DesignCon: SI, PI and EMI Have a Threesome

DesignCon 2018 opened with a keynote panel on the subject of SI, PI, and EMI Challenges Looking Ahead Through 2023. The initials are for signal integrity, power integrity, and electro-magnetic interference.

Istvan Novak pointed out Amara's Law, that we tend to overestimate the effect of technology in the short term and underestimate it in the long term. His guess was that 2023 was perhaps where the lines intercepted five years out, so the panel had a chance of neither overestimating, or underestimating, but getting it right.

The moderator was Steve Sandler of Picotest. His panel was:

• Eric Bogatin of University of Colorado
• Istvan Novak of Oracle
• Ken Wyatt of Wyatt Technical Services
• Al Neves of Wild River Technology

Istvan was actually on a panel session I moderated at CDNLive a couple of years ago, which I covered in Signal Integrity Masterclass.

Each of the panelists took a different look into their crystal ball. However, there was a lot of commonality. If I had to summarize, I'd pick two things:

• SI, PI, and EMI are growing together into a single discipline since you can't fix problems in one of the three without breaking the other two
• This is a huge educational challenge since most people are only experts in one of these domains, some companies even have completely different departments

Eric Bogatin

Eric went first. He put up the Gartner hype cycle as an example of why predicting the future is hard. In the short term, things get hyped way beyond anything they can deliver, and then they crash into the trough of despondency. Eric pointed out that it is then that the magic happens, when some products die, and others climb out to become real businesses. The difference being whether there is a killer app or whether it is just technology in search of a problem.

With that, Eric talked about six aspects of the future where he thinks big disruptive changes are coming, future killer apps:

• Computation technology: Not standard von Neumann processors. The first disruptive technology is quantum computing (Rose's Law, the number of qubits doubles every two years, seems to be in effect but is still on the flat part of the curve). The second is biological computing, doing things that you can't do with pure digital logic.
• Visual interfaces: First, virtual reality and augmented reality...but then connecting electronics directly to the optic nerve in a simlar way to cochlear implants that give the deaf hearing.
• Personal digital assistants: Not your grandfather's Palm Pilot, but a combination of AI, wireless, and human interface.
• Man-machine interfaces: Not keyboards or even voice, but direct brain interfaces. Elon Musk has a company and thinks we are four or five years away from this, perfect for the time frame of this panel.
• Fusion: IoT on steroids with sensors, mobility, dexterity, and connectivity.

• The maker revolution: In Eric's youth, it was Heathkit and Radio Shack, and when they cut back totally, he wondered where the next generation of engineers would come from. But now there is a whole new revolution with things like Arduino and Raspberry Pi. Just a few dollars, open source software, and hardware, powered by the enthusiast, not the professional. Take a look at Backyard Brains, for example, neuroscience for everyone.

Istvan Novak

Next up was Istvan, who Eric introduced as the expert in PI and SI that everyone in the industry goes to. He went back to a keynote from DesignCon in 1995 (he couldn't remember who gave it) where someone predicted that "the computers of the future will look like hairy, steamy, golfballs". Meaning that there would be thousands of connections, they would run hot, and they would be small. Istvan pointed out that one of those was correct. With high-speed serial interconnect, computers are not hairy. A smartphone is bigger than a golfball, but because of the screen. However, they do sometimes go up in flames, so the steaming bit is partially true.

Three specific predictions looking forward:

• Istvan remembered a keynote that he'd given sponsored by a capacitor company that were salivating at how many caps people would need to purchase from them. But the answer may be none if we get better at matching impedance between packaged silicon and the DC source (well, probably not none, but just a few). His prediction is that the number of bypass capacitors will decrease.
• Integrated voltage regulators have been in the making for a long time and are finally starting to appear on some of the largest chips. His prediction is that the next few years will see further applications.
• Allowed noise voltage bands continue to shrink, and advanced components sometimes have loose tolerance. This means that designing for worst case leaves a lot on the table. His prediction: individually characterized systems may appear.

