Issue link: https://iconnect007.uberflip.com/i/1207026
FEBRUARY 2020 I SMT007 MAGAZINE 17 curve and get the yield up high, it's not a prod- uct that's going to fit in that format, even if the material would be ideal. _____________ What's Going to Give? Johnson: What's going to give first? It sounds to me like the answer to what's going to give first is the substrate. Holden: The substrate has always been the unique piece of it. You take the standard chips, and how the customer sees them, and it's often oriented and packaged along with some of the other stuff we add to the circuit. That's an infi- nite variability there working with standard chips and components. Matties: If we put this in terms of an assem- bler today who is putting together their five- year roadmap for their business, what should be their considerations? Korf: We already see 0.4-, 0.3-, and 0.2-millime- ter pitch devices go on boards. With 0.16-mil- limeter pitch, what's driving RDLs are MSAP, SAP, and a bunch of processes trying to get that pitch down. It's also driving electrical test and solder mask capabilities, and it's hard to lay a web down that small in solder mask and get a high-speed, high-volume contact that's good when the pitch gets that small. _____________ Active Embedded Matties: I was recently visiting AT&S in Austria where they have been doing the active embed- ded for a while. It requires new testing and handling challenges and new disciplines. If a company is putting together a roadmap, is that an emerging opportunity? Korf: It's an area that always gets talked about, and I came back from several years in China this year. People would say, "We're ready to do it. Who wants to buy it?" Well, nobody. Only a few companies are doing it. It's only ductor companies, one of the things that you learn is that that sort of volume creates its own unique set of problems and challenges. Holden: There's a new generation of lasers that will cut glass in milliseconds without any sharp edge. I'm not sure if it's an X-ray laser, but you can shave up that one-meter-square piece of glass into 10,000 individual pieces that are all perfect. Afterward, you can sputter additional layers on it, not the initial ITO, but a carbon nanotube. Because of the size, the unit cost is much better than using silicon PCBs made out of wafers as a packaging substrate. Glass can have better mechanical and electrical proper- ties than silicon but be much larger as a panel. _____________ Routers Bauer: Cisco would love to have a memory device that was 512 bits wide and four bits deep, but nobody will make it for them. The world market for that memory device because that device would be perfect for switcher rout- ers. It would help them make them much faster and more efficient. It would make them more reliable because there wouldn't be so many different board assembly operations. However, if one supplier was making them, they would not have to make more than 3,000 or 4,000 wafers a month to supply the world market. For them to come down the learning curve and get their yields up requires a minimum of 5,000 wafers, and they don't even look at a market unless it requires 20,000 wafers a month and be able to sell it at a reasonable price. To get a memory maker to design a 4x512 memory, you can't pay the company enough money to do it. It's cheaper to take the 16x32 memories, line them up on a board, use them as 4x512, and configure them. It's cheaper because you can use the reject die from the high-density memory companies. Doing glass on the meter-square sheet is going to create that same kind of challenge. If they don't have enough volume to come down the learning