Finally, Realizing The Full Benefits Of Parallel Site-To-Site (S2S) Testing

A very common and well-known practice by manufacturers during the IC test process is to test as many of the device die or packaged parts as possible in parallel (i.e. sites) during wafer sort and final test in order to increase test time efficiency and lower overall test costs. The constraints that typically restrict how many test sites can be used at any given time are the design I/O and capacity specifications of the specific chip under test and the limitations of the designated ATE used.

However, one of the difficulties in a multi-site approach is the ability to quickly identify and triage issues that may arise during testing. For example, test hardware interfaces such as probe cards, PIBs (probe interface boards), and sockets may deteriorate on a per site basis over time, leading to false failures of potentially good chips and thus an unexpected decline in yield. The degradation may be subtle at first which exacerbates the problem since the issue may get masked as chip test yield results may not fluctuate drastically when comparing results across multiple sites.
The challenge is how to track this information quickly enough so that when yield issues occur, they can be immediately flagged and corrective action taken. This is where big data analytics come into play. Using big data approaches to analyze site-to-site (S2S) test data for their entire test fleet, companies can set thresholds for yield deviation across their global supply chain to monitor and react to problematic variations in real-time. While this approach would have been traditionally seen as a very costly capability to put into a production manufacturing environment, it is already being done today with no impact on overall test time, enabling companies to significantly improve overall product yield, in some cases by as much as 2%, and enabling them to fully realize the benefits of parallel S2S testing.

Figure 1 below shows an example of a site-to-site yield rule where a yield deviation greater than 10% between one site to the overall average of the remaining sites would issue a warning and notify the operations engineer to take specific actions. In this example, users can also set the triggers for the yield deviation across multiple sites or can monitor site yield relative to total yield.

Figure 1. Yield Impact Rule

Figure 2 shows the results from a triggered S2S rule. When the site to site yield gap was greater than 10% (almost 70% in this case), a warning was issued. This alert enabled an operations engineer on the manufacturing floor to take immediate corrective action, which in this case was to specifically look at test site #4. The operations engineer can either decide to temporarily shut down testing until the problem is fully identified and resolved or they may opt to temporarily disable site #4 while allowing the remaining sites to continue operation. Either way, by taking immediate action, product yield loss at site #4 is immediately avoided.

Figure 2. Real-world example of a triggered S2S rule

Site-to-site rules are an effective and efficient way to use big data analytics to enable real-time, 24×7 monitoring of site specific issues that can improve test efficiencies as well as overall product yield. With the right big data analytics approach, early detection of site specific issues at wafer sort operations (scaling from operations that test two sites per touchdown to thousands of sites per touchdown) can also help improve quality by reducing the number of unnecessary subsequent touchdowns for testing of a perceived bad die and thus lowering the risk of damaging the die.

Today, OSATs who incorporate big data analytics and site-to-site solutions into their environment are confident that they can easily track and immediately respond to site specific issues as they occur even if they are testing multiple parts from multiple customers on multiple testers in parallel. Fabless semiconductor manufacturers who utilize site-to-site data analytics now have an effective way of monitoring all test sites in parallel in order to detect site-to-site variability that can impact yield or quality and reduce overall manufacturing costs.

To learn more about how semiconductor companies are using big data analytics in their manufacturing operations, visit us at http://www.optimalplus.com.