22 May 2012
Expert panel: Innovation, collaboration and competition are hot topics for semiconductor developers
Maria Marced, president, TSMC Europe
Svenn Tore Larsen, ceo, Nordic Semiconductor
Graham Budd, coo, ARM
Marc Cetto, ST-Ericsson
Peter Schiefer, president, power management and multimarket division, Infineon
Steve Wainwright, general manager, EMEA, Freescale
Jim Aralis, cto, Microsemi
You might think that a roundtable at a meeting of the Global Semiconductor Alliance (GSA) charged with discussing the future of the semiconductor industry would have addressed technology issues; the future for Moore's Law, for example. Instead, the discussion addressed complementary topics.
And that's not a surprise: the GSA brings together a range of fabless semiconductor companies, organisations from the silicon supply chain and some 'fab lite' businesses, such as Freescale and STMicroelectronics – each with a different perspective on the problems of creating or enabling leading edge designs.
Moderating the discussion was Maria Marced, president of TSMC Europe. She said the participants had two things in common: they were all working with analogue, mixed signal and rf technology; and they were all incorporating ARM cores in their products. She said innovation was another common feature. "But when do you innovate?," she asked the panellists, "and what are the challenges?"
First to contribute was Serge Willeneger, vp of product management with Qualcomm CDMA Technologies. "We invest more than 15% of revenues into R&D," he claimed, "so it's important that we nurture and fund innovation. That sounds obvious, but not everyone does it."
He also pointed out that R&D failure is an option. "We don't encourage failure," he continued, "but we've had many projects which haven't been successful, but which have helped in other areas later on. Provided there's a purpose to the work, there's always value in pursuing different leads."
Innovation at Nordic Semiconductor takes a different approach, said Svenn Tore Larsen, the company's ceo. "We're a small company," he noted, "with only 170 employees. Innovation for us is integrating our wireless technology into new sensor technologies and we are working on this with four or five universities. Not all apps will have IP addresses; there will still be the need for radio communications. There's a lot of innovation out there," he believed, "but it needs to be discovered before it gets too well developed."
Marc Cetto from ST-Ericsson gave a plain answer. "Innovation is what you have to do." As part of its recent restructuring, ST-Ericsson has decided to work with STMicroelectronics in the development of application processors. "We are doing this to offer a guarantee to content providers that everything will work in the same way," Cetto explained.
He said a critical part of the innovation process is how quickly you can get the results to market. "By being active on whole platforms and by partnering, we can make sure we do that."
Innovation is only one element of the semiconductor business; growing companies is an equally large task. Marced asked the panel where they see future growth coming from. "It has to be the mobile market," said ARM's coo Graham Budd. "It will continue to expand rapidly. But there is also great potential in the smart sensors/grid/meter sector, with products built around microcontrollers." He believes the smart sector requires a different approach to thinking about energy reduction. "There are two reasons," he said. "One is that energy reduction is good for business, but the other is the explosion in mobile computing is driving up energy consumption in the cloud. A potential 'killer app' will create a new way of looking at services, because there will be huge cost savings from reduced energy consumption."
Peter Schiefer, president of Infineon's power management and multimarket division, sees three challenges, each presenting companies with opportunities. "Security, mobility and energy efficiency." Picking up on Budd's point, he said: "If you share pictures, you tend to use Facebook. A cpu in the data centre will consume 60W. That power needs dc/dc conversion, ac/dc conversion and transmission. The 60W consumed by the cpu started out as 220W at the power plant. If you use semiconductors to support renewable energy, you can save 90% of those losses."
Jim Aralis, Microsemi's chief technology officer, sees potential in medical applications. "Prevention and monitoring will need low power radios, sensors and so on. We'll see tremendous growth, but it will take a while to come of age." Like Schiefer, he sees opportunities in the data security sector. In terms of 'killer apps', he believes voice and facial recognition technologies will be in demand. "But there's also potential in timing technology. Systems wake up and go to sleep to save power, but how can you reduce power consumption by 100 times? Timing will be important here."
