Research On Diagnostic Methodology Of Digital-analog Hybrid Integrated Circuits

Integrated circuits are the foundation and core of electronic and electrical systems,and their applications involve all aspects of human society today.The digital-analog integrated circuit is an important branch of the integrated circuit family,which has the characteristics of strong comprehensiveness,wide coverage and complicated diagnosis.Detection and diagnosis are important means to ensure the correct function and performance of integrated circuits.In this paper,the research on the diagnostic methodology of digital-analog hybrid integrated circuits has important scientific significance and application value.Based on the BP neural network algorithm,this paper proposes a digital-analog integrated circuit diagnosis method.The circuit of the temperature acquisition function module in the integrated circuit is used as the carrier to verify the designed method.The main research contents are as follows:(1)The fitting method is used to improve the selection method of hidden layer neurons in BP neural network.This method covers the total number of neurons in the hidden layer,and the number of neurons with the smallest error rate is obtained by fitting analysis,which improves the prediction accuracy of the BP neural network.(2)In order to overcome the problem that the BP neural network tends to converge to the local minimum,a simulated annealing method is proposed to optimize the network.This method performs a large number of sample exchanges on the training samples,and performs network training after comparing the thresholds.The experimental data verifies that the method is feasible and can significantly improve the network prediction accuracy and iteration speed.(3)Completed the analysis of the temperature sampling circuit formed by the V/F conversion circuit.The faulty test nodes of each module of this circuit are selected and classified,and a large amount of data on the normal and fault states of this circuit are collected.(4)The simulated BP neural network and the optimized BP neural network are used to train and classify the collected circuit data,establish a fault dictionary,and using the test data for experimental verification.The experiments on the specific circuit according to the above scheme show that the improvement of the BP neural network by the simulated annealing method and the fitting method can improve the prediction accuracy of the original BP network and reduce the number of iterations.At the same time,the circuit experimental results confirm that the method can locate the specific fault of the circuit Module and component locations.