In this dissertation, a vital step of VLSI physical design flow, synthesis of clock distribution networks, is investigated. Clock network synthesis (CNS) involves large and complex optimization problems to achieve high performance and low power demands of current integrated circuits (ICs). Ineffectiveness of existing methodologies to provide high performance at lower voltage nodes is the main driver for this dissertation research. A design and automation flow for voltage-scaled clock networks is proposed to satisfy tight timing constraints at high frequency (for high performance) and low voltage (for low power) operation.
One implementation of voltage-scaled clock networks is low (voltage) swing clocking, which is a known technique, yet its applicability remains limited to designs with low performance demands. In this dissertation, novel methodologies are introduced to i) apply low swing clocking to legacy designs as a power saving methodology, ii) develop a complete CNS flow for low swing clocking of high performance ICs. These methodologies include slew-driven approaches that are better suited to future transistor and interconnect technologies. Second implementation of voltage-scaled clock networks is multi-voltage clocking, which is another known technique, yet its applicability remains limited to clock tree topology. In this dissertation, multi-voltage clocking with a clock mesh topology is investigated in order to address a missing aspect in the current IC design flows.
Practical considerations of the current IC design flows are also investigated in this dissertation to expand the applicability of the proposed CNS flow. A novel methodology is introduced to facilitate clock gating within low swing clocking. The applicability of low swing clocking to FinFET technology, which is currently the industry norm, is shown to be effective.
