Development Of Optimization Algorithm For RF Passive Device Modeling

Semiconductor device model is an important link between he IC designers and IC foundry,its accuracy directly affects the performance of IC.With the IC integration and device working frequency gradually increased,the device model has become increasingly complex and the number of parameters more and more.This leads to difficulties in extracting the appropriate model parameters.However,the accuracy of the device model is determined by the model's parameter values.In order to extract suitable model parameters,two kinds of methods are summarized: one is direct method which use the physical characteristics of device or algebraic method,and the other is to use the optimization algorithm to search for the appropriate model parameters.Direct extraction method is simple,intuitive,and the solution is unique,but with the model more and more complex,using the direct method to extract parameters become more difficult.Compared with the direct extraction method,it is relatively easy to extract the model parameters using the optimization algorithm.Some algorithms such as gradient descent,Newton iteration and differential method have been applied to the extraction of model parameters,and some results have been obtained.However,those algorithms easily fall into local optimization.Then,the intelligent algorithms,such as Genetic Algorithm(GA)and Particle Swarm Optimization(PSO),are applied to extract the model parameters,which are applications that have achieved great success.However these algorithms have the disadvantage of "premature".So the speed of convergence has been limited,for which the researchers have also improved the algorithm to avoid "premature" phenomenon.Therefore,it is of great significance to study the parameters of the device model extracted by intelligent optimization algorithm.In this paper,the on-chip spiral inductor is taken as an example to study the extraction and optimization of model parameters.Firstly,the concept of optimization and related intelligent optimization algorithms are briefly introduced,and the cuckoo search algorithm(CS)is improved.The experimental results show that the improved algorithm has better performance than GA,PSO and original CS.According to the performance of the optimization algorithm,we decided to improve the CS applied to the semiconductor device model parameters extraction.Then,an on-chip inductor is introduced as an example to describe the process of device modeling,and an automatic optimization program is designed based on the improved algorithm and model structure.At the same time,according to the characteristics of the device model parameters,the sensitivity analysis experiment of the model parameters have been done and combined with the sensitivity of the model parameters proposed cross-model operation parameters to improve the convergence rate.In addition,in order to further improve the speed of automatic optimization,Optimization of the direction of the experiment has been done,and improves the cuckoo search algorithm update formula.Finally,the designed optimizer is implanted into the independent development platform.The performance of the optimizer is verified by the measured data of 53 on-chip spiral inductors.The results show that the optimization results meet the engineering requirements.In order to verify the performance of the improved algorithm,the original CS and the improved CS are used to optimize the same device respectively.The performance of the two algorithms is compared with that of the improved one,and the convergence speed of the improved algorithm is improved.