NoC Silicon IP for RISC-V based chips supporting the TileLink protocol
Limits of IP block strategy exposed
Limits of IP block strategy exposed
Richard Goering
EE Times
Mar 14, 2005 (9:00 AM)
URL: http://www.planetanalog.com/showArticle?articleID=159903637
Santa Cruz, Calif. The lesson of Barcelona Design Inc., the analog automation pioneer now disbanding, is that good technology and plenty of venture money won't save a company with an unclear business model and overly ambitious goals.
But at a deeper level, Barcelona's failure also calls into question the idea of building a company around analog intellectual property (IP). Barcelona's vision was one in which system-on-chip designers would buy or synthesize large analog IP blocks and plug them into complex mixed-signal designs as easily as they might put analog devices onto a pc board.
"It's a sad tale of how great technology can go wrong," said Joseph Costello, chairman of Barcelona and the former chief executive officer of Cadence Design Systems Inc. Costello recently confirmed that Barcelona, which had garnered about $44 million in venture capital funding, is winding down its operations and seeking a buyer for its technology.
In the wake of the Barcelona experiment, observers said that analog IP must be offered within the context of a broader solution that includes tools, services and lots of support. Most analog IP development remains in-house at large companies, and a lot of analog functionality is still going off-chip.
"I think if someone figures out a way to deliver soft [analog] IP that is easily customizable and can be ported quickly, there would be a market for those soft-IP blocks," said co-founder Mar Hershenson, who left Barcelona in 2003 for "work and personal" reasons and is now involved with an unidentified startup. Hershenson, who also teaches part-time at Stanford University, said she was unsure why Barcelona failed. But she noted that hard IP is difficult to reuse because "the specs change and processes change so often."
"The problem at Barcelona is that they often looked more like a business school experiment rather than a company," said Gary Smith, chief EDA analyst at Gartner Dataquest. Nevertheless, in Smith's view, Hershenson "still stands out as one of the best minds in analog automation." As a Stanford graduate student, she did groundbreaking work in nonlinear convex optimization and its impact on automating analog design.
"Barcelona had the potential to be the catalyst in developing a true RT [register-transfer] flow for analog design," Smith said. "The analog EDA market is still struggling, and there is no one around to match Mar and her vision."
Founded in 1999 by Hershenson and Stanford professor Stephen Boyd, the Sunnyvale, Calif., company claimed breakthrough analog synthesis technology and pioneered a Web-based pay-per-use model. Customers were to license Barcelona's IP along with its synthesis tool, then use the tool to optimize circuits for a specific process.
Some big names in electronic design automation saw good potential there. Besides Costello, Barcelona claimed backing from Robert Dobkin, Linear Technology's CTO; Abbas El Gamal, founder of Actel and Silicon Architects; and Buno Pati, former CEO at Numerical Technologies.
But even "wonderful new technologies can screw up, even with all kinds of luminaries to guide [them]," Costello said. One problem from the outset, he noted, was that the Barcelona team didn't have much EDA or semiconductor industry experience. That's not something that can be supplied at the corporate-board level.
Barcelona had some initial success with op amps, but that wasn't a big enough business, Costello said. So it started chasing the holy grail of large IP blocks, such as phase-locked loops and A/D and D/A converters. "No one had ever done compilers for blocks of that size," he said.
But those large IP blocks were complicated to build, Costello noted, given the need to crank out "tons of variations" for different processes. And there wasn't as much of a market as the company had hoped. "It was more or less a service and consulting type of business, and it was a niche," he said.
Last June, Barcelona announced that it was swapping its IP-based model for a more conventional EDA licensing scheme. As the focus shifted back to the company's equation-based synthesis tools, Barcelona cut its 60-some staff roughly in half. But Costello said that the switch to an EDA model never really happened.
Instead, he said, Barcelona started looking at ways to extend its analog compiler technology into the digital domain. When it became clear there was no "quick hit," the board decided to call it quits.
Would a more consistent business model have saved Barcelona? Possibly not, observers said, unless accompanied by a better understanding of what it takes to make analog IP a successful business.
