Title: Automated Nonlinear Macromodelling and Applications

Speaker: Jaijeet Roychowdhury, University of Minnesota

Date/Time: Monday, August 8, 2005, 2:00-3:00 pm

Location: Building 980, Room 95 (Sandia NM)

Brief Abstract: Over the last few years, techniques for automatically abstracting small, computationally efficient macromodels of nonlinear systems have started making an impact on electronic circuit design and CAD. In this talk, we will describe three nonlinear macromodelling research areas that have shown particular promise. The first, trajectory-based piecewise polynomial (PWP) macromodelling, has been applied successfully to a variety of stable non-oscillatory nonlinear circuits (including op-amps, comparators, and I/O buffers) to generate useful general-purpose macromodels faster by about an order of magnitude.

The second topic pertains to algorithms for reducing time-varying systems using multitime differential equations combined with Krylov-subspace methods.  Applications include RF circuits such as mixers and sampling circuits such as switched-capacitor filters. Another promising and timely application - addressing a critical missing need for digital switching noise prediction - is abstracting computationally-efficient interference macromodels of switching digital gates.

The third area to be described concerns macromodelling oscillatory systems.  We will provide a brief description of numerical techniques to abstract nonlinear phase macromodels of oscillators. We will describe the application of these macromodels to predict phase noise and injection locking in oscillators, as well as uses for fast phase-locked loop (PLL) simulation. If time permits, we will also describe recent biological and nanoscale applications to predict spontaneous pattern formation and image-processing behaviors in very large systems of coupled oscillators.  Open problems and challenges will be mentioned in the course of the talk.