Ken Wyatt

Ken made the strongest prediction that SI, PI, and EMI will merge since we can't really partition them into three anymore, they all interact with each other. As we get up to 50GHz we will need to consider everything starting from the physics. 

Another big change is that the whole world is becoming more interconnected with time, resulting in very complex electro-magnetic environments especially in hospitals, factories, and cars. Even in the home where Bluetooth is smart and can co-exist with WiFi in the same spectrum (by dynamically avoiding hopping into WiFi channels that are in use). But then WiFi gets blown away by leaky emissions from a microwave oven.

Another big challenge going forward is self (or platform) interference. The DC to DC converters desensitize receivers for mobile and GPS, for example. Radiated noise couples to the receiver and can lower the sensitivity by a product killing 10-20dB. Autonomous vehicles are perhaps the most hostile environment of all, with 30+ processors, dozens of controls, collision radar, two-way radios, not to mention all the old stuff like airbags. They all need lots of Si, PI, and EMI analysis.

The challenge is that most product designers don’t understand the basic concept of how energy moves in transmission lines, such as the need to have defined path for return current. Excellent digital designers can’t produce a product that meets environmental performance and regulatory performance. That was a perfect segue as to why he wrote a book: EMI Troubleshooting Cookbook for Product Designers. This is a big challenge: how to connect better with product designers in the trenches, plus connect with colleges to bring real-world examples to upper-level students.  

Al Neves

Al said that he doesn't know much about brain-computer interfaces, and he certainly doesn't need one that tells him when to eat or drink a beer. But he does know something about signal integrity. He reduced his predictions to a single number: 2Ω. For 40GHz and up, 2Ω is the future of signal integrity. If you want a 20dB return loss margin, you need around 2Ω TDR variations from 50Ω. Signal integrity is really hard, and the future is simplification.

The first problem is that all field solvers today have issues, and some have really serious issues. Plus, people don't know how to use them and don't identify the materials and the cross-sections properly. Management doesn't get it either.

Like Ken, Al thinks we need to go back to basics and study magnetics again to understand the flow of energy in our systems. At the highest frequencies, there is energy leaving the system and the S-parameters we use don't work for leaky systems, so that's another area for improvement.

Steve's Question

Steve had one question, which was "What do engineers need to be doing now to prepare themselves for the new era?" The new era being when SI, PI, and EMI all overlap. He pointed out that at a workshop he ran on power integrity, only ten percent of the attendees were power engineers. We are all starting to have problems with disciplines that we don't understand. How do we think like a clock?

Eric: Eric thought merging SI/PI/EMC is sometimes a challenge because that's how companies are organized. However, practitioners need to get experience in the holistic approach and most are trained in just one area. "DesignCon is a great place to learn (Yay! We're in the right place) about the areas you don’t know. Cross-fertilization is really important."

Istvan: Also thought education was key. "We all have to learn outside our disciplines, and in addition to attending DesignCon, each company needs to do better integrating these things. 10 years ago SI, PI, and EMI reviews were done differently and separately, and then what one review suggested was undone by the others. This needs to be integrated and, eventually, it will happen."

Ken: "I agree about education. We each have a unique skillset in SI or PI. After 30 years of delving into EMC world, I am reaching out to learn more about PI and SI. Ralph Morrisson (turning 93) just published his 15th book on grounding and shielding, and we all need to study how electric fields move in circuit boards to understand how to minimize SI and EMI. If we each start looking at the other fields, that is key. Need to get more designers in the tranches to attend DesignCon and other training like this."

Al: Al agreed at one level that education is important, but not from the educators. "They teach the basics, the math. But they don't teach the art of engineering because the professors are not practitioners, they don't build things and take measurements. If I want to learn something, I buy all the books, go to DesignCon, hang out with people who are experts. That's how I learned business when I realized being a good engineer was not enough."

Ken agreed that professors are not practitioners, and that turned out to be the end of the discussion. Steve thanked all the panelists and it was time for lunch (if you had a full event ticket, which I didn't. No soup for me).

Sign up for Sunday Brunch, the weekly Breakfast Bytes email.