An efficient diagnosis method to deal with multiple fault-pairs simultaneously using a single circuit model

Abstract

This paper proposes an efficient diagnosis-aware ATPG method that can quickly identify equivalent-fault pairs and generate diagnosis patterns for nonequivalent-fault pairs, where an (non)equivalent-fault pair contains two stuck-at faults that are (not) equivalent. A novel fault injection method is developed which allows one to embed all fault pairs undistinguished by the conventional test patterns into a circuit model with only one copy of the original circuit. Each pair of faults to be processed is transformed to a stuck-at fault and all fault pairs can be dealt with by invoking an ordinary ATPG tool for stuck-at faults just once. High efficiency of diagnosis pattern generation can be achieved due to 1) the circuit to be processed is read only once, 2) the data structure for ATPG process is constructed only once, 3) multiple fault pairs can be processed at a time, and 4) only one copy of the original circuit is needed. Experimental results show that this is the first reported work that can achieve 100% diagnosis resolutions for all ISCAS'89 and IWLS'05 benchmark circuits using an ordinary ATPG tool. Furthermore, we also find that the total number of patterns required to deal with all fault pairs in our method is smaller than that of the current state-of-the-art work.

An Efficient Diagnosis Pattern Generation Procedure to Distinguish Stuck-at Faults and Bridging Faults

“…Next we describe a Groups-Switch Cell design which allows us to put the multiplexers and related logic for all fault pairs in all groups into only one circuit. Similar design can be found in [15] where the main targets are stuck-at-faults only. Figure 8 shows how to apply Groups-Switch Cells to a circuit such that all Sel lines and multiplexers for all fault pairs can be put into one circuit.…”

“…For example, the work in [11][12][13][14][15] deals with stuck-at-faults only. With the advance of process technology, time-independent or DC types of defects become more and more complicated and the diagnosis analysis tools often report multiple fault types such as stuck-at, bridging and open faults as defect candidates.…”

“…In this paper we address the problem of distinguishing the two most commonly used fault types, i.e., stuck-at faults and bridging faults. Several diagnostic test generation methods based on circuit-model modification have been proposed [11][12][13][14][15][16][17]. The circuit model modification approach transforms the problem of distinguishing a fault pair into that of detecting a specific stuck-at fault by adding some logic to the circuit during the test generation process.…”

“…One drawback of [16] and [17] is they both select only one fault pair to process at a time and thus may generate a large number of diagnosis patterns if the number of fault pairs is large. Also both methods need to process two copies of the original circuit during pattern generation, hence may require large CPU time as illustrated in [15] where it has been shown that much more CPU time is required using a two-copy circuit model than using a single-copy circuit model.…”

“…In [15], an efficient diagnostic ATPG method also using only one copy of the original circuit is proposed. This method further allows one to put all fault pairs to be distinguished into one circuit and process multiple fault pairs simultaneously.…”