Variability-Aware Design of CMOS Miller OTAs Using an Evolutionary Optimization Approach

Abstract

Evolutionary algorithms (EAs) have emerged as powerful tools for optimizing multidimensional and non-linear design spaces inherent in analog integrated circuits. This paper presents a simulation-in-the-loop EA for automated design of Miller operational transconductance amplifiers, in which robustness to process and mismatch variations is treated as a primary design objective. Unlike traditional approaches that rely on limited analytical transistor models for process variability, the proposed approach incorporates Monte Carlo simulations within a worst-case-based fitness function, enabling the synthesis of designs that satisfy all specifications under statistical variations. The algorithm enforces key performance constraints, including gain, bandwidth, and stability, while simultaneously reducing silicon area and slightly lowering current consumption compared to a carefully optimized heuristic design. A design example in a 180 nm CMOS process validates the approach. Extensive simulation results demonstrate that the proposed method provides an effective and flexible framework for robust analog circuit design under variability and complex trade-offs.