Methodologies for modeling simultaneous switching noise in multi-layered packages and boards

This dissertation described a methodology for modeling the simultaneous switching noise (SSN) in multi-layered packages and boards. The circuit models were first derived for the multi-port power/ground planes. These models were then expanded for multi-layered structures arising in packages and boards. Transmission lines were then incorporated into the models based on superposition theory to account for the return current effects.; A test vehicle was measured to quantify the plane-to-plane noise waveforms. It was shown that the SSN is caused by plane bounce, which is caused by the return currents of the signal transmission lines. The modeling method was correlated with measurements, showing the validity of the modeling methodology.; The modeling methodology was then extended to account for via transitions and splits on planes by accounting for the return currents on the planes. It was demonstrated that using the methodology discussed in this dissertation, a system can be modeled to simulate plane bounce and its effect on driver and receiver switching and the SSN at all levels of assembly, i.e. chip/package/PCB structure, can be computed.; To further validate the modeling methodology, core and I/O switching noise in a functioning computer system was measured. In the core power distribution system (PDS), the noise at twice the clock frequency was found to be the dominant noise. In the I/O PDS, the switching noise was caused by the return currents flowing on the I/O power planes. The modeling methodology was applied to the functioning system and showed good model to hardware correlation, demonstrating the application of the methodology for modeling complex and realistic systems.; Placement and optimization of discrete decoupling capacitors on the printed circuit board power and ground planes were described.; Both discrete and embedded decoupling capacitors were used in the simulations to understand their effects on reducing SSN. Based on simulations, it was demonstrated that switching noise can be reduced by using thin and lossy dielectric material, which provides sufficient attenuation for suppressing the plane resonances.