EE Times has an interesting story on Intel’s wireless stategy to use CMOS.
In early May, Intel Corp.’s chief technology officer, Pat Gelsinger, offered his connect-the-dots vision of a wireless world with ubiquitous, seamless connectivity from network to network and from air interface to air interface. From Wi-Fi to cellular to ultrawideband, Gelsinger’s theoretical nomad could roam freely without ever enduring the inconvenience of a network disconnect.
Gelsinger’s vision rested on two radical underpinnings. The first was that the proposed wireless connectivity would essentially be free. The second was that Intel would be providing it. In vanilla CMOS.
That’s right. A 35-year-old PC-oriented processor company with no radio-frequency expertise was going to usurp RF behemoths like Philips, Motorola, Analog Devices, STMicroelectronics, Intersil’s wireless-LAN division (now a part of GlobespanVirata) and Broadcom, as well as specialists such as Atheros, Bermai, IceFyre and XtremeSpectrum.
Some have speculated that Gelsinger was still high on the March launch of Intel’s Centrino WLAN platform.
Ironically, Centrino doesn’t actually contain any Intel RF expertise per se: The radio comes from Philips and the baseband from Symbol Technologies. But that doesn’t bother Gelsinger, who claims ownership of all aspects of Centrino. “We own the product-which means we brought an awful lot of the RF expertise to make it a product, even though many of the core technologies come externally,” Gelsinger told EE Times.
Besides, the CTO is confident that the work currently under way at Intel Labs’ R&D division in Hillsboro, Ore., will substantiate his RF-in-CMOS predictions over the coming years, despite industry skepticism-primarily from RF incumbents-that borders on hysteria.
Gelsinger outlined how the building blocks were being put in place to realize the vision of what the company calls “Radio Free Intel.”
From the researchers’ point of view, their day of vindication came with the presentation of an all-digital-CMOS 10-GHz agile-RF frequency synthesizer at the 2003 Symposia on VLSI Technology and Circuits in Kyoto, Japan, in early June. The phase noise in this device, which is based on a 5-GHz CMOS voltage-controlled oscillator (VCO), was specified as -120 dBc at a 1-MHz offset. “That’s when the world stood up and took notice,” said Steve Pawlowski, an Intel Fellow who leads the Hillsboro R&D group along with Kevin Kahn, communications and interconnect labs director.
At one time, “If I was doing a 5-GHz radio with a 500-MHz CMOS process, that was next to impossible. Now I have a 5-GHz CMOS process.” Moore’s Law also provided the multiple gigaflops of processing power to enable such functions as software-defined radios with reconfigurable architectures and smart antennas, and the clout to do RF processing in CMOS.
Wireless talent had to be recruited, which became a bigger issue than anticipated. Many RF engineers perceived Intel as a PC company where their talents would be eclipsed by the success of the processor and computing group. No one had a reason to move-”especially not to the Pacific Northwest from San Diego and the Bay Area,” Pawlowski said. “When we first started, we looked for senior talent,” he said. “All told us that ‘CMOS isn’t the technology. You want III-V materials [silicon germanium, gallium arsenide and so on].”
So Pawlowski and Kahn turned to university professors. “We told them what we were doing and offered them research grants to work on it,” said Pawlowski. Intel also gave students the chance to come in-house and implement those designs that looked promising.
The idea worked out. “The results are much better than what we would have gotten had we gone looking for the best-and-brightest RF guy,” said Pawlowski. “The students were pretty fresh.”
The university grants continue, with Gelsinger’s group alone issuing 400 per year. Instead of acquiring startups, Gelsinger is sticking with a simpler strategy of funding research and dispersing dollars among companies devising technologies that could help in the dissemination of Wi-Fi. “I can give a research grant of $50,000 to five different researchers and have them solve my problems,” he said. “I now have five of the top people on the planet working for me for $250,000. If I buy one company I’ll have to at least spend a few million.”
Intel expects that Centrino–a bundle that includes a Pentium-M processor, a wireless chip and a chipset–will become more widespread. Especially as 90 nanometer process technology becomes mainstream and radio parts get integrated with the processor. Intel will come out with a Wi-Fi chip that can connect to 802.11a and 802.11b networks in the third quarter. A similar chip compatible with 802.11b and 802.11g technology is set to arrive in the fourth quarter.
Related Daily Wireless articles include Teresa Meng’s Journey, Meshing Around at Intel, Sensor Nets, Intel’s wireless plans, Intel Researching the Last Mile, Craig Barret on China, Berkeley’s Wireless Research Center and Behind the Curtain at Intel Labs.
.gif)
