Setting Standards For The Chip Industry

Industry veterans examine the inner workings of how standards are created.


For all the advances in semiconductor design, and the astonishing scales on which the industry now works, when it comes to standards committee meetings, not much has changed. Advice from a 91-year-old retired engineer can sound surprisingly like advice from those active today. Standards were then, and continue to be, a mix of technical compromises and corporate politics, as well as passionate arguments over engineering details.

Typically, the only thing everyone agrees on in a standards group is that standards are proliferating.

“I call it the standardization jungle,” said Jyotika Athavale, director and principal architect at Synopsys, and IEEE Computer Society 2023 president-elect. “There are so many new standards emerging, it’s hard to keep up and understand the whole mind map of how they all connect together.”

Everyone agrees that the massive jungle should be tamed, although it’s more a dream than a roadmap. Speaking of her specialty, Athavale said that there currently appear to be no efforts to consolidate the safety standards. However, the IEEE CS Functional Safety Standards Committee did publish the mind map of the various standards within the functional safety community.

“Most successful standards groups wind up in a streamline/consolidation phase eventually, mainly around proposed extensions,” said Rob Aitken, distinguished architect at Synopsys. “It’s rare that you get a VHS/Beta case where standards overlap completely. Instead you get something like Gen-Z/CCIX/CXL situation where the standards covered similar but not identical space and CXL eventually absorbed Gen-Z while CCIX sort of faded away – or the way that IEEE 1809-2009 effectively merged Verilog and SystemVerilog. Generally standards consolidate when there is enough pressure, usually from users but sometimes from suppliers, to force action.”

Retired engineer Guy Kelly helped establish the IEEE FORTH language standard in the mid-1980s, as well as being part of the ISO/IEC FDIS 14443-2 noncontact smart card standards meetings in the early 1990s, which had a less happy outcome. Despite taking place nearly 35 years ago, Kelly’s experience sounds so familiar it could have happened yesterday.

“You’re there to set a standard for existing technology,” said Kelly. “Ours was existing, but wasn’t compatible with the other folks. We went into it thinking, ‘We’ve got this excellent technology we can contribute,’ but the other participants’ systems weren’t up to it. One of the things that you have to worry about when you’re sending data back and forth to a non-contact device, like a smart card, is you put it in the EM field, it powers up from the field, and then starts to communicate via the field. If that process takes longer than a few milliseconds, you can pull it out too soon, which is called ‘tearing’ because you’ve torn the communications. Most people had to write software on their end to handle a torn transaction, trying to get you to repeat it. We did some interesting buffering, and we had a higher communications rate, so the card itself could notice that it tore, and cancel or fix the transaction based on buffer storage.”

Despite that innovation and his position as technical support to the committee chair, who also happened to be his boss at the company, their design did not persevere. While this may have been frustrating from a corporate point of view, it was exactly what should have happened.

“Officers have to have a neutral position,” said Athavale. “Even if you’re representing a company, you’re going to have to be neutral, and just be a technical person in your role as officer.”

Thus, Kelly and his boss had to sit by while their newer, more efficient solution was voted down by the incumbents. “One of the driving factors [from our competitors] was, ‘We’ve got most of the market so it’s going to be us,” Kelly recalled. “You built hundreds of thousands of cards. We’ve built millions of cards.’”

While that might seem like an obvious case of big guys throwing their weight around, it also illustrates some of the human psychology underlying standards adoption. Several sources warned that first-time participants, especially if they’ve just come out of an academic environment, should not be naïve about what drives standards debates. In corporate life, with its tight budgets and even tighter deadlines, human factors often trump engineering considerations, and often what wins is the solution that will require participants to do the least amount of additional work. “Time to market is less if the committee picks what you just did,” said one veteran.

Having lost at the meeting, Kelly and his employer decided on their own workaround. “The two big players each got what they wanted, so the resulting standard had two parts, A from one, and B from the other. We were hoping to get C in there. But nobody else was really interested, because they were locked into the earlier design. Most people built what they called a ‘dual reader’ to handle parts A and B. We built what we called a ‘tri-reader’ to handle parts A, B, and C.”

To this day, the company, Cubic, still makes cutting-edge smartcards. And, standards decisions still can result in “augmented,” alternative designs.

