基于几何规划的模拟器件建模与电路性能优化

基于分段线性函数,提出一种新的对模拟器件进行建模的方法,并提出乐观、悲观两种数据拟合算法,两种算法均能较好地拟合MOS管参数。模型建立之后,利用几何规划对电路性能进行优化。乐观算法会得到较好的电路性能,但可能不满足约束条件;悲观算法能够严格满足约束条件,但是得到的电路性能未必最优。在此基础上,提出一种基于响应曲面方法的折衷优化算法,能够得到综合考虑电路性能和约束条件的电路设计参数。最后,以两级放大器为优化实例,证明该方法的有效性。

Analog device modeling and circuit performance optimization based on geometric programming

A new method of analog device modeling based on piecewise linear function is presented, including two types of data fitting algorithms, the optimistic algorithm and pessimistic algorithm. MOS transistor parameters could be properly fitted by both algorithms. With these models, circuit performance could be optimized via geometric programming. The optimistic fitting could yield better performance; nevertheless, design constraint violation may occur. On the other hand, the pessimistic could strictly satisfy the design constraints, yet the yielded performance may not be optimal. In addition, a response surface modeling based algorithm is presented to take to tradeoff between the design constraints and circuit performance. Finally, the capability of the method is illustrated through an example of a two- stage amplifier.