Steve Wainwright, Freescale's general manager, EMEA, said the connected car was an important area. "One of the apps that will be useful will be reducing fatalities on the road. How do you move safety from being passive to active to predictive? How can cars be connected beyond mobile telephony?" Addressing the analogue and mixed signal issue, he said Europe was doing well 'because there's a higher entry point for these technologies than there is for digital'.
Aralis picked up the issue. "Companies have always benefited from differentiating through analogue technology. But it's becoming a problem because there aren't as many analogue designers as the industry needs, although Europe has always done well when it comes to analogue education."
He also sees the opportunity to use analogue devices as the basis for more system oriented products. "3d packaging allows designers to use the appropriate technology," he pointed out, "and to take a system in package approach, rather than pursuing the more expensive SoC route. Some companies are already recognising this by de-integrating rf from the baseband chip and housing it in an analogue front end."
Nordic's Larsen added: "We have the basic analogue technology we need to integrate more digital technology and sensors, but we need to work with leading suppliers in order to take advantage of the opportunities." He warned developers that two factors were critical: cost and power consumption. "We need to understand how to meet these factors with the facilities available through foundries like TSMC."
Wainwright contended that talk about analogue and digital technology was becoming redundant. "The divisions are disappearing. Analogue people are getting more involved in understanding architectures and are recruiting microcontroller experts. Looking at our road map, it's getting harder to work out which side the products are coming from. Devices are getting smarter and smarter, integrating cores into products which were once only analogue."
Schiefer claimed analogue and mixed signal design was about 'special people'. "They have the ability to combine sophisticated analogue and digital functions," he said. "We need young people to join the industry, but we see problems – even in Europe – about how to attract them."
A concept close to the heart of the GSA is ecosystems; the collection of complementary companies providing support to each other. "Why are some successful?," Marced asked the panellists, "and does the size of the ecosystem matter?"
Willeneger responded first. "Those which succeed are those which meet needs and sometimes it's to do with standards." He said ecosystems can arise through design or, effectively, through accident. "GSM, for example, was politically motivated, while the Windows/Intel ecosystem (Wintel) developed because of the de facto market position. But efficiency is important and that's why the ARM ecosystem is in a position to challenge Wintel."
Should ecosystems be open or closed? Cetto said ecosystems can only be successful. "Apple is a closed ecosystem," he offered, "while others are more open. Both approaches work, but the key is that an ecosystem has to deliver a good user experience."
Cetto talked from his experience of porting an application – in this case video streaming website Netflix – to 172 different Android platforms. "We ended up with 172 different code trees, while Apple has only one. The problem is that the 172 code trees could easily rise to 500 in the next three years, so there will be the need for more partnerships in the future."
Wainwright said that, when you look at applications processors, the need for an ecosystem was obvious. "There are hundreds of customers dealing with complex multicore processors and they need an infrastructure for which someone takes responsibility. But it is also important for there to be options; companies have to dedicate time to find partners at all levels – from development tools all the way through – in order to help customers design products."
He also believes the ecosystem shouldn't result in one element being overloaded. "Distributors, for example, shouldn't be overwhelmed with requests for technical support."
The ARM ecosystem is perceived as one of the more successful environments of its type. How did it develop? "Partly because of standards," said Budd, "and partly because it's an enabling mechanism. But it's also about business; providing opportunities for companies to create value around a point in the market, spread development costs and get products to market more quickly. An ecosystem encourages innovation."
Marced asked Budd how he thought the ecosystem might develop. "We now have more than 950 partners working together," he said, "and scale is important. But it is also important that the market which enables the ecosystem is large." Budd also believes the ARM ecosystem is somewhere which can help small companies succeed. "There are places where small companies can add value," he noted.
As the cost of developing leading edge chips increases, the number of companies who can foot the bill decreases. Some see this as discouraging competition. Aralis believes this will stifle innovation. "And this industry needs innovation in order to keep moving forward."
But Willeneger countered. "Qualcomm might make it look like there's no competition, but there is. Qualcomm is in the position it is because it has invested in innovation and it's false to think there's no competition."
The problem, in his view, is that only two handset manufacturers are profitable. "That has an impact on the market and we all need to nurture diversity," he concluded.