Issue link: https://iconnect007.uberflip.com/i/994883
12 DESIGN007 MAGAZINE I JUNE 2018 and play it with something else versus throw- ing away the whole system. Should I do it for cost reasons? Performance reasons? And then there's the make versus buy decision, right? For this function of my system, should I just buy a board and connect it to my system? Or should I design it into the system? You see that with your home PC. It's got ded- icated graphics built into it now, but you still may want to put an add-on graphics card, but those used to be things that weren't necessar- ily integrated into one system. They've done that over time. Shaughnessy: It sounds like some of the big- gest challenges are the trade-offs make. Like, should you add another board? What are some of those challenges? Wiens: I'd loosely group the design challenges into four categories: managing design com- plexity, enabling multi-disciplinary team col- laboration, optimizing product reliability, and leveraging IP across the organization. Regard- ing design complexity, yes, the trade-offs are a huge one and, frankly, because of that a lot of people kind of avoid it. I mean, they make decisions very early on to segment something. They take their high-level requirements, and then they start building an architecture from that. They decide what's going to be hard- ware, what's going to software, and how that's going to interplay with mechanical structures, cabling structures, all of that is decided way up front, and then the question is: Do they ever revisit that? If they don't revisit it, then they were either really good and they got it right up front, or as the design progressively refines itself, as the engineers refine it from that high- level architecture down into the physical imple- mentations, they may find things. For instance, the architecture indicates, "I decided to put these two boards over here." But maybe they don't fit. Maybe they are too far apart for those high-speed signals from the driver on one board to the receiver on another board. There are things that you don't know at that top level that as you progressively refine you get smarter and smarter, and so you want to make those trade-offs progress through the design. If you locked and loaded your design way up front, then you've lost that alternative. That's the case of most teams. Because it's so painful to make changes after that initial archi- Figure 1: A fully integrated process for designing multiple PCBs and associated cables.