More on the future of 3G Basestations Norden------------------------------------------------------------------------
Technologists define future cellular components
Stephan Ohr
San Diego - Technologists from the Nokia Research Center, Lucent Technologies' Bell Laboratories and the Virginia Polytechnic Institute took a stab at defining the component requirements for new-generation cellular systems at a wireless communications show here last week. Software radios and third-generation, or "3G," cell phones will be among the 21st-century services to depend on new wireless components and packaging techniques, the experts said.
Such services would benefit from smart antennas and advanced component technologies. But improving the bandwidth of A/D converters, increasing the DSP Mips rate and reducing the passive-component count on the cellular handset's RF front end will be the toughest job, the experts agreed.
Their remarks were delivered during the Sunday plenary session at the third annual Wireless Communications Conference, part of the International Microelectronics and Packaging Society (iMAPS) conference and exhibition here.
James Drehle of Hewlett-Packard Co. (Colorado Springs, Colo.), speaking as past president of iMAPS, said the semiconductor technologist needs to know more about the requirements of the wireless-system designer, and "the designer requires more knowledge about processes." Wireless technology-especially cellular telephony-will be a driver for IC component and packaging development, he said.
Shipments of cell phones surpassed PC shipments in 1996, confirmed Kari-Pekka Estola, vice president of Nokia and director of its electronics laboratory (Helsinki, Finland). More than 100 million cellular handsets will ship in 1998, he said, rising to 140 million in 1999.
Cell-phone usage is highest in Europe; as of March, penetration in the United States was only 21 percent, according to statistics from the World Bank. Thus, while the Asia-Pacific region-especially China-represents the largest growth opportunity, North America will account for some of next year's growth. Estola predicted there will be 2.6 billion cell-phone users worldwide in the year 2015.
The projected 3G cellular services may be a driver for North American business, said Estola, though Europeans are increasingly using GSM for data transmission and Internet Protocol services. The 3G services will use higher data rates, he said: 64 to 144 kbits/second in rural outdoor areas, 384 kbits/s in the urban outdoors and up to 2 Mbits/s indoors over short distances. Those higher data rates will allow the integration of multimedia services with cellular voice transmission.
That prospect has companies dreaming up a slew of new services; for example, local maps and restaurant guides could be downloaded onto a cell phone screen. Such services would give rise to software packages for voice recognition and specialized Web browsers, among others. A 3G search engine might allow you to play a few musical notes on a keyboard and download a list of symphonies and songs that use that theme, Estola said.
Some of the support technologies for such futuristic scenarios are already moving into place. An MPEG-7 compression standard, which would allow multimedia browsing, for example, could be defined by 2000.
While an obvious attention-getter, 3G phones would not stand alone, Estola said. They would be part of a service package that would include packet networks and interconnectivity with other computerized appliances, such as faxes and printers.
If 3G phones are one preoccupation of the wireless world, the software radio is another. Design requirements for such an animal were sketched by Jeff Reed, an associate professor at Virginia Tech (Blacksburg, Va.).
In principle, the software radio makes it possible for RF modulation schemes, signal parameters and data protocols handled by the RF front end to be configured and adjusted on the fly. While the content of those cellular systems would be dependent on DSP, Reed said, the reprogrammability applies to the RF front end, not just the baseband portions. He cited work done by Harris Semiconductor (Melbourne, Fla.) in software but suggested that the actual arrival of software radios will lag 3G phones by several years.
The software radio includes four major blocks: narrowband A/Ds and D/As, for voice conversion; microcontrollers and DSPs, for compression and modulation; wideband A/Ds and D/As, for driving RF components; and RF converters and antenna drivers. Thus, the enabling technologies would include faster data converters, more powerful processors, Java and other forms of downloadable software, said Reed.
The biggest "system engineering" issue, he said, lies in the RF front end-the transmitter/receiver section. The antenna, for example, is less than 10 percent of the wavelength of the carrier frequency. Because of its limited size, Reed said, the antenna can not easily support the multiple frequencies (900 MHz and 2 GHz) required for dual-mode phones. Obtaining full-duplex capability-the ability to send and receive from the same antenna at the same time-is something of a "black art," he said.
