巨集模型與稀疏復原之電源供應網路暫態增量分析技術

Abstract

隨著VLSI(Very Large Scale Integrated Circuit)科技的發展，由於晶片上巨大的電源節點數目，電源供應網路分析是一具有挑戰性的任務。傳統上，在設計過程中，電源供應網路需要被頻繁的驗證且錯誤通常都發生在晶片工作時。因此，設計者強烈的需要一個可以有效捕捉電路修改後暫態行為的方法。我們整合巨集模型技巧、稀疏復原機制(正交匹配追蹤法)、暫態的連續性和一提出的自動偵錯系統來發展一個有效且可信賴的RLC電源供應網路暫態增量分析器。此分析器不但可以處理既有電路元件值的變動，而且可以處理網路拓樸(Topology)的修改。此發展的增量分析器不但可以比重做聚集模型暫態分析法速度快上大於一個量級且保持良好的精準度。

As the advancement of VLSI technology, power grid analysis becomes a challenging task because of numerous amounts of power grid nodes on a chip. Typically, power network needs to be frequently verified during the design, and violations often occur while the chip is acting. Therefore, an effective method that is able to capture the transient behavior of modified power network is highly needed for designers. This work integrates macro modeling techniques, sparse recovery mechanisms (orthogonal matching pursuit), continuous characteristic of transient analysis, and a proposed adaptive error control system to develop an efficient and reliable RLC power grid transient incremental analyzer. The developed analyzer not only can deal with the change of existing element values but also can handle the modification of topology. Compared with redoing transient analysis by using the macro modeling technique, the developed incremental analyzer can achieve orders of magnitude speedup with negligible loss of accuracy.