There's the old story about the frog in a pot of water which comes to a boil slowly enough that the frog doesn't realize what's going on until it's too late. The water in the Semiconductor Test pot is getting warmer and warmer for a few reasons:
- Labor shortages: For a lot of reasons, it's hard to find new test engineers to replace ones that are aging out of the workforce. - Devices: As we all know, the parts we test continue to get more complex and are release at a faster cadence. - Skill set and tools: For most companies, the skill set in test organizations is very narrow and most of the tools we use are, to say the least, dated or home-brewed.
How does this manifest itself? - Poor quality test programs - Test Escapes or lower yield - The inability to choose the best test solution because "it's too hard to change" - Poor integration (still!) with design and DFT teams
These are all problems that are solvable with existing technology if only we'd do things differently. Like the frog, though, most of us just can't seem to get out of the pot. It's a cultural problem.
This paper will talk about where we are today and show some ways forward. Most of these are software related, such as a more disciplined test development flow, including code management and testing. It also depends on redefining "test engineer" to "test team" to include the new realities of EDA integration, Data Analytics, and massive amounts of test program code for characterization and production. Lastly, it depends on a realistic way to apply AI to practically help the test engineering ecosystem instead of having it perpetually be limited to Data Analytics.
