Issue link: https://iconnect007.uberflip.com/i/1428512
66 PCB007 MAGAZINE I NOVEMBER 2021 some 3D microscopy onsite; how did that change the ability to track your quality? Did you get the results you expected? Stepinski: e use of those tools gave us a differ- ent perspective on the manufacturing process. at factory was able to eliminate the stacked microvia failures and hit tolerances that had not previously been hit in the market by lever- aging those tools to build the quality in; and continue to do so. Johnson: ere's a correlation, then, between test and inspection, and the processes? Stepinski: Yes. e important thing is to have a data link. is is the new buzzword for our market. Very few shops have this. Maybe a few in Asia, but in the U.S., we're missing this right now. A data link allows you to have a reposi- tory for all this information so you can corre- late things together, do analytics, and be able to come up with very good sampling plans that minimize your overall costs, raise the quality level, and add value to the customer. I think it's a big opportunity and it's not a hard thing. Because it's common in other industries, there are a lot of other prior learning and resources that you could leverage to accomplish this. Happy Holden: I agree. We need to have good AI case studies because it's not easy to originate all the ideas. But it's a lot easier to take some- thing that's working—somebody else's idea—and apply it or brainstorm it. Stepinski: ere are a lot of compo- nents to artificial intelligence to make it successful. e first is the inherent process capability, how much process variation is there. You can't do artificial intelligence with garbage equipment. It just doesn't work out. You need a suf- ficient level of process control in order to know that when you optimize some- thing, the process hasn't changed. e next step is the data engineering, mak- ing sure that the data coming to you is good and in the right format to do something with it. Typically, in these kinds of artificial intelli- gence practices, that's the biggest component. Aer process engineering, it's the biggest so- ware component. Once you get through that, then you're into the data science, and that's the core algorithm to figure out what's going on. It's probably the smallest piece of the puzzle. en you follow that with the analytics and the feedback. You need all those things to make it work. Can you do it beyond registration? Absolutely. You can do it with plating, and with etching. But you need to get the first piece working, and know that the process is capable. You have to do your studies and identify this, because arti- ficial intelligence requires a good Cp to have a good Cpk. If you have a poor Cp, you won't have a good Cpk. Just basic statistics, right? So that's what it's for; it's going to give you the best Cpk for the native capability of the equipment if it's all executed properly. I think this is a great opportunity in the market. ere is a lot of potential everywhere to add more value and it reduces costs without the new tool CapEx if the Cp is already there. Matties: As you were hitting ratios and such that others haven't achieved, that created a new market opportunity where customers started coming to you because of your capa-