Semiconductor test data collaboration: One source of truth, faster decisions
Imagine it’s Wednesday morning in Singapore and a test engineer in Singapore spots a yield excursion on a lot that left the German fab two weeks ago. In most companies, that means a file transfer, a chain of emails and a meeting on Friday.
Now imagine the other version:
- She looks at the data for the batch of chips that were tested. It is the same cleaned-up view her colleagues in Germany and the US are looking at.
- She finds a strong correlation between a fab WAT parameter and the wafer sort bin, and shares the chart and analysis in one click.
- A test development engineer in the UK left a comment on that test months ago. It appears whenever that test appears on future lots, so she sees it before she needs to ask anyone.
- The alert reaches engineering, the subcon and operations at the same time. Nobody needs to be talked through what happened.
- Her manager opens a Bird’s Eye view across every team and subcon instead of requesting a status update.
By Wednesday afternoon, it’s a decision, not a meeting invitation.
The difference is one set of data
Nothing in the second version is magic. Every step comes from the same design decision: the whole company works from one set of data instead of their own copies of it.
In yieldHUB, that means a relational database holding production data from every subcon, updated several times a day, with everything linked: fab PCM/WAT data, wafer sort and final test data, characterization data for new products, MES data, genealogy, inline fab data where available, and the comments and knowledge that engineers add along the way.
That last item matters more than it sounds. The unusual failure pattern in the story can be investigated quickly because the fab parameter and the sort bin live in the same database and can be correlated in one view. But the reason our engineer in Singapore doesn’t have to ask anyone about that test is the comment a colleague in the UK left months earlier. Observations are saved in contextual threads throughout the system, so they appear wherever the test appears, on every lot that follows.
Knowledge that stays when people move on
Our founder and CEO John O’Donnell has described the problem this solves better than I can:
Before yieldHUB, I spent many years in the industry and saw a recurring challenge: data silos. This happens when engineers work in isolation, keeping critical yield data on their own computers instead of in a centralized location. If their knowledge and data aren’t properly documented or accessible, the company may lose crucial information. The next person who takes over is left scrambling to piece it together when an issue arises, wasting valuable time.
With one shared platform, knowledge isn’t lost when someone changes role or leaves. It stays within the company, attached to the tests, lots and products it belongs to, and it builds up every day.
What the second version is worth
Our users report an increase in productivity of around 20 percent. That is an extra day per week, per engineer, spent solving problems instead of gathering data.
“It used to take us three hours to convert multiple STDF files. We now have our data in five minutes. The Test Engineer can get the desired results without wasting time.”
Byungwoo Han, Principal Engineer, Test Development, ADTechnology Inc.
That second version is what I was in Singapore talking about back in early July: what changes when teams in different countries work from one set of data instead of their own copies of it.
Which version is your Wednesday?