Complete design management is more than managing the data.
By Ann Steffora Mutschler
Managing the people, the data and the technology are just as important as meeting the market window given that without these, the entire project wouldn’t function. Throw huge data set sizes, different cultures and business management issues into the mix and the challenges are many.
Fortunately, these are issues that the semiconductor industry has been refining for decades while leveraging new technologies, techniques and procedures as they are identified.
Synopsys has a long history of work in these areas, according to Glenn Dukes, vice president of professional services. With a global design organization that is a couple hundred people strong, there are cross-geography people on most engagements. “We started this very intentionally some years ago. The fundamental start to our process was centralizing all data.”
There is no physically distributed design data—all of it resides on shared and very secure design centers, accessed with thin clients to view the data. The secure thin client allows the global engineering teams to view the data, and there are multiple levels of firewalls that limit what the team can do. They can’t copy or read the data, and where they can write it is even controlled.
Frank Schirrmeister, group director for product marketing of the System Development Suite at Cadence, noted there is a difference between chip and board when managing design data. “The board teams, because of the BOM, have to keep track of in the design, getting all the supplies for all the passive components and all the active components of the chips integrated on the board. You need to make sure that those all are product lifecycle managed through PLM tools.”
For the chip-oriented teams, a lot of the work goes on remotely already anyway, he said. For instance, with Cadence’s emulators, there are users with boxes in very odd places—in Brazil or Mexico. It’s highly distributed and assessable from everywhere in the network.
An interesting and recent example of a successful global collaboration was the ARM, Cadence and Samsung 14nm finFET test chip collaboration, said Graham Bell, senior director of marketing at Real Intent, where the team was spread out around the world in Taiwan, Cambridge, Austin, San Jose, Grenoble and Seoul, depending on which part of the design they were involved with. ARM had the IP and other base-level libraries. Physical implementation was done in Seoul, and design work was also done in Taiwan. There was all kinds of front-to-back analysis for the 3D finFET.
And this is just one example. “There’s a lot more collaboration going on and with the growing design set data there is a need for an organized, secure way to have access to the designs,” he said.
Another area growing in importance, particularly for the automotive and aerospace/defense semiconductor industry, is the business management aspect.
The myriad of new subsystems in cars for driver information and infotainment, as well as security, parking assist and other applications require quite a lot of processing power, explained Martin Hall, senior technical specialist at Dassault Systèmes. “This is a way the manufacturers are differentiating the vehicles without changing the entire base platform and just having these additional features in there.”
Considering the various systems within these vehicles, engineering teams also must be aware of AUTOSAR and ISO 26262 [functional safety standard] and track the requirements all the way through the development cycle, he noted.
“You’ve got to look at it from a governance perspective because you need some sort of high-level budget management tool to define the milestones and the phases and the deliverables that you need in order to get the product to market. You have a master project or program for a family of semiconductors. Your projects can be subtypes, so you can have a project that is following the software development process. And you can have templates within the software projects that support the ISO standards, or for aerospace and defense it is D0178B where you need, at certain milestones within the project, to have the software development plan or verification plan in place, and certain requirements have to be tracked and traced in the design and source code,” Hall said.
IP is another design component that must be managed effectively as semiconductor developers have catalogs of IP—not just from the traditional hardware subsystem IP point of view—but also software and high-level modeling that speeds up the architecture and development process.
Dassault currently is working to develop partnerships in the area of IP management around the IP-XACT standard and looking at how its IP catalog can be tied to the work a system integrator would be doing to elaborate the design based on different component views, based on IP-XACT. “This is a system-engineering approach to doing some front-end simulation, but based upon governance around what’s in the IP catalog. Typically the IP-XACT format is a technical view of the IP, but what we’re trying to offer is the governance layer of the IP. The engineer may not be able to use an IP because of contractual reasons or IP export control, so there’s a lot of these kinds of considerations as well as the maturity level of the IP,” he added.
The human element
Last but not least, once the fundamentals are in place to be able to share the data, Synopsys’ Dukes explained, “one of the benefits of that brings actually is you can provide a level of transparency where if everyone is working with their local data and you only see—especially at a management level—what they disclose to you, that doesn’t tend to build trust. The more transparent you can be with each other, where I can really see what you’re doing, I can see your data, you can see mine as appropriate, then that can help to build trust.”
At the same time, it is still important to have very effective program management on top of that, he said. “The kickoff meetings, regular communications meetings, periodic program reviews are very important, and we tend to co-locate the teams for the most critical of those. For well-defined tasks you don’t need to be co-located, but when you are doing something new or creative or level-setting at the beginning of projects or doing the most detailed reviews at the end, co-location can really help improve communication. Even with transparency of data a lot of our communication is done through body language. Strategically co-locating is important, building trust is important, and the infrastructure helps with that but isn’t completely sufficient.”
Finally, just being respectful of different cultures and time zones is very important. “It’s almost second nature today but we’ve been doing it for so long,” Dukes concluded.