Surrogate-based Optimization-aided Design for Low Power Analog Circuits

Automatic and flexible analog integrated circuit design is still an open topic. Simulation-based optimization is accurate but relies on the efficacy of heuristics to find a solution. Equation-based optimization is faster, can work with both gradients and heuristic algorithms but the accuracy depends on the set of equations. We propose an equation-based surrogate model that combines physics-based transistor model, curve fitting approximations for device parameters, and circuit analysis describing the metrics of the circuit to optimize. Two design examples are provided as proof of concept of the proposed optimization framework, a 2nd-order active-RC Butterworth filter, and a Capless-LDO. The results prove this technique to be effective finding solutions for different processes technologies and flexible to the specifications. Multi-objective optimization is used to generate a Pareto front representing the design trade-off.