“That’s a classic maneuver,” said Aitken, upon hearing the story. “It can be really frustrating to have a standard that limits your technology. It effectively becomes a market play. If the market is interested in what you’re doing, then it’ll say, ‘We’ll buy this stuff because it does what the standard does and then some.”

Of course, it doesn’t always go that way. “The market could say, ‘We just really want things to work, so please don’t change anything ever,'” he noted. “Then you have to go back to the official standard.”

Nevertheless, the tactic of agreeing to a standard and then going beyond it is still employed today. In fact, it underpins a basic approach to standards work.

“As long as the definition of the standard is at an abstraction level that allows people to differentiate their solution based on that standard, it is okay,” said Aparna Dey, senior product marketing group director for strategic alliances at Cadence. “That’s how we work together to develop a standard infrastructure, for example, a standard API and data model allows people to differentiate, interoperate, and build on top of it. If somebody is proposing a specific implementation, the committee can ask them if we can go to a higher abstraction level which allows people to make their own differentiation or to contribute it as a reference implementation. For example, Cadence contributed and continues to contribute to the Si2 OpenAccess API Standard and the reference implementation database for IC design. The reference implementation of the database is not the standard; the API is the standard, but the reference implementation, because it is stable, has wide acceptance in the industry, and it allows companies to focus on their applications rather than the underlying infrastructure, which has been provided.”

NDAs and anti-trust
Engineers attending their first standards meeting may feel like they’ve stepped into a different world. “It’s not like your normal job,” said Rich Weber, fellow and director of engineering at Arteris. “It’s a lot more like politics. The IP-XACT committee actually met in the same room in Dublin, Ireland, where the Irish wrote their constitution. So it really is like international politics.”

In fact, during the historic discussions to create a unified UVM, OVM and VMM standard, which involved all the major EDA players, things got to the point where votes had to be secret because people were concerned about their careers being ruined if they voted the wrong way.

Frank Schirrmeister, vice president of solutions and business development at Arteris, said a standards committee should be staffed with “Alte Hasen,” a German expression that literally translates to “old rabbits.” Because rabbits were hunted for food, its deeper meaning is that those who were smart and savvy were able to survive — in other words, industry veterans. “They have the experience and the gray hair to prove it. They will be able to tell you just by looking at things very early whether you could run into a potential deadlock somewhere.”

Industry veterans know the engineering as well as the legal guardrails, which are the guidelines that junior staffers need to learn before participating in a standards meeting. The two major legal pitfalls are first, the obvious challenge of sharing proprietary information in a standard setting, and second, the potential to run afoul of anti-trust laws.

“The architects are already aware of what is proprietary and what is not proprietary, but we make sure that they are aware of the rules, and what we are ready to contribute to the community,” Cadence’s Dey said.  “Standard committee members are usually senior level engineers because they have the expertise to contribute and develop specifications, especially trying to solve a complex problem.”

As Synopsys’ Athavale explained, the IEEE-Standards Association requires a call for patents at the beginning of every meeting. “Anyone can speak up and respond to that call, or they can inform the chair offline. We encourage early identification of patent disclosures. It is similar for copyright. If you share content in the standards development meeting, then the SDO is automatically given a license to that information. So you have to be careful not to disclose anything that’s confidential from your company, because if you upload something or you present something, then you’re giving IEEE-SA or ISO or whichever committee you’re on license to use that as theirs. Those patent and copyright rules are read at the beginning of every meeting.”

In addition, to keep everyone on track with policies and procedures, every working group includes a staff member from the sponsoring SDO.

As for the second issue of possibly running afoul of anti-trust laws, unfortunately, a standards meeting can appear perilously close to the legal definition of price-fixing, which is why participants can never flat out say, “If we go with your idea, the device will be more expensive.”

“We just focus on the technical topics and don’t talk about anything else with respect to products or pricing or marketing or sales or any other aspect,” said Athavale.

Yet there is one market-related aspect that entices many companies to participate. It’s not just vendors and their competitors who join in, but their own customers.

“Customers share information in these bodies about their requirements and what they would like to see in a standard,” said Dey. “It helps us figure out what we need to do in our products to be compliant.”

Nevertheless, there’s a fundamental contradiction at the heart of standards meetings. “Normally, it’s not the goal for competing companies to work together,” said Weber. “How do you handle that?”

