Issue link: https://iconnect007.uberflip.com/i/1295812
74 DESIGN007 MAGAZINE I OCTOBER 2020 connects. Chiplets may become a moderately- used term, but we may also call it other things. Holden: For me, this is an old topic because by 1972, we were putting multiple gallium arsenide LED dies on a PCB and wire bonding them over for second-generation calculators and things like that. Everybody thinks this is a 21st-century technology, but most of them aren't aware that back in the early '70s, we were doing a lot of chip-on-board and multiple chip-on-board. Wesling: If you read chapter eight, you'll see that we still see wire bonding as a high-use application for bonding chips, three- and four- layer tiers, etc. Wafer-level packaging and flip chip may be coming in, but there's still going to be a lot of wire bonding. Johnson: Happy's comments harken back to a time when all manufacturing took place at the same company. And because it was all under one roof, you could develop your own protocols and communications. It's different now. The various functional roles are fulfilled by third- party specialists. The need to communicate with multiple OEMs at the manufacturing level about this very detailed information is critical. For the PCB manufacturers, this technology could be the point where, no matter how well our systems work, they're going to break. Wesling: That's why we've put in a supply chain chapter, which is not too developed, but we'll soon have one that's more extensive. One pain point is when only one company fol- lows a standard. If we look back in five years to what we've published now, part of what we publish will not have been adopted. These are not standards; these are options for com- panies, consortiums, and university teams to get together and see if it works. Consequently, you'll see lots of potential solutions. Probably only a few will get implemented. The other thing is in the supply chain. Our supply chain is spread out. Somebody does design, somebody does fabrication, and some- body else does assembly. This becomes a chal- lenge in both supply chain and security. Look- ing into the crystal ball and knowing what will happen is the tough part. That's why this roadmap extends to about 600 pages because it covers lots of possibilities—not all of which are going to happen—that can keep people in jobs for the next 10 years, for example. You want to look at how it affects what your focus is, which is substrates and PCBs, and— more specifically—do they go away and get replaced by some other multi-chip substrate? Does polymer end up going somewhere? Does low-temperature processing start coming in here? What kind of things do you see that you can pull out of here that might be interest- ing? Once you get some draft ideas, you might kick them back to the people in the working group—because they're all listed at the end of the chapter—and say, "I'm thinking this. Is that still what you're thinking?" This morning, I thought, "This is going to be about setting the ground rules and projections, and then every few months, coming up with something else—either digging more deeply into something that wasn't covered before to say what could happen or correcting what you thought earlier." Over the next months, we hope to have many of the updated chapters out for the 2020 version. Several of the chapters are being fully rewritten, but most of them are just being updated. Johnson: The change to heterogeneous inte- gration and the use of chiplets really is based on using an interposer, which ties together unpackaged chips into a system. Is that basi- cally a PCB inside the system-in-package (SiP)? Wesling: Yes. It's between silicon and the next level—the interposer (Figure 2). Johnson: The interposer employs an intercon- nect design methodology—more than you would normally use in an active piece of sili- con. Does that mean that PCB designer experi- ence is going to be valuable on the wafer design side? Do the design tools need to change? Wesling: Things like that are analogous to what PCBs have done in the past but at much higher