Issue link: https://iconnect007.uberflip.com/i/1014812
AUGUST 2018 I DESIGN007 MAGAZINE 13 I think there's always a cost trade-off when you try to integrate too much into a chip, and if you do integrate too much, then the question is: Does that narrow the market for that par- ticular device? I think we all know that for chip makers in general, if there's no volume, there's no opportunity. The memory guys know that in spades. That's why we don't have 128-bit wide memories put into switcher routers for the internet. They're still using relatively nor- mal memory devices and having to construct them onto substrates to create "synthetic" wide band memory, simply because there's not enough demand worldwide to drive the manufacturing above 4,000-5,000 wafers a month. The memory guys will tell you that the minimum that they must have is 10,000 wafers a month to come down the learning curve, and they don't really want to touch anything that's less than 50,000 wafers a month. I think that's going to be a different trade-off and that may be an area, the SiP area, where we begin to see some expansion of some wafer-level packages. Andy Shaughnessy: You speak at conferences and moderate panels, and you must hear a lot of questions. What are the big challenges that you're hearing about, especially from the design side? Bauer: Well, I think that in the packaging arena, most of the challenges that I hear about are materials-related. There are some that are pro- cess related, and most of those in terms of uni- formity, repeatability and consistency, those sorts of things. Warpage is an issue, especially in stacked 3D packaging. Most of those issues all come back to the materials, so there's a lot of interest in the materials. I will say that the problem is that nobody wants to pay anything more for the materials and that's always a par- adox in the conflict between the materials sup- pliers and the manufacturers. Especially since the volume of material going into any particu- lar package has reduced dramatically over the years, even as the number of packages has gone up. It's been a dichotomy for the material suppliers as well. I think Happy is probably familiar with how long it has taken BCB (benzocyclobutene) to really become an established player, and BCB really came along in the 1980s as the substrate material for BGA packages. It's now starting to hold its own, but even today, when it comes down to price, people will go to FR-4 or FR-5 if they can. I think that that's the biggest chal- lenge that I hear on a consistent basis—find- ing materials that'll do what you want them to do at a reasonable cost. And things like nano materials which I mentioned earlier. Nano materials are getting a lot of interest now, not necessarily for broad-based assembly, but cer- tainly for assembly of high value-added com- ponents onto boards after it has been assem- bled and tested, and then they can put the high value component down after the fact. Shaughnessy: We hear a lot of designers talk about how co-design is such an advantage. Do you see a lot of that going on? Bauer: Happy and I have both been talking about that since the early 1980s. I remember the head of hybrid design at Tektronix when I first started, about 1984. He said hybrids wouldn't exist in 5-10 years because we'll be able to integrate it all into the silicon. I think we all know where that went. Whether you call it a hybrid or a multi-chip module or a sys- tem in package, we're still doing it. I think the talk about co-design in most com- panies is just that. Unfortunately, there are two types of design that I would call design. One type refers to designing a product, where you have to incorporate all the manufacturing tech- nologies together. The other type is the actual Whether you call it a hybrid or a multi-chip module or a system in package, we're still doing it.