Don’t Rush: Get ‘Acclimated’ With Each Level of SI

Estimated reading time: 8 minutes

During DesignCon, I met with Al Neves, the founder of Wild River Technology, and a serious fly fisherman as well. Wild River has developed some of the most cutting-edge signal integrity measurement and optimizations test fixtures around, and Al has authored and co-authored a variety of award-winning papers presented at DesignCon over the years.

Al is not shy about saying what’s on his mind, so I asked him to discuss what he’s seeing in signal integrity design and measurement these days. As Al explains, some engineers are getting ahead of themselves by rushing to take on complex SI challenges before they’ve mastered their foundational knowledge. Like climbers on Mount Everest, these engineers need to spend more time getting “acclimated” at base camp before heading for the summit.

Shaughnessy: Al, how’s it going? It looks like you’re having trouble with your phone.

Neves: Yes. Apple guys, if you're listening, increase your phone's battery so that by the end of a conference day, the battery is still working. Thanks.

Shaughnessy: Probably a signal integrity issue. Speaking of which, how’s business, and what are the big drivers in your segment now?

Neves: The company's overwhelmed with business right now. The whole NVIDIA AI push is superheating the industry. What people don't realize is that you just can't compartmentalize technology, so everything is going to be getting faster. We have three groups of businesses that are really interesting. First, we have the folks going from 10 to 32 Gb/sec who couldn't meet 32 Gb/sec because they don't have a signal and engineering design team. We help those customers; their projects are interesting and still pretty challenging. And then we're doing a lot with the 112Gb people, and it looks like this year we'll be doing a lot of work at 224 Gb PAM4, the really high-speed stuff requiring extreme signal integrity.

Shaughnessy: It sounds like your customers, and probably some other SI engineers, are getting ahead of themselves. They’re moving on to higher speeds without addressing the foundations first.

Neves: Exactly. Imagine climbing Mount Everest. You spend a year or two getting into super good shape. You were already in decent shape two years before that. And then when you get to base camp, where you stay for a little bit to acclimate. There's an acclimation period—a base camp—in achieving a level of technology. You need to stay there for a while to clean some things up, then learn the whole concept and get mastery. For example if you created some test vehicles to go from 32 gig NRZ to 56 gig PAM4 to 112 gig PAM4, you’ll want to throw enough time in there so that you plateau. Why? Because sometimes you need to clean things up. Sometimes, there are surprises. The real insight into some of these problems comes when the engineers aren't so rushed; they need to allocate more time to get acclimated at that level and spend a lot more time on material identification for their EDA tools, for example. Feed your field solvers and feed your EDA tools!

At the paper we gave the other day, Kristoffer Skytte of Cadence asked how many people in the audience use anisotropic Dk. There were only 10 or 15 people using anisotropic X and Y decay. What we're saying is this: They aren’t doing enough foundational stuff. Now, what we espouse is making use of a material ID platform, where you create a platform and you extract a full 3-Dimensional material ID model. That's not that easy to do. Your base camp may be your material ID platform. It's a small platform that gets you X, Y, and Z axis decay. That's the fundamental stuff that you need to feed field solvers. And now you build a much more comprehensive test vehicle. At this point of “acclimation” you have a complete library that you can also use for serial link analysis to determine your link margins. you’ve created a foundation that your engineers can feed off of for doing really high-end level work that's suited for 112 and 224 gigs. And at the end of that, you have really good signal integrity. You have climbed the mountain and now looking down, experiencing deep success.

Shaughnessy: Some engineers think, “Hey, we can fix it in the lab.”

Neves: If your engineer says “We're going to run a series of experiments” when his back is against the wall and he needs to do 112 or 224 gigs, but he's never done serial link design, that's really bad. We see a lot of our customers struggling at the beginning, and if it’s done wrong, it's a trajectory that's really bad for the project. It is also bad for the engineer, the manager, the senior manager…

Shaughnessy: Get acclimated at 112 gigs before you move up to 224, or whatever realm you’re playing in, right?

Neves: Yes, and then you need to evolve your technology. A friend of mine said, “If you think it's going to take you six months to do 56 gig, then you should have started it two years ago, because it's going to take you two years.” He went on to say, “You're going to do it, and you're going to fail. And then you're going to do it and you think you're a success. And then you’ll find out you still failed.”  The mountain top is often elusive, you can often see it, but it is so hard to get there. I guess you can think of my team as a bunch of Sherpas; we help you get to the top!

Shaughnessy: But you're learning along the way. It just takes a while.

Neves: It takes a lot of investment in time and effort. And you can't cut corners for foundational extreme signal integrity. The retainment of engineering staff is always a problem in these companies. I'm of the belief that engineers leave because they're not appreciated and they work in an environment that is not methodical and systematic that serves up with a lot of surprises. A lack of systematic approach fosters stress, real stress… the kind that injures the core of your soul.

For example, let’s say you have something fabricated. It gets built, and it comes out on Friday afternoon. It doesn't work; there are all kinds of problems. With no test vehicles, there's no way to test it. You have to go into the lab and probe it to figure this stuff out. The pressure's enormous. And the probability of failure is 100%. Is that engineer going to have a nice weekend? You think he's going to enjoy his beautiful kids and his wife, wash his car and take his wife out to dinner and be happy? No. He's going to be unhappy. He's going to tell his wife, “I'm thinking about changing my job.” And she says, “Hey, you've been looking, and you've been unhappy a long time, do it!” 

Now contrast this with the engineer who designs something and it works. There still could be some problems, but he says, “OK, we've got this under control. I know what to look at. I have test vehicles that are going to help me. I’ve got a lot of work to do next week, but I have a lot of confidence we're going to execute.” That's why engineers leave; they get burned out because their projects aren't managed correctly. And, sometimes, the engineers are partially responsible, but often it is how the company is managed, and failure is embedded deep into the culture.

Shaughnessy: Wow. Can I steer our talk to a lighter note? Any interesting trends in the air at DesignCon?

Neves: Absolutely. Another thing that we're seeing at DesignCon is that crosstalk is insidious. Our old models of crosstalk don't fit anymore. Yuriy Shlepnev presented a really good talk on crosstalk. We used to look at crosstalk as just a coupling from something that's too close, but that's no longer the case. Now what you have is a lack of localization; the electromagnetic energy “sprays out” and victims nets pick that energy up. Now, everything's an antenna, and everything's high frequency. And then things are not well behaved any longer. The energy then goes crazy; the fields are not orthogonal to each other, and the field densities look squirrely. What happens is you have insidious margin eaters with wideband crosstalk problems in these systems. But you didn't think that would happen.

Crosstalk is going to be a much bigger deal as not only speeds increase, but we’re going to PAM6. Ransom Stephens, an author and physicist, wrote a paper a few years ago explaining that “crosstalk is back.” We need a practical way of looking at crosstalk and coupling and understanding the electromagnetic propagation of energy. I think learning your foundational elements, understanding crosstalk, and having a systematic approach is how you get your head right. Go to base camp, Camp 1, Camp 2, and get acclimated, then climb to the top when you’re ready.

Shaughnessy: I like that. “Get acclimated” is great advice. Thanks for talking with us, Al.

Neves: Thanks, Andy. Always great to catch up. And it was great seeing you and Kelly Dack playing guitar and singing in the lobby last night.

Shaughnessy: Thanks, Al. They didn’t chase us out.