The great SDR wait: software defined radio is getting closer, but real products are still years away — with a range of technical and regulatory issues still to be resolved - Wireless

Telecom Asia, April, 2002 by John C. Tanner

Cellular operators in Japan and South Korea are expecting major traffic surges as football fans from all over the world come to town for the co-hosted 2002 World Cup football tournament in June.

Many will be bringing mobile phones, and most of those phones probably won't work since, by coincidence, the two co-hosts of the World Cup just happen to be Asia's two most difficult wireless markets to roam to -- Japan represented by PDC, cdmaOne and ARIB W-CDMA, and Korea by cdmaOne and cdma2000 1x.

It's scenarios like this that make the prospect of software-defined radio (SDR) extremely appealing. SDR, in simplistic terms, is the concept of taking radio functions usually handled by analog circuits -- modulation techniques, frequency hopping, error correction and encryption -- and allowing them to be performed by software running within DSPs, ASICs and FPGAs (field programmable gate arrays).

In English, allocating these functions means that changing a radio's parameters could be done with a software upgrade, either via wireline or even over the air. The radio could even be designed to re-program itself based on certain conditions, such as its RF environment. Conceivably, a handset could move from a GSM environment to a cdmaOne or cdma2000 1x environment simply by automatically changing its software. The same handset could also work in both environments at once.

For this alone, SDR has been hyped for the last couple of years as the ultimate harmonization solution for incompatible wireless systems, the number of which is growing thanks to 2.5G and 3G deployments. The truly universal handset would now have to support GSM/GPRS, IS-95A/B, PDC, IS-136, EDGE, ARIB W-CDMA, ETSI UMTS, and cdma2000 1x. In a few years handsets will also have to support TDD interfaces like UTRA TDD and TDS-CDMA.

The catch, as usual, is that SDR systems are years away from becoming a commercial reality. The technology is complicated, demand is uncertain, and most regulatory environments will have to make changes in their licensing and type approval rules to allow terminals to travel across borders. However, with a number of component manufacturers releasing platforms to facilitate SDR product designs over the past few months, the SDR movement is far from inactive.

Emerging technology

Despite over 30 years of existence as a military technology and some early commercial deployments by companies like AirNet and ArrayComm, commercial SDR is mostly a firmware proposition at present. Vendors currently use SDR as a way to reduce the number of platforms they have to build for different radio systems, or to make them more easy to upgrade as wireless standards evolve.

But that may change soon with companies like Xilinx, Spectrum Cell, Tropian and others announcing the availability of components and platforms that are ready for system designers to use in designing and developing their own SDR-capable products. A few companies are already close to the product stage, most notably Australia's Advanced Communications Technologies, which recently demonstrated a two-way voice call over its SpectruCell SDR base station, which can be configured to support GSM, CDMA, UMTS, and W-CDMA. ACT is preparing to stage field trials of the technology in the US later this year.

Hard science

Japan's NTT, meanwhile, has a prototype SDR base station that supports both PHS and wireless LAN interfaces, and expects that the base station can be developed to support other mobile technologies.

It's slow going, of course, but then SDR is a jarringly complex technology that's complicated even further by the ongoing migration to 3G, which itself is a complex undertaking.

"Everyone knows about Moore's Law with processor performance doubling regularly," explains Paul Ekas, marketing manager for California-based Morphics.

"You also have Shannon's Law, which says that the complexity of algorithms evolves at an even faster rate. What that means is that the algorithms you need for 3G to leverage spectrum most efficiently outstrips the capabilities of existing processor architectures that haven't been able to catch up."

To illustrate the point, he adds, ARIB W-CDMA requires over 7,900 MOPS (millions of operators per second) for a single-channel 384 kbps in ARIB W-CDMA. "DSPs can't handle that right now, and it may be another five years before they can. ASICs can handle MOPS on that level, but they take at least two years to design and develop, and once they're programmed, they're very inflexible," Ekas says.

David Squires, director of the DSP Center of Excellence of Xilinx, says FPGAs can help in that regard. "[FPGAs] are able to provide ten to 50 times the processing capability for DSP algorithms," he says.

Because the idea of using FPGAs for SP applications is relatively new, there's some doubt in the industry as to how well FPGAs can work with SDR, as FPGAs are regarded as highly flexible but power-hungry, requiring much more power to operate than DSPs or ASICs. Squires chalks this up to a matter of education. "Not everyone in the field knows the capabilities that [FPGAs] have. This is changing with time," Squires says.