Issue link: https://iconnect007.uberflip.com/i/1505220
24 DESIGN007 MAGAZINE I AUGUST 2023 If you think of a digital twin at the highest level, it's the entire system. A digital twin of an airplane represents the entire airplane, but it also represents all the bits inside, including the electronics, the cabling, and everything else that connects it. A digital twin of a system represents a hier- archy of smaller digital twins. If you want to build a real digital twin, you have to build a digital twin of the environ- ment for that aircra as well. If you're at a design shop laying out a PCB and you may now know what it plugs into—okay, fine. at's your end-prod- uct; that's your system. But if you ultimately are working at, say, a large mil/aero company, then you're designing everything from elec- tronics all the way up to the full body, working with lots of partners who are sharing their vari- ous derivatives of the digital twin with you to help make smarter decisions. So, you can model digital twins at different levels of abstraction. Now, you would prob- ably not want to do a full thermal model of an entire F-35, all the way down to the individual interconnect on those PCBs. At that scale, it's just not possible to mesh that whole thing out, so they would have to do different levels of abstraction, or create sub models. Even in a PCB, if you wanted to simulate every single interconnect on a board with a few thousand parts and maybe 10,000 connections, that still would take a really long time to simulate all that just for signal integrity, let alone the impact of signal integrity on power integrity, or power integrity on thermal. You have to start thinking about the different disciplines and how they interact together to produce a real digital twin of the design. Dan Feinberg: I can see how digital twin could be a powerful tool for designers, but I imagine it could be used well beyond the design phase, for example, on the final product. Absolutely, and the Interna- tional Space Station is the per- fect example. NASA said, "We need to maintain this thing, but we can't bring the space sta- tion down every time we want to look at it." What they would really like is a digital twin, so they can at least do root cause analysis to understand some of the problems remotely. If you create a high-fidelity digital twin of the product, then you've got something that you can query when it's actually out in the market. If a product is out there and it's having yield failures, why is that? Well, let's look at the data on what's happening in the real world as far as percentage failures, then look at what's failing, use that data and apply it to the digital twin, and figure out what's going on. You can also create a digital twin of a pro- duction line, for instance, and you can send feedback from the production line back to the product engineering team. If you combine the production line model with the product model to be built, you can use that digital twin to train folks on the manufacturing line. You can model what a human does on the manufacturing line using the product that they're going to build. You know [what I mean]: I take it from this cart, and I put it in this hardware box, and I make sure that my elbows don't hit anything when I move around. Now, I say this as if we're speaking of a seam- less digital thread of technology, but in real- ity, we have broken digital threads all over the place. e idea is flawed that there's a singular digital twin model that can universally model everything. You have different models for dif- ferent things: IBIS models for signal integrity, BCI ROM models for thermal, VHDL AMS or SPICE behavioral models, CAD models for David Wiens