Surrogating circuit design solutions with robustness metrics

• A new robustness model that estimates robustness of both design variability and performance uncertainty.
• A two-way mapping design framework that integrates design space robustness and performance space robustness.
• A surrogate-based optimization procedure for the proposed design methodology.
• Promising robustness improvements compared with the conventional robust design approach.

With the increase in device variability, the performance uncertainty poses a daunting challenge to analog/mixed-signal circuit design. This situation requires a robust design approach to add large margins to the circuit and system-level specification to ensure correct operation and the overall yield. In this paper, we propose a new robust design approach by using norm metrics to quantify the robustness for both design parameters and performance uncertainty. In addition, we adopt a surrogating procedure to achieve robustness in design space and to reduce uncertainty in performance space. The end result of the proposed method is a Pareto-surface that provides the designer with trade-offs between design robustness and performance uncertainty. One advantage of this new approach is the ability to take into account the strong nonlinear relationship between performance and design parameters. Considering a set of highly nonlinear circuit performances, we demonstrate the effectiveness of this robust design framework on a fully CMOS operational amplifier circuit.