Dynamic scan testing: a new paradigm
Jr. Clay Samuel Gloster · 1993
The cost of testing can be divided into two major components: test generation and test application. For a given part, test generation is a one-time cost, while the cost of test application is expended repeatedly. Therefore, in order to minimize the total cost of testing, the optimum strategy is one that minimizes the cost of test generation as well as the test sequence length. This research presents a strategy which attempts to reduce both the cost of generating a test as well as the time required to apply this test. The key to dynamic scan testing is the fact that we supplement an existing sequential circuit, called the object machine, with a scannable test machine to improve controllability and observability of internal machine states. Another key is a novel dynamic algorithm for test generation which takes advantage of the knowledge of the test machine hardware to reduce test generation costs and the overall length of the test sequence. This thesis introduces a new cellular scan test machine which is found to perform well within the dynamic testing paradigm. It is a test machine which connects the flip-flops into one or more circular chains with each chain containing an exclusive-or gate at the beginning of the chain. One input of each exclusive-or, is directly controllable from the circuit inputs. The thesis also presents a dynamic scan test generation algorithm, DYNASTEE, which reduces the average test sequence length per fault when compared to existing static test generation algorithms for scan architectures. It allows the mode control pin to cycle freely between object and test machine operation as required. The scan register is shifted for minimal time to reach a state required for testing. Test sequence length is further optimized via a scan chain ordering algorithm. This thesis demonstrates the flexibility of dynamic scan testing by presenting its effectiveness for both partial and multiple scan chains as well as introducing a technique for random test generation. The thesis concludes with results of experiments which validate the theory that dynamic scan testing can significantly reduce the cost of testing and is a viable alternative to traditional static scan testing approaches.