Research On Noise Detection And Parameter Estimation Of Semiconductor Devices Based On FARIMA Model

Recently,science and technology have been developed greatly,and electronic equipment has been widely used in many fields.The basic unit of electronic equipment and semiconductor device,its reliability and stability of the system has a direct connection to the semiconductor device before applied to all kinds of electronic equipment to ensure its reliability,avoid the defects of semiconductor device caused by electronic equipment failure and affect the stability of the whole electronic system,may cause irreparable loss.Therefore,the reliability analysis method and parameter estimation method of semiconductor devices are attracting more and more attention from many scholars at home and abroad.In order to master the detection noise of semiconductor and analyze its corresponding parameter estimation more accurately,the need to build a proper model for semiconductor research object,and the appropriate model selection is difficult to grasp,even if the model is more accurate,there will be resources,calculation method of realization,feasibility and other aspects of the problem.At present,the methods of modeling can be divided into three categories: the method based on the signal processing technology,the time series analysis model method,the subspace class method.The method based on signal processing is the main means of time-frequency analysis,such as the Hilbert yellow transform(HHT),which has been developed in recent years.The subspace class method is to process the signal and express it with the state space model,which is mainly concerned with the theory and technology of the decomposition of the matrix.The time series analysis model method uses the parameters of the model to reflect the characteristics of the system.At present,the widely used model is the autoregressive moving average(ARMA)model.The three methods have their own advantages and disadvantages.In this paper,the FARIMA model is used to model.By example,it is proved that the FARIMA model is more accurate than the traditional ARMA model for semiconductor noise detection data.By analyzing the comprehensive problems of resources,computing methods,and feasibility,we divide the current modeling methods into three categories: signal processing technology,time series analysis model and subspace method.The three methods are analyzed and compared one by one,and the FARIMA model is used to model the model at the end.Secondly,the modeling and theoretical research of noise signal are briefly discussed,and the partial autocorrelation function is used to calculate the model.The definition of the partial autocorrelation function(PAC)and the random simulation of the AC and PAC characteristics of the FARIMA model are carried out.Finally,the FARIMA model is used to estimate and study the noise parameters of semiconductors.