All... very interesting 3G article...
Politics dim hopes for unified 3G
standards
By Loring Wirbel
Would-be developers of the third generation of the digital cellular
personal-communications-system (PCS) phone, known as 3G, are
perplexed. That's because the closer the regional standards bodies come to
defining specifications, the more elusive the defining purpose for such a
system becomes. When the British government and the European
Telecommunications Standards Institute (ETSI) started floating concepts for a
"Future Public Land Mobile Telephone System" in the early 1990s, the idea
was to unify baseband air interfaces and provide enough frequency spectrum
for wideband data services capable of hitting speeds of several hundred kbits/
second. But as the former FPLMTS became the Universal Mobile
Telecommunications Service (UMTS), politics and design inertia colluded to
make the goals for 3G fuzzy at best.
In the first place, differences in goals within the standards bodies for Europe,
Japan and North America are calling into question whether air interfaces can
be unified. The wideband code-division multiple-access (W-CDMA)
standard touted by Nippon Telephone & Telegraph's DoCoMo division
seems to have amassed the greatest amount of support in all regions, with
OEMs and semiconductor companies flocking behind W-CDMA for
European markets in particular.
However, the partisans supporting Global System for Mobile telecom (GSM)
and hybrid time-division multiple-access (TDMA) systems do not appear
willing to go down without a fight. And because backward compatibility with
IS-95 (CDMA) and IS-136 (TDMA) remains a key goal, baseband
specialists are saying that a single air interface for global use may prove
impossible to achieve.
Last month, ETSI sought to bring peace to the new International Mobile
Telecommunication-2000 (IMT-2000) proposals by suggesting a dual
standard, with a hybrid concept called TD-CDMA used for
time-division-duplex services and W-CDMA reserved for frequency-division
duplex services. But that compromise does not begin to address the problem
in North America.
Meanwhile, the supporters of traditional IS-136 TDMA are hardly giving up.
Hardware developers and carriers have formed the Universal Wireless
Communications Consortium to promote an incremental move from IS-136
to 3G. At a recent forum in Atlanta, the group proposed moving first to
IS-136+, offering higher-quality voice compression and 64-kbit/s 1-MHz
channels for packet- and circuit-switched data. That strategy is similar to
what the CDMA Developers Group (CDG) proposes for the IS-95B
follow-on to IS-95 CDMA.
The TDMA camp also proposes IS-136 HS, which would have scaled data
channels, ranging from 144 kbits/s for high-speed automotive applications to
2 Mbits/s for fixed cellular. Again, CDG has a similar concept for CDMA.
Both groups pledge to submit all proposals to the International
Telecommunications Union for IMT-2000 consideration. They seem
oblivious to the fact that ETSI and Japanese government decisions have all
but shut out North American TDMA and CDMA proposals.
In fact, the Western Hemisphere may have written itself out of the picture for
full integration with Europe and Asia. The United States leads the way for
Canada and Latin America and, even though U.S. officials agreed to support
the IMT-2000 standard, the politics of the PCS auctions intervened. Rather
than use large swaths of the 2-GHz band for voiceband and wideband
services, portions of the IMT-2000 spectrum were parceled out to
PCS-auction winners. Former FCC chairman Reed Hundt's vision that PCS
would become the equivalent of 3G phone service fell victim to an industry
that wanted to create a fast-deployed down-and-dirty upbanded digital
cellular standard.
"At this point, there does not seem to be any clear way to prevent the
Western Hemisphere from being different from the rest of the world, which
might make a SIM [subscriber identification module] card necessary for a
true global phone," said Chris Wallace, vice president for technical standards
at Nokia Corp. "Carriers in the U.S. aren't going to rip out infrastructure just
to get to IMT-2000 compliance."
But if developers have to settle for programming interfaces and gateways for
software roaming among existing air interfaces, the complexity of
microcontrollers and baseband processors required for implementing those
interfaces may cause designers to sour on 3G. Even the purported key reason
for moving to 3G-the addition of wideband channels through assignment of
wider frequency bands centered on 2 GHz-is finding some naysayers, who
question market maturity for data services rather than the core baseband
DSP capability.
For all the talk about accessing Web sites through cellular phones, wireless
data for cellular networks remains largely a nonstarter, said Qualcomm Inc.
chairman Irwin Jacobs. Cellular Digital Packet Data for analog Advanced
Mobile Phone Service (AMPS) networks remains all but unused today, he
said, while specialized mobile-radio packet services like Ardis and RAM
Mobile Data are struggling to keep data services alive.
Qualcomm is moving to an enhanced version of CDMA, IS-95B, that
supports data speeds of up to 115 kbits/s. But the chip sets Qualcomm offers
for IS-95B are being touted first and foremost for low power dissipation, not
better data throughput. "We want to be ready when interest in data in a PCS
or digital cellular environment takes off," Jacobs said, "but that interest is still
dominated by circuit-switched cellular modems. Packet data over wireless
remains a small market."
