Algorithmic nonlinear macromodeling: Challenges, solutions and applications in Analog/Mixed-Signal validation

Analog/Mixed-Signal validation at the system level is becoming increasingly important as more electrical bugs are caused by the interaction among various circuit blocks. While hand-crafted behavioral models and linear models are still most widely used among designers, there is an increasing need for automatic behavioral modeling tools which capture low-level nonlinear behaviors in the circuit. This paper discusses challenges and difficulties of algorithmic nonlinear macromodeling, and reviews a series of recently developed techniques. In particular, we study the behavioral modeling problem from the perspective of projection in the state space defined by voltages and currents. We review a few nonlinear macromodeling techniques from the projection perspective, and demonstrate the model accuracy and computational efficiency compared to transistor-level models and linear models.