Issue link: https://iconnect007.uberflip.com/i/1219242
28 DESIGN007 MAGAZINE I MARCH 2020 ions and reviews, even when a first-order anal- ysis would have allowed them to proceed. The truth is that the experts don't want to be both- ered with simple problems any more than de- signers want to wait for the expert's opinion, so why not break the logjam? We have found that combining DRC and first-order simulation can be incredibly effec- tive. We use DRC to scan the design and locate areas of interest, and then use first-order sim- ulation to see how that part of the design be- haves. It turns out not to be about the details of the I/O behavior, but about how the inter- connect behaves because that's what system designers are really designing anyway. The exact details tend to vary from case to case, but the consistent focus is on iden- tifying a suspect area of the de- sign, using the right level of EM modeling to capture particular types of effects, and then uti- lizing simulation to excite and observe the area's behavior as simply as possible. We have been able to create entirely au- tomated processes for customers this way. A PCB designer checks in a board at the end of the day and a set of simulation results are wait- ing the next morning. Shaughnessy: On a different note, do your cus- tomers come to you with demands for cost- aware design capability? We ask people if they know the total cost of their designs, and they usually have no idea. Westerhoff: I think most designers don't have good, reliable access to cost information. In most companies, those details live in some oth- er department, and by the way, costs change with time, so the cost to build a board when I design it probably isn't what it costs in pro- duction. If you're talking about total cost, in- cluding engineering costs and other overhead, I don't think designers see that at all. The bits and pieces of that data are all over the compa- ny, and by the time it all gets pulled together, it's probably only shared at the executive level. What does it mean to design products for extreme volume manufacturing? Chances are the people who designed my fitness track- er had a pretty good idea of what production costs were going to be because they expected volumes in millions, and a fraction of a penny mattered. The people who designed my lap- top probably had a different idea of volumes and cost sensitivity. In my "big iron" network- ing days, the first revision of a design was about functionality, performance, and getting to market; cost reductions came later because volumes were lower and the product lifetime was much longer. Let's look at this from the standpoint of an SI person and consider the case of a popular consumer product. Volumes are high, and design cost is critical. One rule of thumb for high-speed design is to provide a clean return path by routing signals on a ground plane. How- ever, adding ground planes—if it comes to that—costs money, and your cost target may not be able to support that. There's a strong incen- tive to bend the rules, reference traces to pow- er, ground shapes on the same layer, or some- thing else. The trick is that it raises the model- ing and simulation bar significantly. To make sure the design will work, you'll need to mod- el signal return paths explicitly, something we call power-aware simulation. That will let you take cost optimization to a whole new level, but I don't see a lot of people going there yet. I wish we heard more about this because I think it's fascinating. Shaughnessy: It seems like this could be a real brass ring for managers looking to reduce board costs for high-volume design. Westerhoff: I agree. We'll see where this goes. Shaughnessy: This has been great, Todd. Thanks for speaking with me. Westerhoff: Thank you, Andy. DESIGN007