Design Centering by Yield Prediction

Design centering is an appropriate design tool for all types of electrical circuits to determine the nominal component values by considering the component tolerances. A new approach to design centering will be presented, which starts at the initial nominal values of the circuit parameters and improves these nominal values by maximizing the circuit yield step by step with the aid of a yield prediction formula. Using a variance prediction formula additionally, the yield maximization process can be established with a few iteration steps only, whereby a compromise between the yield improvement and the decrease in statistical certainty must be made in each step. A high-quality interactive optimization method is described which allows a quantitative problem diagnosis. The yield prediction formula is an analytical approximation based on the importance sampling relation. This relation can also be used to reduce the sample size of the necessary Monte Carlo analyses. Finally the efficiency of the presented algorithm will be demonstrated on a switched-capacitor filter.