The third-party analog IP market is a small one. In 2003 it totaled $26.6 million, said Jim Tully, chief semiconductor analyst at Gartner Dataquest, compared with $1.02 billion for semiconductor IP. Moreover, Tully said, analog IP was flat from 2002 to 2003.
"There is considerable user demand in the market, but no vendor has found a way of satisfying that demand as a standard-packaged piece of IP," Tully said. "You need a lot of service bundled with the IP in order to make it work in the customer's environment."
Jos-Maria Moniz, head of customer services and communications for Chipidea Microelectronica SA (Lisbon, Portugal), one of the largest providers of analog and mixed-signal IP, thinks the market is larger than most surveys indicate. Research studies miss the analog and mixed-signal component of key markets like cellular, wireless and digital media, he said.
Chipidea alone had $18.8 million in revenue in 2004 and is rapidly growing, Moniz said. But he emphasized that analog IP "is not a building-block market. It is about delivering complex analog subsystems." Using a suite of proprietary design tools, Chipidea's 140-person engineering team offers what Moniz called "full solutions."
Analog IP involves more than blocks going inside SoCs, he noted. "We are witnessing an interesting trend for our larger, complex analog subsystems, which are staying at 0.18 micron and going off-chip," he said. "Customers are increasingly asking us to deliver the IP for a standalone analog solution."
Competitor Qualcore Logic Inc. offers digital and analog IP and design services, with about 60 percent of its business in analog. The company has around 130 people in India and 30 in its Sunnyvale headquarters.
"To be in this business, you need to have a very broad portfolio of IP and be able to provide customization services with it," said CEO Mahendra Jain.
He noted that considerable engineering development is needed for analog IP, and he expressed doubt that a tool such as Barcelona's could fully automate it.
Indeed, "the analog IP market is really more of a services-based market," said Phil Dworsky, director of marketing for DesignWare IP at Synopsys Inc.
"You can't just provide a piece of the answer and be successful," said Felicia James, general manager of Cadence's Virtuoso business unit. "You have to look at [analog] IP within the broader system."
At least one potential Barcelona customer liked the technology but hated the business model. "They wanted to own the tools and the IP," said Michael Bourland, an analog IC designer in Austin, Texas, who declined to identify his company affiliation. "Imagine if Cadence, or more appropriately Synopsys, told RTL developers that they owned the IP developed using their tools?"
Even with good technology, Bourland said, Barcelona's tools couldn't match the performance, size and quality of a full-custom design. But they could have solved some time-to-market problems, he said.
Barcelona today consists of a few people who are winding things down and shopping around the technology. "I think there's still a home for it," Costello said. "It has real applicability to both analog and digital design.
"The morality tale, is that all the king's men can still mess up big time. Wonderful technology that can add tremendous value to the world doesn't necessarily make a giant-home-run company."
Related News
Breaking News
- HPC customer engages Sondrel for high end chip design
- Ubitium Debuts First Universal RISC-V Processor to Enable AI at No Additional Cost, as It Raises $3.7M
- TSMC drives A16, 3D process technology
- Frontgrade Gaisler Unveils GR716B, a New Standard in Space-Grade Microcontrollers
- Blueshift Memory launches BlueFive processor, accelerating computation by up to 50 times and saving up to 65% energy
Most Popular
- Cadence Unveils Arm-Based System Chiplet
- Eliyan Ports Industry's Highest Performing PHY to Samsung Foundry SF4X Process Node, Achieving up to 40 Gbps Bandwidth at Unprecedented Power Levels with UCIe-Compliant Chiplet Interconnect Technology
- TSMC drives A16, 3D process technology
- CXL Fabless Startup Panmnesia Secures Over $60M in Series A Funding, Aiming to Lead the CXL Switch Silicon Chip and CXL IP
- Blueshift Memory launches BlueFive processor, accelerating computation by up to 50 times and saving up to 65% energy
E-mail This Article | Printer-Friendly Page |