Multi-Objective Design of Analog Integrated Circuits Using Simulated Annealing with Crossover Operator and Weight Adjusting
Keywords:Analog design, CAD, Meta-heuristic, Simulated annealing, Optimization
This paper approaches the problem of analog circuit synthesis through the use of a Simulated Annealing algorithm with capability of performing crossovers with past anchor solutions (solutions better than all the others in one of the specifications) and modifying the weight of the Aggregate Objective Function specifications in order to escape local minimums. Search for the global optimum is followed by search for the Pareto front, which represents the trade-offs involved in the design and it is performed using the proposed algorithm together with Particle Swarm Optimization. In order to check the performance of the algorithm, the synthesis of a Miller Amplifier was accomplished in two different situations. The first was the comparison of 40 syntheses for Adaptive Simulated Annealing (ASA), Simulate Annealing/Quenching (SA/SQ) and the proposed SA/SQ algorithm with crossovers using a 20-minute bounded optimization with the aim of comparing the solutions of each method. Results were compared using Wilcoxon-Mann-Whitney test with a significance of 0.05 and showed that simulated annealing with crossovers have higher change of returning a good solution than the other algorithms used in this test. The second situation was the synthesis not bounded by time aiming to achieve the best circuit in order to test the use of crossovers in SA/SQ. The final amplifier using the proposed algorithm had 15.6 MHz of UGF, 82.6 dBV, 61º phase margin, 26 MV/s slew rate, area of 980 μm² and current supply of 297 μA in a 0.35 μm technology and was performed in 84 minutes.