Parameter Optimization of Single Flux Quantum Digital Circuits Based on Monte Carlo Yield Analysis

摘要

We have developed a parameter optimization tool, Monte CarloJosephson simulator (MJSIM), for rapid single flux quan- tum (RSFQ)digital circuits based on a Monte Carlo yield analysis. MJSIM cangenerate a number of net lists for the JSIM, where all pa- rametervalues are varied randomly according to the Gaussian distribu- tionfunction, and calculate the circuit yields automatically. MJSIM canalso produce an improved parameter set using the algorithm of thecenter--of gravity method. In this algorithm, an improved parametervector is derived by calculating the average of parameter vectorsinside and outside the operating region. As a case study, we haveoptimized the circuit parameters of an RS flip-flop, and investigatedthe validity and efficiency of this optimization method byconsidering the conver- gency and initial condition dependence of thefinal results. We also proposed a method for accelerating theoptimization speed by increas- ing 3σ spreads of the parameterdistribution during the optimization.