針對造成嚴重電源供應雜訊之元件診斷與修改

Abstract

本論文提出一種自動化的方法來診斷造成電源供應雜訊違規(Power-Supply-Noise violations or PSN violations)的風險元件(risky cells)。我們還可以找到工程變更命令(Engineering Change Order or ECO)，如:調整大小(resize)或移動(move)風險元件以修復電源供應雜訊違規。我們應用重疊原理(superposition principle)將診斷和修復建模為線性規劃問題(linear programming)。我們不僅考慮電源供應雜訊約束，還考慮時序約束，以確保我們可以修復所有電源供應雜訊違規，而不會導致新的時序違規。有了我們的診斷結果，IC設計師可以花費很少的代價執行工程變更命令。實驗結果顯示，我們可以用只有不到2％的候選元件(candidate cell)修復所有電源供應雜訊違規。

This thesis proposes an automatic method to diagnose risky cells that are responsible for PSN (Power Supply Noise) violations. We can also find ECO (Engineering Change Order) method (resize or move) to perform on risky cells to repair PSN violations. We apply superposition principle to model the diagnosis and repair as a linear programming problem. We consider not only PSN constraints but also timing constraints to make sure that we can repair all PSN violations without causing new timing violations. With our diagnosis result, designers can perform ECO with very little circuit modification. The experimental results show that we can repair all PSN violations with only less than 2% of all candidate cells.