The Pitfalls of Necessary Assignments

Abstract

It has been shown that, finding necessary assignments during a search process, such as Automatic Test Pattern Generation (ATPG) process, can significantly improve its search performance. The techniques to find necessary assignments include static learning, dynamic learning, and recursive learning. All these techniques did improve ATPG performance. However, in our experience with real circuits, we found that necessary assignments can create unnecessary requirements in an ATPG process. Sometimes, these unnecessary requirements are not justifiable such that a testable fault may be mistaken as untestable.