Issue link: https://iconnect007.uberflip.com/i/1219242
MARCH 2020 I DESIGN007 MAGAZINE 27 it takes extra work, I don't have time, and I'm not going to do it." Shaughnessy: We run into designers who con- fuse DRC with DFM; they run a DRC and think of it as DFM. Westerhoff: That's understandable, as both are essentially rules-based automated checks. DRC tends to focus on design functionality; if you don't change that, the design won't func- tion. DFM tends to be more about the ability to manufacture in volume, but there's certainly some overlap. I think another area of confusion comes af- ter DRC identifies a potential issue. What then? DRC doesn't always find hard errors. Some- times, it flags things that need further con- sideration; they might be a problem, or they might not. Those issues need review by a de- signer, and we find they tend to fall into one of three categories: obvious problems, with no need to analyze in more detail that need to be fixed; obvious non-problems, with no need to analyze in more detail that can be marked as such; and unclear problems where models and simulations are required to determine what needs to be done. The good news here is that the first and sec- ond outcomes take the issue off the table with- out placing any burden on those scarce SI re- sources we talked about, and they improve the quality of the design as well. The third out- come is the most interesting; most of the time, the only option is to pass the problem to ex- perts for analysis and wait. Of course, if the ex- perts are too busy, you make your best guess and hope for the best. But what happens if you don't need expert- level simulation to resolve the issue? We have been working with our customers to put simu- lation directly into designers' hands in a sim- plified way. We let designers see simulation re- sults daily while the board is being routed with the goal of being able to find problems and cor- rect them while layout is in process. We want to avoid finding problems only after layout is com- plete because that's when changes to the design are the most difficult and time-consuming. But how? There aren't enough experts to go around. The trick is that we don't need to do the same level of accuracy, signoff quality sim- ulation that an expert would normally do. We can do something far simpler because we're looking for things that are obviously out of whack (excessive ringing, excessive crosstalk, bus delays don't match). We call this first-or- der analysis. We model interconnect accurate- ly using 3D EM solvers, but we don't do the same level of model setup and post-processing as signoff analysis because we're looking for outliers. Shaughnessy: Designers typically haven't want- ed to simulate in the past. Why would they want to start now? Westerhoff: The traditional approach has been to go get more experts and more sophisticated tools, but we have reached a breaking point. The percentage of SI/PI experts in the design community continues to shrink, and design complexity continues to increase. Many de- signers have avoided picking up SI tools be- cause the learning curve was too steep, and it's only getting steeper. We have reached a point where we think we need an expert to do anything, but they're too busy, so now what? Here's a weird question: How many homeown- ers people can patch a wall or replace a light switch anymore? Shaughnessy: I can, but I have to. I'm a land- lord. Westerhoff: How about an outdoor faucet? Last winter, I forgot to drain one of my outside fau- cets. When it got down to 2°F, the water in- side the head expanded and cracked it. It was so frozen that it didn't leak immediately, but as the weather warmed up, it created a nice, frozen waterfall on the backside of my house. Here's the point: The faucet itself was worth $10, but it cost me $250 to have a plumber fix it because I didn't have the tools or skills to do that kind of repair myself. I think SI is in a similar situation. Design- ers are bottlenecked waiting for expert opin-