典型的類比電路自動化設計方法大致可分為兩大類：模擬基礎 (simulation-based)以及方程式基礎(equation-based)方法。模擬基礎的方法有高度的準確性，但需要消耗大量的模擬時間。方程式基礎方法雖然執行速度快，但因為估計公式無法考慮到較高階的電路效應使得結果不準確。因此，本論文提出一套結合方程式基礎與模擬基礎優點的類比電路自動化設計方法。利用本論文提出的「全域性方程式設計參數空間化簡方法」，能快速地縮減設計參數空間，留下小部分最佳解可能存在的區域。之後，再用模擬基礎的最佳化方法，對剩下的小部分區域作搜尋。在實驗中，所提出的演算法用於兩級式運算放大器自動化設計與鎖相迴路行為模型參數設計上。在第一階段的演算法中，原本的設計參數空間被化簡為小範圍的設計參數空間，然後再用模擬基礎的搜尋方法對化簡後的設計參數空間作搜尋。從實驗結果可以看出，提出的演算法能有效地提高模擬基礎最佳化方法的效率，且結果也保有高度的準確性。 Traditional analog design automation approaches can be classified into two categories. One is simulation-based optimization. It has high accuracy but requires more simulation time to find the global optimal solution. The other is equation-based optimization. It has short execution time but the accuracy is often limited due to the simplified equations. In this thesis, a combined method is proposed to take the advantages of both simulation-based and equation-based approaches for analog design automation. By using the proposed “global equation-based design space reduction” method, the design space with the global optimal can be simplified to a small region quickly in the first stage of the proposed algorithm. Then, simulation-based optimization is applied to search the small region to accurately identify the global optimal point. As shown in the experimental results on several analog circuits, the proposed method has higher efficiency than traditional global search approaches without sacrifice on accuracy.
