Research On Pseudo-Random Test Method Based On Spectral Transform

Integrated circuit testing is an effective way to ensure the high quality of integrated circuits,and it is also an indispensable part of the process of integrated circuit development and design.With the rapid development of integrated circuits,integrated circuit chip integration and complexity is getting higher and higher,chip test test data is growing rapidly,testing becomes more and more difficult,the test cost is also getting higher and higher.The growth of the test data makes the automatic tester capacity and bandwidth requirements in the testing process is increasing,but also increase the test application time and test power consumption.Therefore,how to reduce the cost of testing has become an important research hotspot in integrated circuit testing.This paper mainly reduces the test cost from the point of view of reducing test application time.The main work of the paper is as follows.(1)A pseudo-random test generation method based on packet spectral transform is proposed.In this method,the test vectors in the original test are divided into several groups according to the rules of the test vectors with smaller Hamming distance.Each group of small test sets is subjected to spectral transform to extract a principal component.Finally,LFSR is used to generate weighted pseudo-Similar to the principal component of the pseudo-random test vector set.Experiments on the ISCAS’85 reference circuit and the ISCAS’89 reference circuit show that the efficiency of the pseudo-random test vector generated by this method is higher than that of the quasi-full maximum distance pseudo-random test method and non-grouped spectral transformations.Under the premise of achieving the same fault coverage,the test application time of this method is 32.65%less than that of the quasi-maximum distance pseudo-random test method,which is 20.29%lower than that of the non-grouped spectral transformation method.(2)A passive pseudo-random test method for partial spectral transform is proposed.In this method,the vector of the original test set is cut into small pieces,and the main component values are obtained by performing spectral transformations on the small pieces.For the small segments directly compatible with the original test set,the incompatible segments are used to store the pseudo-random sequences generated by LFSR The main component of the adjacent segment is the same as the test vector.And then block the test set,divided into valid and invalid block,add control information to remove the invalid block.Experiments on the ISCAS’85 reference circuit and the ISCAS’89 reference circuit show that the test time of this method is 54.23%less than that of the controlled LFSR pseudo-random test method,which is 38.8%less than that of the quasi-maximum distance pseudo-random test method.