Fast synthesis of analog circuits based on evolutionary optimization of ANFIS space mapped model

A methodology for fast synthesis of analog circuits based on adaptive network fuzzy inference system (ANFIS) and evolutionary optimization is presented in this work. The method is based on establishment of a surrogate nonlinear model which improves the efficiency of computational tasks and also increases the design accuracy. Furthermore, development of an ANFIS coarse model provides further reduction of computational cost. Owing to generalization property of space mapping approach the design method could be used on various circuit designs. Two design experiments including precise wave shaper and two stage CMOS OTA are used as examples of multi-variable cost function optimization to demonstrate the validity of proposed method. HSPICE implementations based on 0.18 ¿m technology are used. The proposed model is compared to standard neural network space mapping. Considerable reduction of computational cost and increase on design accuracy is resulted for various optimization criteria.