A hybrid approach for analog design optimisation
Lioua Labrak, Thierry Tixier, Y. Fellah, Nacer Abouchi · Conference proceedings · 2007
This paper presents a new method to improve analog design automation and thus design reuse. A top-down constraint driven methodology is applied to design complex analog system. This approach formulates the design problem as a multi objective optimization problem (MOOP). In the last decades optimization has been introduced in the field of analog design. Nevertheless, the knowledge needed to build an optimization problem dedicated to an analog circuit is still out of reach of designers. Our approach is a new efficient way to capture design performance metrics allowing an automatic optimization problem formulation. Based only on performance specifications (e.g., gain of 80 dB) and preferences of the designers (i.e, maximize or minimize a given performance metric), a multi objective optimisation problem (MOOP) is built. An example using hybrid algorithm consisting in coupling a pattern search based algorithm and a conjugate gradient based one to solve the MOOP for a CMOS two stage opamp is given.