Industry sources note that every in-person standards meeting is essentially two meetings. One takes place in the hotel conference room and is on the record. The other takes place in the hotel bar and is off the record. The latter is often more productive, because it’s not only where the most lobbying for votes is done, but where people can be frank about what’s really driving their objections. In other words, it’s as important to understand people’s needs as it is to understand technical issues. Ultimately, every standards meeting has two problems that need solutions — the technical ones, which are discussed in the conference room, and the human ones, which are hashed out in the bar.

“You want to have the balance favoring engineering because that’s where you get a good standard, but the reality is that the politics aspect of it is important,” said Aitken. “If it’s going to succeed on its own, and not just be some pristine document that everyone agrees is wonderful and no one follows, it’s going to have to be adopted by human beings. They will be motivated by all the things that human beings are motivated by. A lot of times engineers call that politics, but it’s really just human relations.”

Advice from the trenches
Athavale welcomes and encourages participation, especially across diverse groups. She and many others suggest starting out as an observer before jumping in, or participating in a smaller study group before joining a full working group. The IEEE Standards Association has free training classes available online, which cover the entire lifecycle of standards development, as well as a handbook covering policies and procedures for designated representatives.

Nevertheless, there are still interpersonal dynamics that may need finessing.

“If you’re trying to make an argument, it’s important to bring the necessary supporting collateral to that argument. It helps get the point across,” said Jake Wiltgen, functional safety and autonomous solutions manager at Siemens EDA. “People are different types of learners. For example, I’m a visual learner. I like to see block diagrams and supporting data. A lot of people like that.”

Wiltgen also recommends getting as much experience giving technical presentations as possible, some of which may come from academic experience, as well as industry experience. “These are very technical folks, and they will often challenge. That’s a good thing, but you need to be prepared to answer when those questions come up.”

Aitken added that the reverse is also true. Standards committee participants shouldn’t be afraid to challenge even those who are their seniors. “In some cultures, that’s a massive faux pas. You can’t tell a person who’s senior to you that they’re wrong. But you should still say, ‘Hey, no, this is what’s going on,'” he recommended. “It’s important to be diplomatic, and to avoid damaging the relationship, so you might not use those exact words, but the message still needs to be conveyed.”

Athavale also encourages everyone to be confident — especially women and others who may feel underrepresented. “I’d like to see more women in standards and standardization. More than the diversity, it’s the inclusion part. People need to feel included. People need to feel like they should be able to speak up, that they should be able to participate. They should have a voice,” she said. “For people participating from countries where English is not their first language, it’s a little bit of a struggle. But from what I’ve seen, people are very encouraging, very supportive, and understanding, and willing to help build their confidence to speak up.”

Kelly, the most alte of the Alte Hasen, gives these final words of encouragement: “The best advice is to listen carefully, be friendly, don’t diss anybody or their technology. Be prepared to defend yours in terms of the benefits that it has and why they might want to include it in the standard.”

Related Reading
New Standards Push Co-Packaged Optics
Speed, density, distance, and heat all need to be considered; pluggables still have a future.
Standards: The Next Step For Silicon Photonics
More data and denser designs are opening the door for photonics.


Cliff Greenberg says:

It is interesting to note that Ms. Athavale’s concern about safety standards neglects to mention the “grandfather” of semiconductor equipment Safety Standards, S2 from the SEMI organization. This document was first released in 1991 and since then has undergone continuous review and improvement with the participation of safety people from all aspects of the industry: equipment designers and suppliers, end users, third party safety specialists. I’ve been participating in this avocation for over 30 years and we have a common goal: to reduce hazards and control possible bad outcomes. Standards membership is free and is international as well: differences in approaches in various nations can be brought in for discussion and assessed, used for reference so design engineers can have a document that reflects best practices worldwide. Before my retirement I sat at the table with my competitors’ safety people and our suppliers and customers with a common goal. We achieved this result to such an extent that NIOSH contacted SEMI to find out how the industry had the lowest incidence rate of recordables compared to comparable industry sizes.
There are ongoing issues since this topic is less glamorous than some other specialties and, while we have several women involved, the percentage is still lower than I would like to see, but we keep soliciting new people to help this effort.

Art Scott says:

Standard to measure and improve Recommendation ITU-T L. 1318 outlines a method and fundamental metric for expressing integrated circuit energy efficiency, the Q factor. The Q factor could be applied to measure and improve the integrated circuit technology behind information and communication technology itself.
Tufts U “HotGuage” advanced hotspots —

Leave a Reply

(Note: This name will be displayed publicly)