Abstract:
SPRIM (Structure-Preserving Reduced Interconnect Macromodeling)
is a simple modification of the widely-used order-reduction
method PRIMA for large-scale RCL networks arising in the simulation
of state-of-the-art electronic circuits. With the same computational
work as PRIMA, SPRIM generates reduced-order models that are twice
as accurate as the corresponding PRIMA models and at the same time,
unlike PRIMA, preserve the block structure of the original large RCL circuit. SPRIM was first proposed for the special case of RCL circuits driven by only current sources. While the extension of SPRIM to the general case of RCL circuits driven by current and voltage sources is straightforward, establishing the theoretical properties of SPRIM in this general case is not, and in fact, requires different mathematical techniques. In this talk, we first describe SPRIM for general RCL
circuits and then discuss its theoretical properties. We also present some interesting observations about rank deficiencies in the Krylov subblocks used in the projection approach underlying SPRIM, and demonstrate how these rank deficiencies can be exploited to further reduce the size of the SPRIM models.
This seminar is easily accessible to persons with disabilities. For more information or for assistance, please contact the Mathematics Department at 743-3500.