Studies On Test Compression Methods Based On FDR Code For Digital Circuits

The huge amount of test data leads to the requirement for high memory performance,huge number of I/O channels and high frequency of automatic test equipment(ATE),and also causes long test application time.Thus,the test cost is high.How to reduce test cost is an important research subject,about which two test compression scheme are proposed in this thesis.Firstly,FDR(frequency directed run-length)codes can compress test stimuli data effectively.However,the number of faults detected by a test pattern is small in average,since don't care bits in the test set are filled by all Os.That is the test quality is low.In order to improve the test quality and further increase test compression ratio,based on FDR codes this paper proposes a scheme where the test response of the previous test pattern is utilized to fill the don't care bits in current test pattern.This scheme increases the randomness of the X-filling.Thus,the test quality is improved.In this scheme,the test compression ratio is also related to the order of test patterns.This paper also proposes a test pattern reordering approach based on the nearest neighbor algorithm to decrease the number of conflict bits between the test response of the previous test pattern and the current test pattern.Difference vectors can be obtained by the difference between test patterns and the test responses of their previous test patterns.The test data volume can be reduced by using FDR codes for the difference vectors.Experimental results show that for larger ISCA'89 circuits the proposed approach increases the test quality by 5.9%and test compression ratio by 2.5%in average compared to those of FDR codes.The hardware overhead,which is only an 2-input XOR gate can be omitted.Secondly,scan tree technique can reduce test data volume and test application time effectively by broadcasting test data into compatible scan cells which are in the same layer of scan tree.Based on scan tree technique and FDR codes this paper proposes a scheme by applying the aforementioned method to scan tree technique.Test patterns of scan tree are made diffrence with test responses of their previous test patterns in this scheme.The test patterns and test responses can be composed of value of any scan cell in every layer of scan tree.For the scan cells in the same layer,their test stimuli values are the same and test response values are different.The test data volume can be further reduced by selecting the scan cells which have the minimum difference distances with test stimuli values.The experiment results show that the proposed approach increases test compression by 3.02%and test quality by 4.9%in average compared to those of the combination of scan tree technology and FDR codes.