Exploiting Implicit Information From Data for Linear Macromodeling

In macromodeling, data points of sampled structure responses are always matched to construct linear macromodels for transient simulations of packaging structures. However, implicit information from sampled data has not been exploited comprehensively to facilitate the identification process. In this paper, we exploit implicit information from the sampled data for linear marcomodeling. First, in order to include complementary data for a more informative identification, we propose a discrete-time domain identification framework for frequency-/time-/hybrid-domain macromodeling. Second, we introduce pre-/post-processing techniques (e.g., P-norm identification criterion and warped frequency-/hybrid-domain identification) to interpret implicit information for configurations of identifications. Various examples from chip-level to board-level are used to demonstrate the performance of the proposed framework.