Research On Optimization Of Diagnostic Test Pattern Set And Fault Pattern Set

Fault testing is a very important part in chip development,and it is distributed in all steps of development.Each chip needs to be tested for defects before release to ensure that defective chips do not enter the market.Normally,faults are detected by mounting the chip on an automatic test equipment and applying different test patterns to the chip.In addition,for chips that fail the test,they need to be diagnosed to determine the cause of the failure in order to improve the manufacturing process and increase the yield.Test pattern set optimization is of great significance in reducing chip test costs.As the size of the chip decreases,the complexity of the circuits on the chip increases,which increases the number of test patterns required to apply on the chip,which not only increases the test time,but more importantly increases the cost of the test.Therefore,reducing the number of test patterns during testing can effectively reduce testing costs and reduce the time required for testing.At the same time,as the complexity of the circuit increases,diagnosing faults in the chip becomes more and more challenging,and it also takes more time.In order to reduce the time for fault detection and fault diagnosis and improve test efficiency,the quality of test patterns must be enhanced to improve the ability of test vectors to distinguish faults.Model-based diagnosis is a new intelligent diagnostic reasoning technology,which is called an abnormal revolution in diagnostic theory and technology by experts in the field of artificial intelligence.The method of solving the minimal hitting set is a key step in the process of model-based diagnosis.A large number of experts and scholars have obtained fruitful results in the research of the minimal hitting set.Solving minimal hitting set is a classic combinatorial optimization problem.It has important applications in practical problems related to scheduling,allocation,and set covering problem.In this paper,through an in-depth study of the minimal hitting set and the test set optimization problem,an optimization method that can effectively reduce the test set for fault detection is proposed.In addition,for the diagnosis test set which is used for fault diagnosis,a more efficient generation algorithm is proposed:1)Fault test set reduction method based on the minimal hitting set solution method: For the automatic test pattern generation tool Tetra MAX ATPG 2018,which is widely used in the industry,it is found that the test set generated is redundant,so there is a possibility of further reduction and necessary.By modeling the relationship between the test patterns and all the faults in the circuit,the problem of test pattern reduction is transformed into a problem of set coverage.This method can not only obtain complete results,but also ensure the minimum size of the final test set.Experimental results show that compared with Tetra MAX ATPG 2018,this method can effectively reduce the size of the test set and guarantee the fault coverage of the test set.2)Global diagnostic test set generation method based on the minimal hitting set solution method: For simple and effective method COMEDI,it relies too much on the sequence of test patterns and does not have enough globality.Therefore,a global diagnostic test set generation method is proposed in this paper,which selects valid test patterns through an overall comparison between the corresponding output responses of the test patterns.This method improves the independence of the test patterns.At the same time,when the scale of the fault to be diagnosed is reduced to a certain degree,the method of solving minimal hitting set is used to select test patterns from candidate set cluster and it can speed up the process of test generation.Experimental results show that the efficiency of this method can reach up to 5 times of the COMEDI method,which effectively improves the efficiency of test generation.1)The method reduces the scale of the fault test set,and the theory and practice show that a good reduction effect is achieved;2)The method optimizes the generation process of the diagnostic test set,and the experimental results show that the efficiency of the generation process has Great improvement.