Yield optimization of a MMIC distributed amplifier using physically-based device models

It is shown that by using a physical model to generate correlated parameters and response modeling to overcome long response times, even complex circuits can be optimized for maximum yield. A monolithic microwave integrated circuit (MMIC) distributed amplifier was simulated and optimized for maximum design yield. It is demonstrated that fitting the modeled circuit response to a simplified model that is a function only of the statistically perturbed circuit parameters enables large speed increases with little observed degradation in accuracy. Modeling devices at the process level allows meaningful data about the statistical parameters of devices to be deembedded. Improvements to the simulators, some of which now accept correlated and user-defined statistical data entry, allow realistic yielded determination during the design phase.<>