Research On Optimization Of Fault Test And Diagnosis Of Integrated Circuit

Test and diagnosis problems of the integrated circuit are very important fields in the semiconductor manufacturing process.Every chip made with new technology needs a fault test to ensure that the defective chip will not enter the market.With the reduction of characteristic size,the complexity of the circuit on the chip is getting higher and higher,which leads to the increase of the number of test patterns for fault test.As a result,the time required for fault test and the cost of test increase.Therefore,it is necessary to compact the test set to save the cost of the chip test.At the same time,the defective chip needs to be diagnosed to determine the cause of the fault during the fault test,so as to improve the manufacturing process of the chip and increase the yield of the chip.The high accuracy fault diagnosis program can help the physical fault analysis process to quickly and accurately find out the fault location of the chip,which can help reduce the cost of mass production.The fault test of the chip can generate a set of high-quality test patterns for a specific fault type by using the automatic test pattern generation technology.Cadence,Mentor Graphic and Synopsis chip design companies are the representative companies providing commercial automatic test pattern generation tools,among which Tetra MAX ATPG is the most powerful and convenient automatic test pattern generation tool developed by Synopsis chip design company.However,Tetra MAX ATPG can only generate the test set with the highest fault coverage for the specific fault type,and the test set contains a large number of redundant test patterns.Traditional static test set compaction methods only aim at fault coverage,and can only get a reduced test set of a specific fault type.But one goal is difficult to apply to all situations.In real life,multiple goals need to be met,including test cost and fault coverages of multiple fault types.Based on the abovementioned problem background,this paper proposes a static test set compaction method based on the multi-objective particle swarm optimization algorithm to solve the limitation of the single objective test set compaction.Firstly,the problem of test set compaction is remodeled and transformed into a particle swarm optimization problem.Then,the length of the test set and multiple fault coverages of various fault types are set as objectives,and the multi-objective particle swarm optimization algorithm is used to solve the multiobjective test set compaction problem.This method can reduce the test set at the same time under multiple fault coverages of multiple fault types.The experiment results on benchmark circuits show the effectiveness of the method.The fault diagnosis can find out the possible fault locations for the defective chip.In the initial stage of new technology production and improvement,the defective chip may have multiple faults at the same time.Due to the interaction between these faults,the fault diagnosis method based on the simulated single fault model may produce a large number of candidate faults.The early fault diagnosis method based on test scores will eliminate the real faults contained in the set when reducing the candidate fault set,which leads to the decline of diagnostic accuracy.In real life,high accuracy fault diagnosis technology is needed to reduce the cost.Therefore,this paper proposes a new method of combing test scores to make better use of diagnostic information in order to reduce the number of candidate faults and improve the accuracy of diagnosis.This method uses the output response of fault free circuit and the output responses of single modeled faults to calculate the score of each test pattern,so that the same test set can distinguish more faults and improve the diagnostic ability of the test set,so that the number of redundant faults in the candidate fault set is reduced and the number of actual faults is increased.Meanwhile,the method can complete the test scores statistics offline by using the fault free information,which reduces the running time of fault diagnosis.Compare with the early method,the experimental results show that this method can effectively improve the accuracy of fault diagnosis and improve the efficiency of diagnosis.