The Research And Application Of Don't Care Bit In Test Patterns Applied By Circuit-under-Test

In recent years, the technology of Intergrated Circuit changes with each passing day. IC is required to have strong function as well as high reliability. The density of the IC increases fast while the number of pins outside of the IC increases slowly because of the limit of size. So it is much more difficult to control and observe the IC exteriorly. Design for Testability was proposed to solve the problem. Built-In Self-Test is a widely used Design for Testability method.Recently, a Built-In Self-Test method called Test Patterns Applied by Circuit-under-Test(TPAC) was proposed. A deterministic test set named Mintest is used in this method. When a test pattern is applied to the Circuit-under-Test(CUT), the response of the node inside of the CUT is captured by registers and feeded back to the inputs of the CUT as the next test pattern.The don't care bit in the Mintest test set is researched in this thesis. Then a method using don't care bit is proposed to improve the fault coverage of TPAC. To decelerate the rate of feedback nodes number decreasing, the don't care bit in the next pattern can be evaluated a special value according to the response of the inside nodes.Then each pattern sequence can generate more patterns belonged to the Mintest set.Base on Generic-Feedback method of TPAC, this thesis proposes another method that switching don't care bit dynamically. Generic-Feedback method in most CUT needs multi-team to reach anticipant fault coverage. Some additional patterns are generated following the deterministic patterns to detect some new faults, the number of faults needed to be detected by the remainder deterministic test patterns decreases. So that more don't care bits can be swithed in the remainder deterministic patterns while the fault set decreasing. More don't care bits in test patterns helps the next feedback team to reach higher fault coverage.The results base on benchmark 85 circuits and Mintest with don't care bits show that compared with the original TPAC method, the feedback teams are cut down in Group-Entire-Feedback, and the fault coverage in Generic feedback can be increased without increasing overhead.