Energy-Efficient Digital Signal Processing for Fiber-Optic Communication Systems

Abstract: Modern fiber-optic communication systems rely on complex digital signal processing (DSP) and forward error correction (FEC), which contribute to a significant amount of the over-all link power dissipation. Bandwidth demands are evergrowing and circuit technology scaling will due to fundamental reasons come to an end; energy-efficient design of DSP is thus necessary both from a sustainability perspective and a technical perspective. This thesis explores energy-efficient design of the sub-systems that are estimated to contribute to the majority of the receiver application-specific integrated-circuit power dissipation: chromatic-dispersion compensation, dynamic equalization, nonlinearity mitigation, and forward error correction. With a focus on real-time-processing circuit implementation of the considered algorithms, aspects such as finite-precision effects, pipelining, and parallel processing are explored, the impact on compensation and correction performance is investigated, and energy-efficient circuit implementations are developed. The sub-systems are investigated both individually, and in a system context. DSP designs showing significant energy-efficiency improvements are presented, as well as very high-throughput, energy-efficient, FEC designs. The subsystems are also considered in the context of datacenter interconnect links, and it is shown that DSP-based coherent systems are feasible even in power constrained settings.