SANTA CLARA, Calif. The semiconductor manufacturing community must sing the praises as well as air the difficulties of SoCs to make system-on-chip design reach its full potential. That was the main message delivered here at the IP/SOC 2001 conference by keynoter Joe Pumo, director of system-on-chip design technology for Motorola's Architecture System Platform division.
Pumo has been a key figure in Motorola's Digital DNA initiative, pushing that company's semi sector to make design reuse the norm, not the exception, for rapid system-on-chip development. Now, Pumo is sharing his group's experiences with the rest of the design community and encouraging others to do the same to drive SoC innovation, thus design success, thus more money.
It has taken Motorola two years to implement the Digital DNA initiative, according to Pumo. "We've found that SoC is really a mix of cultural issues, technical issues and business issues," he said. "SoC isn't just about Motorola; it's about the future of the semiconductor manufacturing industry."
Consumerization and miniaturization are two forces working in parallel that are impacting SoC design, said Pumo.
"Consumerization gives rise to enormous market opportunity," he said. "People want a lot of little smart gadgets that communicate with each other and have lots of application code and software on them."
Meanwhile, said Pumo, miniaturization allows semiconductor manufacturers to provide systems through higher levels of performance, which allows functionality to migrate to software.
"If you go back to the traditional ASIC model you talk about dollars, about silicon area," said Pumo. "Now semiconductor manufacturers are looking to get more value out of their silicon through software, application code and IP functionality."
This shift, said Pumo, provides greater flexibility and product cycle times. "But the designer now has a more difficult job," Pumo said. "System-on-chip is a product, but it is also process."
Pumo said through Digital DNA Motorola set out to create application-specific cores optimized for specific markets. Further, he said Motorola is combining cores to create core platforms. "These platforms incorporate the maximum amount of reuse to speed time to market," said Pumo, who said Motorola continues to grow its internal IP repository.
Motorola has also undertaken three driver projects implementing SoCs with cores and platforms to test out methodologies and tools and to find design commonalities.
Pumo said wire delay, power management, embedded software, signal integrity, RF effects, hybrid chips, packaging and physical limitations are all issues in SoC design. In terms of tools, the projects showed a greater need for timing convergence tools for mixed digital and analog, data management and verification. "It is not just timing and power verification; at 0.13-micron we have to start focusing on software that can verify the reliability of a chip when it goes to the cust omer," said Pumo.
Pumo also said that tools are also required for voltage selection. "If you have different voltages on your chip, how do you select different voltages for different pieces of functionality so perhaps your chip will function better?"
Motorola is extremely interested in moving up a level of design abstraction to system-level design, Pumo said. "It is new but extremely important. We have to make it more pervasive."
Pumo said the ASIC design flow of system designer to ASIC designer to verification engineer is starting to blur. "What we are seeing is that it isn't somebody writing RTL and then handing it off to someone else for place and route and then someone else for verification. Some of these chips are getting so difficult to timing-converge. Now we have different design people who understand a lot through the whole flow. They understand system design and are starting to understand more about physical design and what it takes to do a chip."
Pumo said that in initiating the D igital DNA message into Motorola, the company had to encourage its design groups to practice design for reuse guidelines set internally by Motorola's management to grow an IP repository.
"When I originally started thinking about IP repositories, I thought it would be a big cultural issue designers don't want to go through the extra effort of taking IP and all its views and placing it in a repository," said Pumo. "The more I dug into it, I found that was an issue, but there were also bigger business issues. Most business managers don't want to put IP in a central repository because they fear they have to support it, and they are worried about their reputation that if someone pulls a core out of a repository and doesn't use it right that their reputation will be in jeopardy."
Pumo said to get over these issues, the company has established an incentive program that issues a range of rewards for designers and groups contributing to the repository.
Pumo said the company also rewards de signers whose practices conform to use of standards endorsed and sometimes created by Motorola's management.