intended to offer flexible channel wavelength granularity to meet the requirement of high spectral efficiency (SE) in today’s optical networks. However, optical cross-connects (OXC) and switches based on optical wavelength division multiplexing (WDM) are not flexible enough due to the coarse bandwidth granularity imposed by optical filtering. Thus, OXC may not meet the requirements of many applications which require finer bandwidth granularities than that carried by an entire wavelength channel. In order to achieve highly flexible and fine enough bandwidth granularities, electrical digital subcarrier cross-connect (DSXC) can be utilized in EON. As presented in this dissertation, my research work focuses on the investigation and implementation of real-time digital signal processing (DSP) enabled DSXC which can dynamically assign both bandwidth and power to each individual sub-wavelength channel, known as subcarrier. This DSXC is based on digital subcarrier multiplexing (DSCM), which is a frequency division multiplexing (FDM) technique that multiplexes a large number of digitally created subcarriers on each optical wavelength. Compared with OXC based on optical WDM, DSXC based on DSCM has much finer bandwidth granularities and flexibilities for dynamic bandwidth allocation. Based on a field programmable gate array (FPGA) hardware platform, we have designed and implemented a real-time DSP-enabled DSXC which uses Nyquist FDM as the multiplexing scheme. For the first time, we demonstrated real-time DSP enabled real-time DSXC which uses resampling filters for channel selection and frequency translation. This circuit-based DSXC supports flexible and fine data-rate subcarrier channel granularities, offering a low latency data plane, transparency to modulation formats, and the capability of compensating transmission impairments in the digital domain. The experimentally demonstrated 8×8 DSXC makes use of a Virtex-7 FPGA platform, which supports any-to-any switching of eight subcarrier channels with mixed modulation formats and data rates. Digital resampling filters, which enable frequency selections and translations of multiple subcarrier channels, have much lower DSP complexity and reduced FPGA resources requirements (DSP slices used in FPGA) in comparison to the traditional technique based on I/Q mixing and filtering. We have also investigated the feasibility of using distributed arithmetic (DA) for real-time DSXC to completely eliminate the usage of DSP slices in FPGA implementation. For the first time, we experimentally demonstrated the implementation of real-time frequency translation and channel selection based on the DA architecture in the same FPGA platform. Compared with resampling filters that leverage multipliers, the DA-based approach eliminates the need of DSP slices in the FPGA implementation and significantly reduces the hardware cost. In addition, with a processing latency that equals to a few clock cycles, a DA-based resampling filter is significantly faster when compared to a conventional direct-structured FIR filter whose overall latency is proportional to the filter order. The DA-based DSXC is, therefore, able to achieve not only the improved spectral efficiency, programmability of multiple orthogonal subcarrier channels, and low hardware resources requirements, but also much reduced cross-connect switching latency when implemented in a real-time DSP hardware platform. This reduced latency can be critically important for time-sensitive applications such as 5G mobile fronthaul, cloud radio access network (C-RAN), cloud-based robot control, tele-surgery and network gaming. | |
