DESPITE THE GROWING complexity of semiconductor devices, adaptive test deployments for such devices remain elusive. The high cost and difficulties of developing a custom adaptive test solution are likely causes for this sparsity. Indeed, there are many requirements that an adaptive test solution should meet. A good adaptive test system should dynamically adjust to process variation statistics that may not be stationary. It should modulate the test program at high granularities (die level or even sub-die level), have low adaptation latency to achieve real-time adaptivity, and comfortably handle learning from terabyte-scale historical test data. Moreover, it should achieve these goals without violating stringent industrial requirements on permissible test error. Finally, many of the specific details of these requirements are likely to differ across product lines, e.g., consumer products certainly have very different comfort levels for permissible test error than automotive products. However, the advantages of a successful adaptive test deployment are clear: test time reduction, test quality improvement, improved data analysis ability, and acceleration of yield learning. Therefore, substantial industrial interest exists for making adaptive test a reality in the near term. This interest is driving groups from a broad swath of industry to collaborate on adaptive test development. The Test Test Equipment team within the International Technology Roadmap for Semiconductors (ITRS) has created a subgroup specifically targeted at the adaptive test problem 1.
展开▼