Reordering Tests for Efficient Fail Data Collection and Tester Time Reduction
Abstract:
During fail data collection, a tester collects information that is useful for defect diagnosis. If fail data collection can be terminated early, the tester time as well as the volume of fail data will be reduced. Test reordering can enhance the ability to terminate the process early without affecting the quality of diagnosis. In this paper, test reordering targets logic defects based on information that is derived during defect diagnosis. The defect diagnosis procedure is enhanced to identify tests that are useful for defect diagnosis across a sample of faulty instances of a circuit. Tests that are determined to be useful for more faulty instances of a circuit are placed earlier in the test set based on the expectation that the same tests will be useful for other faulty instances of the circuit. The experimental results for logic defects in benchmark circuits support the effectiveness of this approach and indicate that test reordering helps to terminate fail data collection early without impacting the diagnosis quality. The proposed architecture of this paper analysis the logic size, area and power consumption using Xilinx 14.2.
Existing System:
An efficient fail data collection algorithm should ensure that fail data logging stops as early as possible without adversely impacting the diagnosis quality of the reduced fail data logs. The work assumed a given test set and did not modify it in any way. The goal of this paper is to improve the efficiency of the fail data collection algorithm by reordering the test set that it uses. The reordering is done offline before the fail data collection commences. Test reordering has been used for different purposes in the past. The procedures described aim at steepening the fault coverage curve by reordering tests such that the defect coverage would increase by a maximal amount with the application of each additional test. Test reordering has also been explored for reducing power dissipation.
Test reordering has been shown to reduce the size of fault dictionaries. Small fault dictionaries are obtained by reordering the test sets for combinational and scan circuits. This approach manipulates a tree-based fault dictionary, which identifies each fault by traversing the diagnosis tree from root to leaf. Test reordering is used to minimize the lengths of the root to leaf paths in order to reduce the overall information required for distinguishing faults. Current diagnosis procedures use fault simulation to derive the information stored in the fault dictionaries. In this paper, we use a simulation based diagnosis procedure as such a procedure provides more flexibility in selecting the fault candidates and removes the need to store fault dictionaries.
In the approach from, data mining is used to extract a set of rules to establish a relationship between tests and faulty components using a previously available system model. Rules are then ordered based on their confidence to determine the test execution order. With this approach, the average number of tests executed for benchmark circuits is 58.03% of the total number of tests with no loss in diagnosis accuracy. However, this approach depends on the availability of the system model. The work from is based on the premise that the tester data logs used for diagnosis contain only a limited amount of information and that this reduces the quality of diagnosis.
Disadvantages:
- Accuracy is low
Proposed System:
Our reordering procedure is based on information that the enhanced defect diagnosis procedure collects for a sample of faulty instances of the circuit in the presence of logic defects. The number of faulty circuits in the sample is denoted by p. In the following sections, we describe the enhanced defect diagnosis procedure and the reordering procedure based on it.
GTsub: Get a Subset of Important Tests Tsub:
The procedure described in this section is applied separately to every faulty circuit in the sample. The procedure identifies a subset of tests called Tsub, which are the most important for correctly diagnosing the defects in the faulty circuit. Similar to the procedure described, it removes tests from the complete test set T, where the removal of tests from T means that the diagnosis procedure ignores the responses of these tests while selecting candidate faults. Tests are removed one by one. Under certain conditions that are described below, we attempt to remove subsets of tests in order to speed up the procedure.
To determine whether a test (or subset of tests) is relevant for diagnosis, let the candidate faults generated by the complete test set T be denoted by CAND orig. Initially, we assign Tsub =T. By removing a test (or subset of tests) from Tsub, the new candidate set generated by the diagnosis procedure is referred to as CAND new. Procedure GTsub compares the number of candidates in CAND new with that in CAND orig. If the number of candidates is the same, the procedure concludes that the test is not needed for diagnosis and the removal of the test is accepted. In this case, the removal of additional tests is considered with the test excluded from Tsub. If the number of candidates in CANDnew is different from that in CAND orig, the procedure concludes that the test is relevant to diagnosis and it is added back to the test set. The end product of procedure GT sub is a subset of tests Tsub ⊆T such that performing defect diagnosis usingTsub instead of T yields the same number of fault candidates.
Figure 1: Procedure GTreord.
GTreord: Get a Reordered Test Set Based on Tsub:
Based on the results from GTsub for p faulty instances of the circuit, the second procedure, referred to as GTreord, determines the order for the tests in T. Suppose that the application of GTsub to each of the p faulty instances of the circuit in the sample set yields TSUB={Tsub,0,Tsub,1,…Tsub,p−1}.GTreord is based on the premise that a test t j ∈Tis more important for defect diagnosis if it appears in more of the subsets Tsub, k ∈TSUB. With this approach, GTreord assigns to each test tj ∈Ta score that is equal to the number of subsets Tsub, k it appears in. The test with the highest score is determined to be the most important for defect diagnosis, while tests with lower scores are deemed relatively less important. All the tests in Tare then sorted in descending order of their scores to get the final reordered test set Treord. Ties between tests with the same scores are resolved arbitrarily. Fig. 1 illustrates the steps involved in GTreord.
Advantages:
- Accuracy is high
Software implementation:
- Modelsim
- Xilinx ISE