Issue link: https://iconnect007.uberflip.com/i/1472851
12 DESIGN007 MAGAZINE I JULY 2022 eral-case conditions. 2D solvers have become mature enough that designers performing sim- ulations may be using them without realizing it; they're embedded parts of simulation pro- cesses in some tools. Todd Westerhoff: Let's take a step back. Why do we simulate our designs? Why do we simulate anything? Hargin: So that we can predict the electrical behavior of our designs—the performance of our designs in advance. Westerhoff: Precisely. We use simulation to predict whether something will work before we build it, so that we can debug and opti- mize our designs before we commit them to prototype fab. ere's really only one ques- tion, or two, that we ever care about: Will it pass or fail against requirements, and by how much? e "by how much" part is operating margin and it's critically important. No simu- lation is ever 100% accurate; no matter how careful we are, we can't model and predict everything. Every simulation and every simu- lation result has a margin of error associated with it. When we simulate, we're looking to prove that the operating margin at the system level is substantially more than the margin of error in the simulation we performed. at's when we can say the design will work with room to spare. When we talk about field solvers, we need to remember it's the margin at the system level that we care about. We need to model the structure in question (a transmission line, a package breakout, a differential via, etc.) as accurately as required to ensure adequate mar- gin at the system level. But we don't want to model things more accurately than required, because that takes additional time and effort that isn't really moving the design forward; it's just analytical overkill. ere's a huge temptation to model every- thing as accurately as possible because we're engineers and we want to be precise—the hard part is knowing when that extra accuracy is required and spending our time judiciously. at's a hard call when we're dealing with field solvers. Eric Bogatin says the level of accuracy you need in your interconnect model goes up as your system operating margin goes down. I think that's a great way to frame it. Hargin: At faster speeds, s ystem margins decrease and the number things you need to worry about increases. You need to take a lot more care and consider the details in your design, because you need to precisely model the design as it will be built, not as you wish it could be built. at's the focus of my company, Z-zero, and our product, Z-planner Enter- prise. We allow designers to model their PCB stackups more precisely, instead of just design- ing to an approximate stackup and hoping the new product introduction (NPI) engineer can work with a fab house to create a stackup that matches whatever we used for design simula- tions. ere's really not much point in running detailed design simulations if you don't have an accurate description of the board stackup and material properties. Bill Hargin