Package power delivery analysis for unmerged and merged power rails

• Main Author: Tau, See Yee Hung
• Format: Thesis
• Language: English
• Published: 2014
• Subjects: TK5101-6720 Telecommunication
• Online Access: http://eprints.utm.my/id/eprint/48020/25/SeeTauYeemfke2014.pdf; http://eprints.utm.my/id/eprint/48020/

Power delivery design has become important and critical nowadays especially at the package level, interconnect between silicon and motherboard. It is not an easy task to perform power delivery analysis at package level as it could have more than 30 power rails and it is very time consuming to validate each of them one by one at a time. Resonance frequency of power delivery network is critically important to understand the design risk from power delivery perspective. Inaccuracy of resonance frequency may result in over design and lead to increase of decoupling solution cost. This thesis presents a study of modeling extraction for single and multiple power rails by using Sigrity PowerSI (2.5D field solver) through frequency domain analysis, focuses on the changes in resonance frequency within each modeling case and recommends the appropriate way of extracting multiple power rails. The criterion of modeling extraction methodology in power rails modeling especially for multiple power rails extraction has been proposed for unmerged power rail and merged power rail designs. Time domain analysis was carried out to understand the performance of power delivery network base on the design target and the impact of resonance frequency of power delivery network. Coupling noise effect has shown significant impact in merged package design through time domain analysis. The thesis also reports the comparison of RL network with extracted model for unmerged and merged power rails design. The generated RL model is correlated with extracted model in frequency and time domain analysis for unmerged package design. It is also a good enough model to replace the extracted model of merged package design by justifying through time domain analysis.