Jian Wang Carnegie Mellon University

Parameterized Macromodeling for Analog System-Level Design Exploration

As the complexity of on-chip analog circuits continues to increase, the system level design optimization problem becomes extremely challenging. The main difficulty lies in evaluating the circuit performance, which is very hard to abstract due to the large number of design variables and their complex interactions. Transistor-level simulation only provides the information for a fixed design, and is computationally prohibitive for an iterative optimization process. For this reason most system-level design methodologies employ hand-analysis equations, which unfortunately, can yield significant errors for many applications. Well-designed macromodels can provide reasonable trade-off between evaluation speed and accuracy, and is therefore desired in the system-level simulation.

In this talk, we will present a methodology for extracting macromodels of analog circuits. Particularly, the macromodel is parameterized as a function of design variables, so that it can be used in design exploration. During the modeling process, the analog circuit is first characterized using commercial simulator and comprehensive device models, which avoids human manipulation and renders better accuracy. Then a parameterized order reduction technique, CORE, is employed to compress the obtained model so that the evaluation cost can be reduced. Furthermore, a multi-point scheme is proposed where the global model is built as the weighted sum of the low-order local models. With appropriate partitioning, the multi-point scheme can provides a better coverage over large analog design space. Some preliminary results will be given to demonstrate the effectiveness of the proposed macromodeling method.

Bio:

Jian Wang received the B.S. and M.S. degrees in Microelectronics from Fudan University in 2001 and 2004 respectively. Currently, he is a graduate student in Electrical and Computer Engineering at Carnegie Mellon University, working with Prof. Larry Pileggi. His research interests include various aspects of analog system design and design automation.