An efficient decoupling capacitance optimization using piecewise polynomial models

This paper proposes an efficient decoupling (decaps) capacitance optimization algorithm to reduce the voltage noise of on-chip power grid networks. The new method is based on the efficient charge formulation of the decap allocation problem. But different from the existing work [12], the new method applies the more accurate piecewise polynomial micromodels to estimate the voltage noises during the linear programming process. The resulting method overcomes the over-estimation problem, which plagues the existing method. The proposed method has the best of two worlds: it has the efficiency of the charge-based methods and the accuracy of the sensitivity-based methods. Experimental results demonstrate that the proposed method leads to the decap values similar to that of the sensitivity-based methods, which give the best reported results and are much better than the existing charge-based method, and at the same time, it enjoys the similar efficiency of the charge-based method.