High-Throughput Power-Efficient DSP for Fiber-Optic Communication Systems

Licentiate thesis, 2018

Communication networks are a vital backbone of the modern society. Power dissipation of optical communication links and digital signal processing (DSP) subsystems for these links are a major concern as throughput requirements increase, and the number of deployed systems grows. We are approaching fundamental integrated-circuit scaling limits quickly, and it is thus no longer possible to assume that feature-size scaling will enable implementation of ever more complex algorithms. Therefore, approaching DSP from an implementation perspective, and designing low-power high-performance algorithms will become increasingly more important. This thesis considers power- and energy-efficiency improvements in real-time implementation of DSP algorithms. High-throughput parallel implementations of algorithms are presented and improvements in currently-employed major power-dissipating DSP subsystems (chromatic dispersion compensation and dynamic equalization) are considered. Implementation-aware design of advanced algorithms for long-haul transmission systems is considered. This thesis also considers possible power-reduction in short-haul systems through introduction of forward error correction.