Align-Encode: Improving the Encoding Capability of Test Stimulus Decompressors

Abstract

While test stimulus compression helps reduce test time and data volume, and thus alleviates test costs, the delivery of certain test vectors may not be possible, leading to test quality degradation. Whether a test vector is encodable in the presence of a decompressor strongly hinges on the distribution of its care bits. In this paper, we present a technique that provides an on-chip capability to judiciously manipulate care bit distribution of a test vector. We thus propose a hardware block, namely, Align-Encode, to be utilized along with any decompressor to boost the effectiveness of the decompressor. Align-Encode is reconfigured on a per pattern basis to delay the shift-in operations in selected scan chains, in order to align the scan slices in such a way that more test vectors become encodable. The reconfigurability of Align-Encode provides a test pattern independent solution, wherein any given set of test vectors can be analyzed to compute the proper delay information. We map the delay computation problem to the maximal clique problem, and utilize an efficient heuristic to provide a near-optimal solution. Experimental results also justify the test pattern encodability enhancements that Align-Encode delivers, enabling significant test quality improvements and/or test cost reductions even when used with simple decompressors.