Due to increased challenges to achieve better performance of ever more complex semiconductor devices, the cost of test keeps increasing. Some projections even forecast that test costs could exceed manufacturing costs. Despite a huge effort invested in better design using DFT (Design For Test), this is not enough by itself to sufficiently reduce the cost of testing without compromising outgoing product quality and reliability. Therefore, we have to consider techniques which aim to fit the manufacturing and testing flow for each one of the components produced, based on one hand on its performance, and on the other hand on the application it is being manufactured for.
ATTR (Adaptive Test Time Reduction) is such an IIoT solution that tackles the potentially damaging constraints of limited test time and actually increases quality, instead of reducing it. Rather than just removing a so-called “no-fail” test permanently from a testing program or skipping it at a pre-determined frequency, ATTR turns the test on and off at calculated intervals based on a very specific algorithm, so that if a manufacturing variation is observed, the sampling is automatically stopped. Moreover, using ATTR, lengthy tests that are typically removed because they are costly to perform – such as measuring the maximum speed range of a CPU – can be automatically reintroduced at pre-determined intervals in order to verify that quality and performance assumptions are correct.
Until now, one known limitation of such techniques was to comprehend feedback from different testers within the same facility and across different facilities. Due to the very high volume of parts being tested simultaneously in many testers, it is indeed not sufficient to take the feedback from a single tester to detect a manufacturing variation. Through a big data analytics infrastructure, it is now possible to stop sampling on the entire tester fleet once such a variation is observed on a single tester (even if it is located at a different facility). In a similar manner, this ‘cross-facility ATTR’ allows validation of a large number of units (at a scale of millions of units) for a specific test before initiating sampling again for this test. Such innovation allows deployment of ATTR in a safer way in production with real time feedback, and achieves as much as 30% test time reduction without compromising quality or yield.