Tolerance design of controllers for switching regulators

Abstract

An evolutionary approach to worst case tolerance design is introduced here, with a focus on feedback compensation networks for dc-dc switching converters. Assumed that varying parameters values are uniformly distributed and uncorrelated, as provided by the worst case approach, the proposed algorithm, of general applicability, seeks for the set of nominal values and tolerances of the circuit parameters ensuring that the design constraints are met and that a user-defined circuit performance index assumes its optimal value. Design constraints, are fixed in the frequency domain, in terms of acceptability ranges of loop gain crossover frequency and phase margin, to guarantee closed loop stability and the desired dynamic performance. Resistive and capacitive compensation network's parameters values are chosen within a suitable database of couples nominal value/tolerance available on the market, while the nominal values and tolerances of the parameters of the power stage are fixed. Referring to a buck dc-dc switching regulator, two widely used different compensation network topologies are compared in terms of reliability, robustness, and cost of components. Simulation results show the wide usefulness of the proposed method in supporting designer decisions.