4G wireless challenge raises stakes for DSPs

Industry observers expect that developers will introduce 4G wireless communications systems between 2007 and 2010; historically, previous systems have seen a change of generation every 10 years or so. Ultimately, though, the success of 4G systems depends on how economically companies can build complex and overblown adaptive arrays, large power amplifiers and giant digital signal processors.

To make these systems cost-effective, the functions of analog circuits need to be digitized as much as possible, and A/D conversion and digital processing needs to occur as close to the antenna as possible, according to Yukou Mochida, senior vice president at Fujitsu Laboratories Ltd. The digital processing section will start being implemented by lining up processing blocks that use DSPs and FPGAs.

Each of the three generations of wireless technologies has produced a major breakthrough, made possible by progress in semiconductors and devices. Even more will be expected of devices as 3G evolves to 4G. In particular, the system side and device side will have to co-operate more fully than they did with 2G and 3G devices the fourth generation is to see the light of day.

A sizable burden will fall to semiconductor manufacturers to develop high-Bips DSPs, to write the algorithms that will drive the applications and design the architectures that will merge the computer, communications, audio and video worlds.

This week's In Focus looks at players who are or who hope to be at the leading edge of wireless signal processing design. Unlike previous generations of wireless devices, where the communications part and processing part existed as separate physical entities, 4G handheld devices and basestations alike will be based on system-on-chip (SoCs) that combine processing and communicating onto one IC.

Among other requirements, 4G systems demand always-on mobile devices connected to the Internet. Also, although person-to-person communications have dominated conventional communications systems, future communications between people and computers, and between computers and other similar equipment, will be much more common than is evident today. And from the user's perspective, demand for ever-higher speeds and ease of use will continue unabated.

In his IEDM keynote last year, Fujitsu's Mochida suggested that if DSPs have increased in power by an order of magnitude in each generation, from 4 Mips through 40 Mips to 400 Mips, then it might take 4 Gips to accommodate 4G applications.

Research and development into 4G mobile communications systems is already well established. System requirements for 4G devices include support for minimum transmission speeds of 100-Mbits/second downlink and 30 Mbits/s uplink. But, Mochida said, major problems must be overcome if data rates of 100 Mbits/s are to become a reality. Among these problems: The transmission power must be increased without compromising operational quality. Also, signal processing needs to be applied to modems as a countermeasure to severe delay spreads. And, of course, system capacity needs to be high enough to accommodate high-data rate users who might otherwise hinder other users.

According to Mochida, the ideal frequency range for 4G must lend itself to an economical service rollout and offer a high level of mobility. The use of high frequencies makes an economic rollout difficult, however, because high frequencies make cell radii extremely small.

The success of cost-reduction schemes hinges on the rate of IC integration, yet there are many areas that DSP will find hard to tackle. Even today, although 3G devices have not quite reached the point of practical use, expectations are high with regard to developing compact, single-chip solutions that use algorithms to lessen the dependence of RF analog circuits on frequency, as in direct conversion, and reduce the number of components.Where 4G is concerned, the specifications of the radio system itself will grow more stringent in order to make efficient use of frequency resources.

Mochida said he expects that 4G baseband sections will be fully digitized. He also said that it will become commonplace for a signal to be digitized using an A/D converter at the baseband entrance, with all processing after that point done digitally. As digitization progresses, it will spread to the modulation/demodulation functionality of the IF section; eventually, all analog processing will be transformed into digital processing.

Still unresolved, though, is how to realize such an extremely heavy digital processing portion. At the start of 2G, handsets centered around voice communications, and almost all digital processing was taken up by the processing in the voice encoder. Although different voice encoders were used in different regions, they all required a throughput of between 10 and 50 Mips, which was comfortably handled by general-purpose DSPs of the time.

As 2G gives way to 3G, the handset is evolving from a tool for voice communications to a high-speed data terminal. In line with this shift, the range of applications has expanded and now includes video encoding. Throughout this transition, digital processing was applied to more and more sections of baseband processing.

Extrapolating, a throughput of several Gips will be required across the handset. The use of orthogonal frequency division multiplexing (OFDM) for modulation is being investigated for 4G, to enable high-speed data services in a mobile environment. "If we attempt to deal with such high-speed processor-intensive modulation entirely in software, it will require a throughput more than 10 times greater," Mochida said.

"Speculating beyond 3G is tough, when we don't really know what 3G will be when it grows up," said Will Strauss, of Forward Concepts, a Tempe, Ariz.-based market researcher. "Everybody assumes that video will be the defining technology in 3G handsets, and certainly MPEG4 will be part of them."

Strauss said most of the 4G buzz has been about the air interface. "It won't be WCDMA or cdma2000, but likely will be OFDM, and to enable the higher data rates, we will probably have to move to higher radio frequencies, probably above 3 GHz [as opposed to present-day 800-MHz cellular and 1.9-GHz PCS frequencies]." That means the FCC and WARC (the world spectrum allocation/regulatory body) will have to make new frequency allocations, Strauss said.

Strauss said 4G basestations will use smart antennas to directly transmit or receive radio-beam patterns to and from individual users, which will make possible more reliable calls at greater distances from base stations. Greater DSP power will enable better ameliorations of fading and interference from multipath reflections and from other cell phones, producing better quality audio and video. "And we will have biometric security features like thumbprint readers and location-centric (GPS and more) capabilities as well," Strauss said. "Some of these features will begin to show in 2.5G cell phones, but they will be perfected and expanded in capabilities by the 4G time frame."

Fourth-generation devices are not likely to become widely available until about 2010. Strauss predicts that DSP clock speeds will exceed the 3-GHz range by then, with many parallel processing elements accounting for hundreds of billions of operations per second. "Naturally, SoC approaches will be the norm," Strauss added.

Industry observers predict CMOS technology will see a doubling of circuit miniaturization and integration by 2005. This development should secure devices that are can enable full digitization of 3G baseband processing. However, an even greater leap forward is required to bring about the ultimate all-digital handsets for 4G.

While CMOS technology is expected to continue its progress in miniaturization, higher speeds and lower power consumption through 2015, there will be a visible gap between the rate of progress in CMOS and that of the handset. This gap will creating a serious hurdle for the early introduction of 4G, according to Fujitsu's Mochida.

Such a development would mean that compatibility with multiple radio systems could be achieved in software alone, enabling the development of simple terminals that can communicate from anywhere in the world. Users could adapt communications according to end use, with complete freedom to select their own style of services irrespective of network or operator, bringing the ultimate dream of software-defined radio to reality.