Design and Parameterization of Analog Electronic Circuits Using Stochastic Optimization Algorithms: A Comparative Study with BJT, FET, and OECT Devices

Despite the rise and wide spread of digital technology, analog circuits are experiencing a resurgence for specific applications due to their efficiency in processing physical signals with fewer transistors and lower energy requirements. They offer advantages in explainability and stability, particularly in sensor systems. However, designing these circuits poses challenges, particularly in optimizing them against issues like component aging and environmental dependence. Current research aims to enhance automation in analog circuit design using optimization methods, including stochastic algorithms like Genetic Algorithms (GA), Simulated Annealing (SA), and Particle Swarm Optimization (PSO). These methods improve convergence and fault tolerance, while dealing with the high nonlinearity of analog circuits with transistors. The focus of this paper is on solving multivariate optimization problems with multiple objective functions for analog circuits. We evaluate and compare the component optimization of a small-signal amplifier, operational amplifier, and a sigmoid function circuit using gradient-based and stochastic methods for bipolar junction, field-effect, and organic electro-chemical transistors, differing in transistor transconductance and transfer curves. We show that PSO outperforms all other optimization algorithms in our experiments for all considered circuit classes, showing the highest convergence rates between 40–100% (depending on the circuit class) and the lowest optimization times. The OECT components were manufactured with a drop-on-demand ink-jet process and experimentally characterized, finally deriving an electronic functional digital twin model from measuring data. This is a simulation study but uses data and models from real devices.