Design Automation of CMOS Amplifiers Using Precomputed Lookup Tables: A Comparative Study of optimization Algorithms

In this paper, we present a comparative study of optimization algorithms applied to the design automation of a fully differential CMOS amplifier. The design procedure is based on using the $\mathrm{g}_{\mathrm{m}}/\mathrm{I}_{\mathrm{D}}$ methodology coupled with precomputed lookup tables (LUTs) to synthesize the circuit systematically. Symbolic expressions for the performance metrics are generated using a symbolic solver. The optimization problem is formulated and solved using gradient-based and non-gradient-based optimization algorithms. The effect of using a vectorized evaluation of the performance metrics is investigated. The optimization results for five algorithms are compared in terms of the elapsed run time and objective function’s value. The reported results show that a 5,000× shorter run time is achieved compared to SPICE-in-the-loop optimization in addition to providing a better objective function’s value.