Boundary search approach to parameter design for analog circuits

Abstract

This paper presents a novel approach to parameter design for analog circuits. This approach is based on the peculiarity of the design equations which are linear with respect to every one of the design parameters. Once the initial feasible hyper cube of the design parameters is defined, the solution space (reduced cube) within the cube for each of design equations can be estimated by evaluating zeros only at the boundary edges of the initial cube. Furthermore, the solution space of simultaneous equations can be estimated as the intersection of reduced cubes for each equation. The proposed Boundary Search Method finds the simultaneous solution of a set of design equations by iterative reduction of the feasible cube. This method needs only specifications of minimum and maximum values of design parameters, and it may find all feasible solutions. (Conventional nonlinear programming (nonlinear optimization) needs an appropriate initial solution, and it may easily be trapped by a local optimum).