Part of the problem, in the view of Wallace of Nokia, is that telecom service
providers must abandon per-packet concepts to charge for services,
particularly since the TCP/IP world brings with it an assumption of
all-you-can-eat services. There are plenty of new ideas for bringing better
data services to hardwired handsets or programmable mobile platforms,
Wallace said, but those concepts must be combined with a willingness among
carriers to look at some form of flat-rate pricing for wireless data.
At February's Wireless '98 show in Atlanta, there were plenty of new
handheld hardware architectures that suggested a better integration of
wideband data and voiceband service than were seen in first-generation
cellular data phones, like the ill-fated Simon from an IBM-BellSouth coalition.
Nokia was showing off its 9000i GSM smart phone, which treats the handset
as an open platform with user-definable programming interfaces, rather than a
closed-system voice handset. Sony Wireless Telecommunications Co.
garnered a lot of attention with its Cosm smart CDMA phone, with a 320 x
200-pixel active-matrix color display on the back of the handset capable of
displaying HTML- or Java-based small applications from servers in the
cellular network.
At the basestation and central office, software developers are looking for
more realistic server applications for these new data platforms. Fujitsu
Software Corp. (San Jose, Calif.) showed off a server application called
ByeDesk, which could concatenate the Short Message Service datagrams of
GSM into e-mail chunks of up to 2 kbytes, allowing existing digital cellular
users to have data capability before 3G services roll out. And Unwired Planet
Inc. (Mountain View, Calif.) continued to sign up developers for its concept
of a Handheld Device Markup Language, creating stripped-down versions of
Web sites that could be accessed by character-oriented handheld terminals.
Nevertheless, Wallace of Nokia said, development of true wideband
applications for 3G phones remains very much a chicken-and-egg problem.
AMPS reliability
Another sticking point in North American markets, said Richard Sfeir,
director of marketing at IBM Microelectronics subsidiary CommQuest Inc.
(Encinitas, Calif.), is that AMPS proved so robust and reliable in many
regions, neither an 800-MHz digital cellular service nor a 1.9-GHz PCS
service carried the importance it did in Europe, where a unifying digital system
like GSM was necessary. This means that some regional carriers could leap
straight from AMPS to 3G, bypassing two generations of digital cellular in the
process.
But what precisely does a leapfrog to 3G mean? "We know enough to design
RF chains for broader bandwidth and to look to more powerful DSP blocks
to handle data right now, but that's about it," Sfeir said. "We see the
European and Japanese markets lining up strongly behind W-CDMA, but it is
not clear what a global 3G system will look like-or if such a system will truly
emerge globally. We're anticipating a model of small islands of 3G within a
sea of 2G, and that model could have a long lifetime."
This does not mean that baseband developers of DSP architectures or
mixed-signal RF/IF designers are standing still, waiting for a common
standard that may not emerge. Instead, designers in both the digital and
analog domains are discovering that the steps taken to support
multiple-frequency-band and/or multiple-mode phones are precisely those
that can aid in moving to 3G designs.
Consequently, many semiconductor developers have been busy in early 1998
updating GSM and CDMA chip-set designs. IS-136, by contrast, seems to
be relatively mature.
On the baseband front, designers are combining MAC and barrel-shifter
DSP cores with general-purpose RISC microprocessor cores to create
unified baseband processors capable of handling dual or triple air-interface
modes. The on-chip RISC eliminates the need for a separate integer
microcontroller. In many cases, that RISC core has enough horsepower left
to handle the data-processing needs of IS-95B or IS-136+ and perhaps the
multichannel, wideband data needs of such emerging 3G standards as
W-CDMA. Anne Wilke, director of wireless marketing at Rockwell
Semiconductor Systems (Newport Beach, Calif.), said the processing power
of the ARM7 core used in Rockwell's new GSM chip set ensures that
"packet processing will not be a problem in existing baseband designs, if new
support for data is offered in 3G."
Similarly, the use of frequency bands between 1.875 and 1.975 GHz and
between 2.11 and 2.16 GHz in the IMT-2000 proposal is similar enough to
U.S. PCS to require very few changes in the RF chain. Handling multiple data
channels may require greater flexibility in frequency synthesis and
down-conversion blocks, but those are well within the range of modern
design capabilities.
In fact, 3G will be able to take advantage of trends that are becoming evident
in the digital PCS world. Use of specialized process technologies such as
gallium arsenide are becoming reserved for only the power amps and
low-noise amps closest to the antenna. Bipolar technologies can be used for
most blocks in IF and RF frequencies, and standard sub-half-micron CMOS
is finding applications stemming from baseband out to the first
down-conversion block.
New process technologies for CMOS substrates, such as IBM's
silicon-germanium process, make it conceivable that single-chip processors
handling both baseband and RF/IF could be on the drawing board as 3G
standards mature. That will accelerate as handsets turn from current,
superheterodyne or direct-conversion designs to wideband A/D-D/A
conversion designs that bypass I and Q phase balancing steps. That move is
advocated by DSP processor specialists such as Texas Instruments Inc.
(Dallas) and Analog Devices Inc. (Norwood, Mass.), particularly as they
move to designs with integrated RISC.
