Issue link: https://iconnect007.uberflip.com/i/808383
48 The PCB Magazine • April 2017 just making sure that for a volume line, we've got suf- ficient cleanliness at the ap- propriate process step. But yes, we know how to build a volume line to do this today. We're pretty confident. Now we're waiting for the business. We remain close to our customers. The in- dustry is now looking at dif- ferent ways to employ embedded optical and you probably also know that companies have different approaches to their system architec- ture. Some companies make very heavy use of big backplanes; others not so much. Some use smaller mid-planes. Some use harnesses instead of backplanes. They all have slightly different architectures and they are working the benefits of optical interconnect into their designs. The result will be, I think, different solutions and different product. We are gaining a lot of experi- ence with these different approaches and solu- tions. We're waiting for a little more maturity in the industry to see how they're going to do this. We've been quite open with our customers. As soon as you have a design, even a pre-produc- tion type of design, we're ready to go work with you on it and make sure that we can put it in play. Right now, most of the industry is focused on test vehicles and demonstrators but they are quite sophisticated, by the way. They are func- tional and almost products unto themselves. Matties: When do you see this becoming more mainstream, where it's market-accepted and prod- uct-proven in the real world? How long a time frame? Davidson: Well, that's a really good question. We're asked internally and externally all the time. Matties: I bet you are [laughs]. Davidson: Our forecast comes from our custom- ers and potential customers. But we are focused on the 100-gigabit per channel node, which is predicted to be in full production sometime in 2020–2022. There is no com- mercial product yet. As far as the process technology is concerned, we're expecting that in the year 2020 we may need to be in production. We are planning the expansion of our pilot lines and low vol- ume lines now to understand the real business case for this. We want to be ready to go when companies have real products they want to put into production. Matties: Primarily this is tied to the server farms, the big server markets, communications—there's a real need for speed there, of course. Do you see this reaching, say, the automotive industry with the need for speed around autonomous vehicles? Davidson: I think it's certainly going to pen- etrate the infrastructure type of product—the big switches and hubs and other infrastructure product that's going to be needed to support autonomous driving. True autonomous driv- ing probably isn't going to happen until the 5G node is fully deployed. Things like latency in our Internet connections today needs to be quite a bit better for full autonomous. I don't see optical interconnect going into a car first. I see it mostly right now penetrating the types of infrastructure products that we already pro- duce. It'll take a little while longer before any- thing needs to be done inside a car—at least as to high speed requirements. As you know, we're also one of the largest automotive suppliers in the industry with a strong share in safety criti- cal applications. So, we're very familiar with the qualification requirements and reliability requirements of the auto industry. The time it takes to qualify new materials and processes is quite extensive in the auto industry. I don't see them adopting this as a leader; maybe a fast fol- lower as needed. Having said that, the automotive guys are subject to some things today that they were not before. They are buying commercial semicon- ductors that are fast and in big, complicated high IO packages. I see them advancing along the density curve—from where they were yes- TTM SHINES A LIGHT ON OPTICAL INTERCONNECT Figure 2: System enclosure with joint HDP OE2/PhoxTrot eco-system demonstrator.