Global Sources
EE Times-Asia
Stay in touch with EE Times Asia
EE Times-Asia > RF/Microwave

Call to arms on software-defined radio

Posted: 01 Jun 2006 ?? ?Print Version ?Bookmark and Share

Keywords:Rick Merritt? Texas Instruments? cellphones? software-defined radio? Texas Instruments?

Within as little as five years, cellphones will need to incorporate software-defined radios that can flexibly find and use available spectrum. Yet today, the industry lacks a standard way of handling dynamic frequency allocation and is awaiting breakthroughs in analog devices that will make that technology truly practical.

Cellphones shipping in 2010 using the pending Long Term Evolution (LTE) standard, now being drafted by the Third Generation Project Partnership, will support data rates of up to 100Mbps, said Bill Krenik, manager of advanced wireless architectures for Texas Instruments Inc. (TI), in a keynote speech at the Software Defined Radio Forum last April. Those phones may need to support as many as eight spectrum bands, five modulation schemes and five access codes.

"It's easy to see how you could get to 10-12 air interfaces in different bands. That forces us to shift gears," Krenik said. He called for a focused industry standard in cognitive radio along with RF advances in software-defined radio (SDR) to support those products.

Cognitive radio is an umbrella term for systems that can sense, learn and adapt to changing conditions, dynamically allocating which wireless frequencies they use. SDR is an implementation approach to handling some of those tasks by bringing signal processing increasingly into the digital domain.

Presentations from SDR Forum members underlined the fact that both systems and silicon architectures, as well as the regulatory regimes for SDR and cognitive radio, are still in the early stage of development.

LTE phones could require 100 times more bandwidth than today's models, but they are unlikely to get more than a twentyfold increase. "So cognitive radio will be essential," said Krenik. The problem is that researchers are not focused on the near-term issues in dynamic frequency allocation, he told the forum.

"One of my biggest issues with cognitive radio is that while there is some brilliant, sophisticated work being done, it is hard to see how it will translate into something we can bring to market in five years," Krenik said. "At TI, we'd like to see a standardization effort emerge with a group of companies, governments and universities. Then it's healthy, and you have a real potential for market uptake."

Fragmented forums
Various industry groups are investigating cognitive radio, but none have emerged with both the clout and ambition to handle the overall job. The SDR Forum aims to tackle both SDR and cognitive issues, but appears to lack the active involvement of a broad group of top chip and systems companies and governments. About 90 people attended the recent quarterly meeting in San Francisco, a third of them first-timers to the forum.

Many major chip and systems makers, including Alcatel, Analog Devices, Ericsson, Freescale and Nokia, are not members. Indeed, Krenik's talk marked the first major involvement for TI, which is not a member of the forum.

As many as 20 representatives from top electronics companies in Japan dropped out of the forum in late 2004 following disagreements with the group's management, said Takuzo Fujii, a technical consultant for Hitachi Kokusai Electric Inc. and a representative of the SDR Forum in Asia. "In this forum there are no big companies developing silicon. This is a big issue," said Fujii. "Japan made a big mistake not getting involved early in GSM, and in the worst case, they will see similar results here."

Mark Cummings, former head of the SDR Forum and still an active workgroup leader, said plenty of large companies take part in working groups and otherwise participate or observe at different levels, but many keep a low profile to guard their intellectual property.

Bruce Fette, chief scientist at General Dynamics Decision Systems, who chairs the forum's technical committee, said chipmakers such as Infineon have given input to the group, and both Intel and Microsoft sent representatives to the meeting. "The reality is that our work has not focused as much on the semiconductor implementation as the architecture of the network behind the radios," Fette said.

The IEEE, for its part, hosts a variety of efforts tackling pieces of the SDR puzzle.

The recently formed P1900 group is examining a number of interference issues that relate to cognitive radio. The 802.19 group may touch on cognitive radio as it looks at the coexistence of various wireless networks operating in unlicensed bands. Both the 802.22 and 802.11 groups are working on approaches to dynamic frequency allocation in 700MHz and Wi-Fi networks, respectively.

