加权内建自测试(BIST)设计中的多目标优化技术研究-测试计量技术及仪器专业论文.docxVIP

万方数据 万方数据 摘 要 摘 要 随着数字系统的集成度和工艺复杂度的日益提高，数字系统内的器件测试变得越 来越困难。由于内建自测试(Built-In Self-Test, BIST)能在芯片内部完成自测试，使得 BIST 技术在测试方面得到了广泛的应用。但 BIST 在测试时也存在测试序列过长和测 试故障覆盖率不高等问题。因此，本课题重点研究一种既能保障测试故障覆盖率，又 能缩短测试序列长度的测试方案。 本课题主要研究内容如下：在研究 BIST 多目标优化问题的基础上，考虑编码方 式和适应度分配机制，引入带精英策略的非支配排序遗传算法(Nondom-inated Sorting Genetic Algorithm II, NSGA-II)。采用 NSGA-II 算法得到细胞自动机(Cellular Automata, CA)的规则集，获取 CA 结构，同时，通过构造最大海明距离和预设笛卡尔距离来获 取预确定距离测试序列，使每一个测试序列都能够检测到更多不同的故障。保障了测 试的故障覆盖率和测试序列长度间的平衡，实现 BIST 测试矢量生成优化。 针对以上算法的研究和设计，通过 ISCAS’85 基准电路对上述方案进行了仿真验 证，验证结果表明，本课题方案在保障测试故障覆盖率在 99%以上的同时，缩短了测 试长度，具有一定的先进性和实用性。 关键词：海明距离; 笛卡尔距离; 预确定距离序列; 内建自测试; 带精英策略的 非支配排序遗传算法. - I - Abstract Abstract With the increasing technological complexity and integration level of the digital system, it is becoming more and more difficult for device test in Digital system. BIST technology has been widely used in digital circuit’s test with the advantage of self-test ability. However, there are a series of problems in the testing process with BIST such as excessively long sequences ,low coverage rate of testing faults and so on.Therefore, we,in this study, pay particular attention to the program that not only can enhance coverage rate of testing faults ,but also shorten the length of testing sequences. The research approach of the study is as followings: Considering the encoding mode and fitness distribution mechanism, we have introduced Non-dominated Sorting Genetic Algorithm II (NSGA-II) based on the research of the optimization of multiple objects. Then, getting the regular collection of CA by employing NSGA-II, we’ve obtained the CA structure. Meanwhile, to make sure that each testing sequence can detect different faults as many as possible, we obtain the pre-determination distance sequence by structuring the most hamming distance and presetting cartesian distance. By adopting all the measures above, it maintains the balance between the coverage rate of testing faults and the length of testing sequences, realizing the optimizat