Worst case tolerance design by genetic algorithms

An evolutionary approach to worst-case circuit tolerance design is presented in this paper. Thanks to an appropriate choice of the fitness function and of the reproduction operators, the algorithm is able to explore a wide sub-region of the space of parameters without any prior knowledge about position and size of the region of acceptability. Vertex analysis has been used to check the feasibility of any candidate tolerance region. For this reason, if, as for a large part of cases, the region of acceptability is convex and simply connected, the algorithm ensures an optimal design with 100% yield. The search for a "central" design in the feasible region,. namely for a set of parameters' central values and tolerances that make the circuit the most tolerant to uncertainties, is driven by the optimization of the objective function of interest for the designer as, for example, the per-unit circuit cost. Features and limitations of the approach are discussed by means of two examples.