The Construction Method Of Kriging Meta-model And Its Application In Circuit Optimization

In the recent years,the integrated circuit following Moore'low with higher integration degree,more excellent performance,lower power consumption,greatly facilitated the lives of people.But as the manufacturing process into the nanometer era,the structure of large scale integrated circuit makes the circuit design more and more difficult and there is some factors which affect the output results.The parameters of the circuit affect each other and the performance indexes are changing nonlinearly, Especially for analog integrated circuit system and mixed signal system.To obtain the required performance index at the output or a certain node of the circuit, it takes a amount of time to optimize the circuit parameters which greatly increases the time cost.Although Electronic Design Automation(EDA) is helpfull in the design and the optimization of large scale integrated circuit,the EDA softwares are independent of each other,and the increasing size of the integrated circuit makes the optimization time of EDA softwares longer and longer which delays the time of product goes on the market.In this paper,we propose an active set conjugate gradient algorithm to construct kriging model,and then single objective and multi-objective for key nodes are optimized basing on the analysis of the sensitivity of the bandgap voltage referencesource.Finally,on the basis of the analysis of equivalent circuit of TSV-substrate structure,the relevant parameters of TSV are optimized.The main research results of this paper can be summarized as follows:1.From the viewpoint of the metamodel,the numerical model based on the key node parameters and performance index of the circuit is established equivalenting to the physical and electrical model,and after the comparison of the advantages and disadvantages of the various experimental design and constructing model,the Latin hypercube sampling and Kriging model are determined to construct model.2.Determined the spatial correlation model parameter ? by using the active set conjugate gradient algorithm.How to determine the spatial correlation model parameter ? is a key research in the modeling process.In this paper,the active set conjugate gradient algorithm are used to determine the spatial correlation model parameter ?,whose optimization efficiency is higher than the Genetic algorithm.3.From the viewpoint of the circuit,the bandgap reference voltage source are optimized according to the design requirements.Based on the analysis of key nodes,Kriging model is established with the active set conjugate gradient algorithm by the results of the Latin hypercube sampling.After verifing the accuracy of Kriging model,the TC and PSRR are optimized.4.From the viewpoint of the device structure,The TSV-substrate structure are optimized according to the design requirements.On the basis of the analysis of the TSV-substrate structure,the equivalenting circuit and the key parameters of the circuit are determined.In view of the problem of noise coupling in TSV-substrate structure,the equivalenting circuit is optimized by Kriging model which uses the active set conjugate gradient algorithm.