Optimization and Model Order Reduction in Circuit Design

G. Gangemi (STMicroelectronics)

While during the last decades the great enhancements in the field of digital design methodologies and tools have allowed to design larger digital circuits in less time, the analog circuit design methods have not progressed at the same rate. The design of analog electrical circuits needs electronic engineers with a long experience and a wide knowledge of the theories that rule this kind of circuits. However, experimental optimization tools exist; they search the space of solutions for optimal configurations of variables sets, given a circuit netlist provided by the designers. Typical analog integrated circuit optimization problems are computationally hard and require the handling of multiple, conflicting, and non-commensurate objectives having strong nonlinear interdependence. In general it is possible to reformulate integrated circuit design as constrained multi-objective optimization problems defined in a mixed integer/discrete/continuous domain. The hereby employed traditional numerical techniques are becoming too much time-consuming for circuits of industrial complexity. The long computation time required for the optimization of a complete circuit cannot be tolerated especially in the early design stages. For tackling this complexity problem model reduction methods are a promising approach in order to achieve a faster performance evaluation in order to obtain more robust devices within a more efficient design process.

This minisymposium focusses on the usage of model reduction techniques in combination with optimization methods. The results are developed in the EU Marie Curie projects SymTecO (Symbolic Techniques for Circuit Optimization) and O-Moore-Nice! (Operational Model Order Reduction for Nanoscale IC Electronics).

Both projects address Transfer of Knowledge on Mathematics for Industry

Back