Issue link: https://iconnect007.uberflip.com/i/1472851
JULY 2022 I DESIGN007 MAGAZINE 17 the problem was actually with the simulation setup in the other tool; the signal traces were too close to the crop boundary, and that was throwing the results off. Once that was cor- rected, the two solutions correlated well. Interestingly, HyperLynx had set up the simu- lation correctly. e mistake occurred because the other solver just did what the user told it to. e lesson is that you must know how to set up a 3D field solver correctly, or it's "garbage in, garbage out." e big challenge is this: How can we make these powerful simulation tools accessible to system designers? We can't expect system designers to take the time to learn all the ins and outs of setting up a 3D field solver. Shaughnessy: It seems like some of the new field solvers coming out are aimed at this group. Westerhoff: ey claim to be. e challenge here is accessibility, not technology. e issue with 3D field solvers is that their development has always been driven by designs at the state of the art. e expert modeling the intercon- nect for 224G serial link design has very dif- ferent needs than the system designer that just wants to validate a PCIe-5 board that was laid out using rules that sup- posedly guar- anteed it would work. SI experts at the state-of-the- art care about per- formance, capacity, and accuracy first and foremost. It doesn't matter if the solver is hard to use; they can make it work no matter what. e system designer with PCIe-5 is literally the inverse: performance, capacity, and accuracy have to be good enough to produce a meaningful result, but the driving issues are integration, workflow, and usabil- ity. System designers need tools that they can deploy without having to become SI experts. If you walk into the cockpit of a Boeing 737, your first impression is probably, "is is really complicated. I could never fly this!" e fact is, that plane has an autopilot, and with mini- mal training and common sense, you could fly that plane. Most field solvers are still like flying that 737 without the autopilot, so they require a SI expert. We've integrated the HyperLynx solv- ers with our SI tools and put a layer of auto- mation around them so that process of identi- fying the area to be modeled, setting the crop boundary, defining signal ports, and setting up solver options is done for you. It's like using a 3D field solver on autopilot. e challenge has been taking this sophisti- cated technology and developing algorithms that will let system designers generate mean- ingful results, then ensuring that automation works for a variety of real-world designs. We want to bring 3D field solvers to mainstream design by changing 3D analysis from an art practiced by a few skilled crasmen to some- thing more like 3D EM for production use. Shaughnessy: Are you trad- i n g o f f s o m e l e v e l o f accuracy for accessi- bility? W e s t e r h o f f : W i t h some of the automated flows, yes. If you really need that last few percent, you're probably an SI expert already and will turn off the autopilot to fly the plane manually, so to speak. Remember, the goal here is to enable a larger audience to run analysis themselves, reducing the demands on those overloaded experts. ere are a lot of designs to get out the door, only a few of which require ultimate capacity and accuracy. at last few percent comes at an incredible cost in tool complex- ity and user expertise; we want to reserve SI experts for those cases where they are really needed.