Modeling and analysis of high-speed sources and serial links for signal integrity

As the computer and electronics industry moves towards higher data rates, signal integrity and electromagnetic interference (EMI) problems always present challenges for designers for high-speed data communication systems. To characterize the entire link path between transmitters and receivers, accurate models for sources, passive link path (such as traces, vias, connectors, etc), and terminations should be built before simulations either in frequency or time domain. Due to the imperfection of model, data corrections are preferred before time-domain simulations to ensure stability. Moreover, data obtained from models should be compared with measurement results to judge the level of agreement for validations. This thesis presents a new approach to model via structures to help design signal link path while maintaining a low insertion loss and minimizing crosstalk, borrowing the concepts from the transmission line theories. For the models of sources, a dipole model is proposed to represent integrated circuit (IC) radiation emissions while a circuit model for I/O current source is proposed for IC conductive emissions. Passivity and causality are two important properties for passive networks. This thesis also presents detailed algorithm to check passivity and causality for networks with arbitrary port numbers. Data corrections in term of passivity and causality enforcement are applied based on matrix perturbation theory. Last but not least, Feature Selective Validation (FSV) technique is expanded in this thesis to quantify the comparisons of data sets and provide quantitative standard for data optimization.