双极型晶体管模型参数提取的组合优化算法

Combinatorial Algorithms for BJT Model Parameter Extraction

本文讨论双极型晶体管（ＢＪＴ）器件模型参数提取的最优化方法，提出了一种解决全局最优的组合算法，与通常的Ｇａｕｓｓ－Ｎｅｗｔｏｎ法相比，其突出的优点是：在计算过程中只需计算目标函数值，不必计算目标函数的梯度，所获得的解的全局最优性也较好．本文所提出的方法在全局搜索的基础上还同时解决了初值选择与迭代策略，是一种简便、高效的全局优化算法．

This paper presents some combinatorial algorithms for global optimization problems, which can be applied to solve model parameter extraction in bipolar junction transistors(BJT). The main advantage of combinatorial algorithms to a traditional algorithm like Gauss-Newton method lies in that it needs no computation of the gradient of the objective function. Moreover, the global optimality of this approach is also better than GaussNewton method. As the selection of the initial point and iterative straegies proposed in this paper are based on global search, the combinatorial algorithms are more simple, efficient, and suitable for global optimzing.