Neural Network for Circuit Models of Monolithic InAlN/GaN NAND and NOR Logic Gates

This paper presents large signal circuit models of InAlN/GaN monolithic integrated NAND and NOR logic gates. The models are calibrated automatically by artificial neural network (NN). Appropriate setting and properties for NN training is described. Very high accuracy of the proposed models is observed. The features and accuracy of NAND and NOR logic gate models prepared by NN are compared with standard empirical compact models based on the transistor level. The advantages and limitations of NN for circuit modeling are discussed. Good agreement between measurements and simulations confirms the validity of the proposed models and methodology for model generation of logic gates.