Sparse Macromodeling for Nonlinear Networks

Model order reduction (MOR) has proven to be an effective tool in addressing the simulation complexities of modern microsystem such as the ones arising due to large interconnects networks. Traditional MOR methods are frequency domain methods and are therefore limited to linear networks. Recently, time domain MOR was developed extending this concept to nonlinear macromodels. However, the resulting reduced nonlinear macromodel is dense which reduces the efficiency of the simulation. In this paper a new formulation is proposed which allows for the sparsification of the reduced nonlinear macromodel, thus resulting in significant CPU cost saving as is demonstrated in the examples