Cost-Effective Deterministic Partitioning for Rapid Diagnosis in Scan-Based BIST

of-the-art chip designs to improve test quality and reduce the cost of test development and application. Despite such benefits, designers have not adopted BIST as the primary test methodology. Fault diagnosis in a BIST environment is problematic because only limited information is available in a compact signature like that produced with BIST. Previous techniques have focused on extracting information hidden in the BIST signature, such as identification of fault-detecting test vectors.1-3 However, the test signatures’ highly compact nature, coupled with the inability to precisely model fault behavior, have confined such techniques to faults detected by only a few vectors. For faults detected by more than a few vectors, aliasing problems prevent designers from extracting all the relevant information. In a scan-based BIST environment, fault-detecting vectors typically outnumber fault-embedding scan cells (those in which fault effects are manifested). However, such scan cells are still plentiful enough to thwart techniques for extracting diagnostic information directly from the BIST signature. Consequently, researchers have attempted to gather more information by repeating the same test with different applications while adjusting signature analyzer parameters or observation outputs or by increasing signature register size.4-9 In particular, partitioning-based diagnosis schemes4,5,7 have been highly effective—and are possibly the only viable solution—for large industrial circuits. Partitioning-based schemes provide effective fault diagnosis in a scan-based BIST environment, but generating the partitions using pseudorandom approaches makes it difficult to predict diagnosis results.7 Both the linear feedback shift register (LFSR) parameters used for this partitioning and the faulty cell locations affect fault diagnosis time. Furthermore, the randomness of LFSR-generated partitions makes incorporating design-specific information into the diagnosis procedure difficult. Deterministically partitioning scan cells eliminates these complications, but identifying cost-effective hardware implementations of Cost-Effective Deterministic Partitioning for Rapid Diagnosis in Scan-Based BIST Scan-Based BIST Diagnosis