Multiparameter Optimization of RSFQ Circuits Using the Method of InscribedHyperspheres

摘要

RSFQ (rapid single flux quantum) circuits must be designed with large operatingmargins because of variations in the current art of superconducting circuit fabrication. Finding the optimum set of parameter values is a slow and tedious process using existing techniques. We describe a multiparameter optimization routine based on the method of inscribed hyperspheres, in which the operating region in multidimensional parameter space is approximated by a convex hull mapped out with a minimum of circuit simulations. Parameter values are centered in the region, taking into account relative variations in fabrication. This maximizes yield rather than 'critical margin,' although the concept of critical margin is preserved as the minimum length of the operating region along a parameter axis. The method has proven effective for as many as eight parameters. Larger RSFQ circuits have been dealt with effectively by successive optimizations taking the parameters several at a time, provided the critical parameter is included in each optimization.