A New Framework for the Robust Design of Analog Blocks using Conic Uncertainty Budgeting

In nanoscale technologies process variability makes it extremely difficult to predict the behavior of manufactured integrated circuits (IC). The problem is especially exacerbated in analog IC where long design cycles, multiple manufacturing iterations, and low performance yields causes only few design to have the volume required to be economically viable. This paper presents a new framework that accounts for process variability by mapping the analog design problem into a robust optimization problem using a conic uncertainty model that dynamically adjust the level of conservativeness of the solutions through the introduction of the notion of budget of uncertainty. Given a yield requirement, the framework implements uncertainty budgeting by linking the yield with the size of the uncertainty set associated to the process variations depending on the design point of interest. Dynamically adjusting the size of the uncertainty set the framework is able to find a larger number of feasible solutions compared to other robust optimization frameworks based on the well known ellipsoidal uncertainty (EU) model. To validate the framework, we applied it to the design of a 90nm CMOS differential pair amplifier and compared the results with those obtained using the EU approach. Experimental results indicate that the proposed Conic Uncertainty with Dynamic Budgeting (CUDB) approach attain up to 18% more designs meeting target yield.