Crosstalk Minimization in SOC using Genetic Algorithm

Shilpa D.R, Uma B.V

Abstract: The era of high speed and high density designs has led the design engineers various challenges. As the technology is scaling, one of the major concerns of VLSI design is signal integrity. Crosstalk is the major cause of signal integrity, occurs due to the coupling between adjacent interconnects. Over the years many layout level crosstalk optimization techniques such as shielding, buffer insertion, and encoding-decoding schemes have been proposed. But addressing this problem at high level of design abstraction is also advisable. This work proposes high level synthesis procedures to eliminate capacitive crosstalk based on genetic algorithm. Crosstalk aware simultaneous scheduling, binding and resource allocation is performed to reduce coupling parameter. Coupling parameter is estimated by having the knowledge of data transition patterns and their correlations. The proposed genetic algorithm with ASAP and ALAP schedules as the initial solutions has reduced crosstalk by 12.2241% for 3x3 matrix multiplication.

Keywords: Genetic Algorithm, Miller Capacitance Factor, High level Synthesis, Crosstalk.