A Unified Approach to the Sensitivity and Variability Physics-Based Modeling of Semiconductor Devices Operated in Dynamic Conditions—Part I: Large-Signal Sensitivity

We present here and in the companion paper (Part II) a general framework for the modeling of semiconductor device variability through the physics-based analysis of the small-change sensitivity. We consider a very general class of dynamic device operation, i.e., the periodic or quasi-periodic large signal (LS) time-varying regime, and we evaluate the sensitivity of both dc and harmonic components of the device dynamic working point with respect to process or physical device parameters. The proposed technique is based on the linearization of the physics-based device model around a nominal parameter, and extends to the dynamic case the already established Green's function approach to the numerically efficient dc sensitivity analysis. As an example of application, we consider a class A GaAs MESFET microwave power amplifier; the sensitivity of the LS working point with respect to doping and gate work function variations is evaluated through the proposed approach and compared with the result of repeated LS amplifier analyses, showing that the numerically efficient small-change sensitivity approach provides reliable predictions for parameter variations up to 10% of the nominal value.