EAGER: Storage-Based Logic Built-In Self-Test with Magnified Test Data

电子芯片制造的新技术产生了新的缺陷机制，增加了缺陷的发生率。同时，汽车系统等应用对可靠性提出了更严格的要求。检测缺陷的发生对于电子芯片的可靠性以及使用它们的许多产品的正确和安全操作至关重要。用于电子芯片制造测试的测试装置需要具有全面的缺陷检测和诊断能力，但同时还要满足与存储和测试应用时间相关的成本限制。每个测试都需要最大限度地利用来检测尽可能多的不同类型的缺陷，以确保成本有限的测试集能够实现可靠性目标。这样的测试集被称为焦点测试集，是本项目的主题。该项目将帮助学生为设计和制造当前和未来技术中可靠的电子芯片的挑战做好准备，这对许多依赖电子芯片的美国行业至关重要。PI将提供一个榜样，并鼓励女学生攻读电气和计算机工程领域的学位。通过实习机会，学生将获得学术和行业深度的综合训练。为了管理缺陷行为的复杂性和可变性，使用故障模型作为缺陷的抽象表示，并针对从不同模型中检测故障进行了一套全面的测试。然而，来自不同模型的故障具有相似的测试，并且由于考虑了额外的故障模型，缺陷检测的附加价值减少了。本项目将探索两个互补的方向来生成集中的测试集。(1)使用不同的故障模型来提供全面的缺陷覆盖。然而，如果所有类型的故障都被认为同等重要，那么大量的测试结果，并且它们对于检测或诊断缺陷并不都是必需的。相反，不同的故障应该作为彼此的补充。本项目将考虑的问题是如何将目标故障划分为具有相似（但不相同）检测条件和相似（但不相同）测试的子集。生成集中的测试集的过程将从每个子集中选择不相似的代表作为测试生成的目标，从而产生用于缺陷检测和诊断的测试。(2)在基本的试验数据压缩方案中，每个存储的试验用于应用一个试验。早期的工作表明，相同的存储测试数据可以以多种方式使用，以应用更多的测试。这样可以减少测试数据的量。这个项目将利用的新观察结果是，基于相同存储的测试数据应用更多的测试对于提高集中测试集的质量是有用的。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

New technologies for manufacturing of electronic chips give rise to new defect mechanisms and increase the prevalence of defect occurrence. At the same time, applications such as automotive systems impose stricter requirements on reliability. Detecting the occurrence of defects is critical to the reliability of electronic chips, and the correct and safe operation of the many products that use them. Test sets used for manufacturing testing of electronic chips need to be comprehensive in their ability to detect and diagnose defects, but at the same time, satisfy cost constraints related to their storage and test application time. Each test needs to be utilized to the maximum to detect as many defects as possible of as many different types as possible to ensure that a cost-limited test set is able to achieve reliability goals. Such a test set is referred to as a focused test set, and is the subject of this project. The project will prepare students for the challenges of designing and manufacturing reliable electronic chips in current and future technologies, which is critical to the many US industries that rely on electronic chips. The PI will provide a role model and encourage women students to pursue degrees in the field of electrical and computer engineering. Through internship opportunities, students will get comprehensive training with both academic and industrial depth.To manage the complexity and variability in defect behaviors, fault models are used as abstract representations of defects, and a comprehensive set of tests targets the detection of faults from different models. However, faults from different models have similar tests, and the added value for defect detection diminishes as additional fault models are considered. Two complementary directions will be explored in this project for generating focused test sets. (1) Different fault models are used for providing comprehensive defect coverage. However, if faults of all types are considered equally important, a large number of tests results, and they are not all necessary for detecting or diagnosing defects. Instead, different faults should be used as complements of each other. The problem that will be considered in this project is how to partition target faults into subsets with similar (but not identical) detection conditions, and similar (but not identical) tests. A procedure for generating a focused test set will select non-similar representatives from each subset as targets for test generation, and thus produce tests that are well-utilized for defect detection and diagnosis. (2) In a basic test data compression scheme, each stored test is used for applying one test. Earlier works showed that the same stored test data can be used in more than one way to apply more tests. This allows the volume of test data to be reduced. The new observation that this project will utilize is that applying more tests based on the same stored test data is useful for improving the quality of a focused test set.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.