摘要： 扫描森林是一种有效的扫描结构，它能够大幅度地降低测试应用开销、测试功耗以及测试数据容量。该文针对采用扫描森林结构的待测电路提出了一种新的种子编码方案。在该方案中，伪随机测试向量覆盖电路中的易测故障，使用ATPG对剩余故障生成确定性测试向量，将其中某一测试向量对应的种子编码为LFSR扩展成该向量需要运行的时钟周期数。实验结果表明，提出的方案能大幅度地降低种子存储数据量，最大降幅达到了83.3％。

关键词: 自测试, 扫描森林, 种子编码

Abstract: Scan forest is an efficient architecture which can reduce the test application cost, test power of scan testing and test data greatly. This paper proposes a new seed encoding scheme for CUT with scan forest architecture. Under this scheme, pseudo-random patterns are initially applied to cover the easy-to-detect faults. Test patterns are generated by an ATPG for the remaining faults. For a given seed determined by a generated ATPG pattern, encodes it by the number of clock cycles the LFSR needs to run to reach it. Experimental results show this encoding reduces seed storage by up to 83.3% compared with conventional method.

Key words: Built-in self-test, Scan forest, Seed encoding