Research On Defect Diagnosis Based On Test

With the expansion of the semiconductor industry to the enormous consumer electronics market,the density of components in integrated circuits has grown exponentially,making it difficult to avoid failures in integrated circuits.The fault diagnosis of integrated circuits can detect problems in the design and production process at an early stage,and is important to improve production efficiency.Fault diagnosis based on test set is a commonly used and efficient method in integrated circuit diagnosis.In the fault diagnosis process based on the test set,the diagnosis information is obtained through the fail data generated in the test phase.However,the increase in component density and failure types have significantly increased the amount of fail data required for testing.The large-scale fail data seriously affects diagnosis efficiency and cost.In order to enhance the productivity of integrated circuits,this paper proposes two improvements to the test-based fault diagnosis method: from the perspectives of reducing fail data and improving the performance of the diagnostic method.To reduce the fail data for the diagnostic program,this paper proposes a fail data reduction method combining configuration checking and local search.This method analyzes the logical relationship between fail data and the number of port coverage and proposes the concepts of failure frequency and failure coverage to guide local search.At the same time,it combines configuration checking strategies to avoid repeated searches,thereby reducing redundant iterations to increase exploration capability.This method first obtains an initial solution from the generating process.Then,through the addition and removal method,new solutions are generated,and the failure coverage is updated incrementally to further reduce the initial solution.Finally,the algorithm judges whether the new solution can be a candidate according to the acceptance criterion.The comparison experiment with state-of-the-art fail data reduction method shows that our method can greatly reduce fail data and diagnosis time,and the reduced fail data can also generate more accurate candidate faults during diagnosis.To improve the diagnostic performance of the test-based fault diagnosis method,this paper proposes a fault diagnosis method based on test scores.It is inspired by the fact that when the circuit contains a combination of faults,the test vector with a larger gap from the fault-free output is more likely to contain more diagnostic information.This method puts forward the concept of fault recognition rate.The method first scores the test vector according to the fault recognition rate,meanwhile accomplishes the test set single fault simulation.Then,according to the test score and the output vector which contains a single fault,the algorithm generates the score for faults in the fault list.Experiments show that this method has improved diagnosis accuracy,diagnosis resolution,and time.