Dyspan conference
In an effort to foster a broader discussion, the IEEE launched Dyspan, a new conference on dynamic spectrum allocation networks, last November. The event took place in Baltimore a week before the SDR Forum's annual meeting in Los Angeles, a conflict that stirred ill feelings with the SDR Forum's leadership.

Nevertheless, many executives take active roles in both organizations and consider Dyspan a good forum because it brings together high-level economists, regulators and technologists. "It's got the right mix of people to create ideas, but it's not a standards group," said John Chapin, CTO of startup Vanu Inc., who is active in both groups.

Whatever the venue, plenty of fundamental architectural decisions about cognitive radio are still being hammered out.

The SDR Forum has drafted a concept for an end-to-end reconfigurable network. It recently got $23 million in funding, about $13 million of it from the European Commission, to take at least parts of that concept to the next stage in development. Some 32 organizations in 14 countries are participating in some form, including Huawei, Nokia, Panasonic and Toshiba.

The forum has helped develop the software communications architecture used by the military. But Vanu's Chapin said a commercial counterpart that defines system interfaces and operation-and-management parameters is still sorely needed.

At the San Francisco meeting, Virginia Tech researcher David Maldonado said it is still a matter of debate whether SDR should be implemented in hard-coded algorithms or as a "reasoning engine" guided by an adaptive-knowledge database. "It's not the right time to pick one technology over the other yet," said Maldonado.

Systems will also need a standard language to talk to one another so they can reconfigure themselves for optimal communications, said Cummings, managing director of enVia II, who oversees work on such an effort in the forum. "We have some possible prototypes and are collecting data for how to describe each piece in the radio chain," he said.

But it's in the RF front-end "where the real challenge in cellphone design lies," TI's Krenik told the forum. "That's where we don't have a solution. We have to deal with multiple external filters and duplex switches to link to the antenna. We'd like to see innovation in this area to get to the next level of configurability."

Some startups are pursuing MEMS-based solutions, while existing companies are using more traditional methods for packing discretes into chip-scale modules. Krenik speculated that perhaps at the 32nm process node or later, ADCs might have the sampling speed and granularity to connect directly to a cellphone's LNA.

"There just aren't highly tunable radios out there," said Vanu's Chapin, whose startup typically uses multiple fixed-band radios for its multiband software-defined cellular base stations.

Researchers from UC Berkeley presented a number of ideas behind their Corvus architecture, which defines a physical and link layer for a cognitive radio. It uses a three-channel scheme and requires significant work on A/D, voltage-controlled oscillator and PLL components.

Some tunable radios already exist, mainly for military applications, but they are expensive. Nova Engineering is considering making its $3,000 military radio open for commercial use and academic research. The Stingray fits on a 6 x 3 x 1inch card and spans 200MHz to 2GHz. The company hopes greater component integration could get the subsystem down to $1,500 eventually.

Meanwhile, a spinout from UCLA has developed VHDL code for a tunable-radio PHY chip that is running on a Xilinx Virtex-II FPGA, aimed at use in 802.11n and research prototypes. Silvus Communications Systems Inc. is working closely with UCLA researchers on cognitive architectures.

TI's road map
Texas Instruments (TI) has developed a number of digitally-controlled radios that fall short of being fully software-defined, Krenik said. They include a single-chip, quad-band baseband processor co-developed with Nokia, a single-chip Bluetooth device that integrates power management and a battery interface, and TI's upcoming Hollywood DVB-H tuner/demodulator, which includes power management, analog, memory and CPU features in a device that rides a 1cm2 module.

"We don't call this SDR. It is digital-controlled radio with elements that bode well for our migration to SDR," Krenik said.

It's "a pretty good guess" that TI has similar architectures in development for UMTS, Edge and other key air interfaces, Krenik said. "Once we have a complete suite of radios working in this technology, the next phase is to look at how we can combine them," he added.

By combining as many RF and analog elements as possible, TI will gradually migrate toward SDR. "We are going to evolve to SDR because there is no other course to be on. But whatever we do, our challenge is in the RF front-end," Krenik said.

- John Walko
EE Times

Article Comments - Call to arms on software-defined rad...
*? You can enter [0] more charecters.
*Verify code:


Visit Asia Webinars to learn about the latest in technology and get practical design tips.

Back to Top