On the transmit side, phase noise of the local oscillator and transmitter power efficiency are major concerns. "The PA [power amplifier] seems to be the Achilles' heel of wireless," said Reed. With AMPS phones, he said, it absorbs 60 percent of the battery power; with digital IS-54, the percentage is more like 30 to 45 percent; and with FSM, it's 45 to 50 percent. Reed believes power efficiency is traded for low distortion and wide bandwidth: "A lot of power is wasted in the transmitter requirement."
For receivers, a strategic design decision is the placement of the wideband A/D converter in the RF signal-processing chain. The goal is to place it as far forward toward the antenna as possible, but that decision affects the cost, the component count and the requirements on the DSP.
The problem is that the A/D is a major power consumer, Reed said. The farther forward the A/D converter goes, the higher the sampling rate required; the higher the sampling rate, the greater the power consumption. The front end of a DECT phone, for example, consumes about 200 mW in use. Of that, some 40 mW are eaten by the receiver's low-noise amplifier; the RF mixer-oscillator consumes 50 mW. But the A/D converter uses 100 mW by itself.
The major advantage of pushing the A/D converter forward, though, is to save the power and component count absorbed by the RF tuning filters. The receive filter includes two elements: an initial bandpass filter and a second filter for spurious-image rejection. By putting the A/D converter at the head of the signal-processing chain, a variety of passive components and active tuning elements can be eliminated. And the DSP could do the majority of the work involved in extracting a clean signal.
The DSP could help compensate for cheaper RF components, Reed suggested. Advanced signal-processing techniques-smart antennas, for example-provide interference rejection. Multipath mitigation could be controlled using channel-estimation techniques. Compression techniques provide both RF modulation and voice enhancement, such as speech recognition, echo cancellation and elimination of background noise.
Certainly, Reed sees voice-recognition capabilities integrated into the cell phone. "It's too difficult to get a full keyboard into your handset," he said.
Fortunately, processors cost 4 or 5 cents per Mips, and "cheap Mips" would be an asset for the software radio, said Reed. The 3G phones, for example, would require more than 3 billion operations/s. Data converters that cost $4,000 10 years ago can now be obtained as semiconductor devices for a couple of dollars, said Reed.
Indeed, the celebration of semiconductor technologies was a common theme at iMAPS. Robert Frye of Bell Laboratories (Murray Hill, N.J.) pointed out that the first mobile cellular equipment, rolled out in Chicago field trials in 1978, had 130 ICs and 300 discrete semiconductor components, and occupied more than 2,000 cubic inches. The Motorola MicroTAC cellular handset introduced in 1994 occupies about 14 cubic inches and has 20 ICs and about 45 discretes. A contemporary PCMCIA module, said Frye, includes 15 SMT-packaged ICs and about 30 discretes. It takes up about 2 cubic inches.
The so-called "soft radio," which Bell Labs believes will be out around 2002, will have two ICs and five discretes. Estola of Nokia remarked that 10,000 transistors these days cost about the same amount as a paper clip.
Estola said that today's 2G handsets use an 8- or 16-bit microcontroller running at 10 MHz, a 3- to 30-Mips DSP for speech coding, another 30 Mips for radio-channel encoding and about 4 Mbits of memory. The system draws power from a nickel-metal-hydride or lithium-ion battery.
The baseband elements use 3-V CMOS, said Estola, though the RF transmitter/receiver section uses a proliferation of discretes, passive components and ASICs. Half of the terminal's work is already performed in software, he noted.
In contrast, said Estola, the 3G terminal of the year 2002 will likely use a 16- or 32-bit processor running at 50 MHz. It will pack an Li-ion or lithium-polymer battery and will use 1-V CMOS. Memory requirements will be up to 64 Mbits. If the previous generation relied on SMT-packaged ICs, the next will use micro-ball-grid arrays and multichip modules.
But the most dramatic change will be in the DSP Mips requirement. By Estola's reckoning, 3G will take 200 Mips for radio-channel coding, 30 Mips for speech coding, 50 Mips for voice control (a limited form of speech recognition) and 100 Mips for video coding.
The cell-phone handset of the year 2000 is still likely to contain 100 or more passives-fully 95 percent of the total components, said Estola. |
