Issue link: https://iconnect007.uberflip.com/i/1295812
OCTOBER 2020 I DESIGN007 MAGAZINE 17 would keep pace if we could provide tools with the right capabilities. In the SI space, at least, the shortage of experts isn't getting better, and it's time to do something different. We have tools for the experts; now, we need tools for the broader community and examples of how to apply those tools to common design prob- lems. We provide user workshops that show how to apply HyperLynx to a typical design prob- lem. My point here is that training and educa- tion are just as important as the tool capabili- ties themselves, especially when we're talking about a broad audience. We've been condi- tioned to think about which things are critical when analyzing a high-speed serial link, but most of our input has come from people prac- ticing at the state of the art. In our 25-gigabit serial link workshop, for example, the effect of surface roughness just doesn't matter that much. It's case-dependent, of course, but that's the point. Showing people how to isolate and quantify the different things that affect system margin is every bit as important as having the capability to do that analysis. How are people supposed to know what to do if you don't show them? We all have these ideas about what matters and what doesn't based on what we've been told, but sometimes the actual results are quite surprising, providing additional opportunities to optimize a design. Holden: The overall truth is that two things are conspiring against us. The first one is the schedule. Nobody's going to let us expand the schedule. They constantly seem to shrink because of time to market and who gets there first. The second thing is complexity. To attract customers, everything has to have more fea- tures and be more complex. Johnson: You're absolutely right in terms of schedule. If you bring basic analysis to the desk of the engineer, you helped your schedule by a large amount. If you can put basic tools in that designer's hands, they can make a lot of those decisions. You'll actually win some on schedule there. Westerhoff: I think it's a diminishing return problem. How much can we accomplish if our analysis accuracy is at the 90% level? A lot. If we defer all analysis until the end of the design, we run the risk of finding mistakes that will blow the schedule. When it comes to design analysis, we're saying less is effectively more. Shaughnessy: Less information available quickly is better than 100% of the info two weeks from now. Westerhoff: Exactly. Let's be clear; we're not saying that we don't need SI experts doing what they're already doing because we do. We're arguing that most designs have an "anal- ysis valley" in the middle of the design cycle, where systems designers can't get the feed- back that they need. Putting design-oriented analysis tools in their hands will enable them to make more informed design decisions and help offload the SI experts from what is prob- ably the worst part of their jobs, running post- layout verification on other people's designs. Holden: Mentor and Siemens have access to much of the electronic design space. Do you work with the thermal analysis or placement and routing side teams to help designers juggle the multiple balls that have to be kept in the air? Westerhoff: Absolutely. That's a natural exten- sion of everything that we've discussed here. The problem isn't unidimensional anymore; it requires balancing electrical, mechanical, and thermal trade-offs. How do you extend this design-oriented analysis strategy to play across all those disciplines? That's where we need to go next. Perhaps we can talk about that the next time we meet. Shaughnessy: This has been really interesting, Todd. We appreciate your time. Maybe we'll see each other "live" again. Westerhoff: Thanks for the discussion. This has been great. It will be nice to talk in person again. DESIGN007