Issue link: https://iconnect007.uberflip.com/i/808383
46 The PCB Magazine • April 2017 a long time. The TTM team in Europe has been working on this for a long time now. And you interviewed Marika Immonen on our team back in 2015, I believe. TTM now has technology we can offer that will allow embedded waveguides in boards. This includes polymer waveguides either buried inside a board or built-up on the board surface and with in-plane or 90° connectors. We›re working jointly with several consortia and individual companies to demonstrate this technology. So that's a brief history of where we have been and at least a little preview of what we can offer. TTM has a long experience of fabricating multimode waveguides for short-reach datacom applications. Now as silicon photonics at OEMs is pushing through, we are scaling technology to support their single-mode roadmaps. There we pursue both polymer- and glass- based waveguides. Polymers are very versatile, low cost and easy to fabricate, whereas glass provides low loss at the longer wavelengths and optical compliance with fibers. Single mode waveguides are looked at to provide complex routing between chips or to serve as "bridges" between sub-micron silicon waveguide and 9-micron fiber. In single mode, accuracy and registration both in waveguide fabrication and termination is critical. Matties: And really the impetus behind this is the speed barrier, right? Davidson: Yes, it absolutely is. We're now in production with 25-gigabit backplanes, and by that I mean per channel, so it's some of the fast- est boards out there these days. We know that companies are already working on product for the 50-gigabit node. The generation after that is also being developed now and will be around 100 gigabit per channel. I think it's those speeds that people are most focused on for these opti- cal interconnects. We know that even 25 gigabit is difficult to manufacture because of the tolerances required. The 50-gigabit node will be very difficult to manufacture, again because of the losses. Low-loss materials are required and the tolerances on the copper and dielectric geometries in all di- mensions are going to be very demanding for even the 50 Gb. The consensus from our cus- tomers seems to be that with the 100-gigabit node we really will have to be thinking pretty hard about moving to optical interconnect onboard as more traditional PCB fabrication techniques may not be capable of the tolerances required. Matties: Now, with regard to high-speed materials and the traditional coppers, haven't they broken barriers that have been somewhat surprising to the industry? Barriers that technologists had previous- ly predicted would never be breached with copper? Davidson: Yes, I think that's true. I remember hearing "Oh, geez, we need flip-chip next year or we're out of business." Or in 2000, they said, "If we don't have embedded optical in a year or two, we're out of business." None of those things came to pass. It's remarkable how robust existing technology is and it's always so much easier to make incremental improvements to what exists than to implement a whole new type of technology. This 100-Gb barrier might be a little bit dif- ferent because it does not rely just on opinion regarding what future manufacturing technol- ogy will be able to deliver but on fundamental calculations of losses associated with current materials. Other factors may delay the require- ment of embedded optical. That could be soft- ware solutions such as error correction or ad- ditional functionality such as on-board repeat- ers. But the physical solutions come with severe power and thermal dissipation requirements. Nevertheless, embedded optical is a focus for us and our customers. In fact, we have man- ufacturing techniques now that we're ready to introduce into pilot and volume production. We want to be able to deliver optical intercon- TTM SHINES A LIGHT ON OPTICAL INTERCONNECT Craig Davidson