Nonlinear Identification of Complex Systems Using Radial Basis Function Networks and Model Order Reduction

As wiring density increases with integration and miniaturization of systems and PCB (for example SiP), the ability to simulate precisely the signal integrity (SI) behaviour of complex systems becomes decisive, while the simulation time increases dramatically. Therefore it is essential that the design of such complex systems is supported by accurate simulation models. This paper proposes a methodology for the identification and the fast simulation of non-linear complex HDI/HDP systems and integrated circuits by means of radial basis function (RBF) nets. For this purpose a specific modelling technique is presented with a complete flow that leads to the development of a black box model (BBM). Furthermore a first attempt to combine this black box modell approach and modell order reduction is discussed. This approach allows accelerated simulations of a whole interconnect system (buffer - transmission line - buffer).