However, practical wideband data-conversion systems are a few generations
off, and there are other important limitations to such integration. The biggest
problem is not CMOS' ability to hit higher transmit-power but the noise
considerations that mandate separate baseband and RF/IF blocks in today's
designs. In theory, however, better noise isolation among synthesized blocks
in an ASIC design could allow for handsets in which the vast majority of
chips are implemented in CMOS, thus improving talk time and battery
standby life by an order of magnitude.
The realization of the advances to be made in 2G digital cellular phone
systems, at both 900-MHz and PCS frequencies, has sparked a renaissance
in designs for GSM and CDMA, particularly as semiconductor vendors
recognize that there will be a legitimate second-tier handset business outside
giants such as Nokia, Ericsson, Motorola and Qualcomm. Since the start of
the year, new baseband architectures for GSM have arrived from
IBM/CommQuest, LSI Logic Corp. (Milpitas, Calif.) and Analog Devices
Inc. For RF/IF blocks, the former Plessey group in Mitel Semiconductor Inc.
(Kanata, Canada) has released its Planets chip set, which handles everything
from low-noise amp interface to baseband controller interface, while Philips
Components Inc. has updated several power amps and mixers. Rockwell has
taken the boldest approach of all, having fielded a seven-chip GSM set in late
March that incorporates everything from an ARM DSP baseband processor
to the GaAs power amp next to the antenna.
Embedded RISC
Baseband trends show a mix of hardwired and embedded RISC approaches.
Analog Devices, for example, has gone through two generations of GSM
sets, now led by the AD20msp425, in which a highly optimized DSP handles
GSM signal processing while a codec developed with the Technology
Partnership Ltd. handles voice codec functions. Most DSP players are
adding embedded RISC to their core DSPs, with TI and Rockwell favoring
Advanced RISC Machines ARM7 and IBM/CommQuest preferring a MIPS
R3000 variant. LSI Logic uses a TinyRISC for integer and Oak core for
DSP, but it can add MIPS cores for greater processing power when
necessary.
While CDMA has been slower to develop because of the tight hold
Qualcomm has traditionally held on the IC licensee market, the Atlanta
Wireless '98 show represented the arrival of a legitimate CDMA
semiconductor market (see March 2, page 10). Qualcomm's ASIC products
group updated both its baseband processor and IF devices for the IS-95B
standard, launching its first baseband processor with an integrated ARM
core. The Mobile Station Modem 3000, or MSM3000, processor not only
has the processing power necessary to handle fast packet processing, but has
moved to a new rake-receiver demodulator architecture, dubbed
SuperFinger, that can demodulate six or eight channels simultaneously up to a
maximum speed of 86.4 kbits/s.
But Qualcomm professes confidence despite the new competition. Johan
Lodenius, vice president of marketing for the ASIC group, said that
Qualcomm's long-term expertise in code-spreading algorithms, rake receivers
and demod concepts would allow the company to remain several steps ahead
of the newer CDMA licensees. Qualcomm also released separate transmit
and receive devices for IF, named IFR3000 and IFT3000, with direct
interfaces to the MSM3000. In addition to downconverters and upconverters
in the respective chips, both have wideband data converters to move CDMA
and FM analog baseband to digital baseband and back again. While
Qualcomm plans to stick primarily with those elements of the design that can
be implemented in standard CMOS, ASIC products group president Don
Schrock said he expected that dividing line to be a moving target. More true
RF functions will be absorbed in future generations of the IFx family, he said,
perhaps placing Qualcomm in all chip segments of a design outside power
amps and LNAs.
Larger licensees of CDMA, including LSI Logic Corp. and DSP
Communications Inc., are still playing catch-up, but LSI Logic in particular
can apply GSM lessons to CDMA and eventually to 3G phones. It was one
of the first to add embedded RISC to its baseband DSP-in the case of GSM,
using TinyRisc instead of the industry-favored ARM. Core data-converter
and mixer blocks will be similar in GSM, CDMA and 3G, and LSI Logic will
help customers in the RF/IF domain by developing board-level radio modules
for each air interface.
Lucent Microelectronics (Berkeley Heights, N.J.) is taking a more
opportunistic approach to CDMA for now, developing advanced codecs
using the Enhanced Variable-Rate Codec algorithms in order to replace
Qualcomm voice-codec blocks in some designs. Lucent also is looking at
developing special demod blocks for CDMA, before eventually using the
DSP1600 architecture to go directly against Qualcomm's MSM.
Broad mandate
Qualcomm's latest and smallest licensee has the broadest mandate of all.
PrairieComm Inc. (Arlington Heights, Ill.) intends to offer standard products
in the CDMA realm and synthesizable cores and other elements in the GSM
space that can be licensed. Standard baseband chip sets are mandated in the
CDMA market, said president John Diehl, since competitors are fewer and
system designers need more help. PrairieComm will bypass GSM handset
chip sets to work solely on GSM base station controllers in conjunction with
Motorola's Semiconductor Products Sector. PrairieComm will license
"Inner-G" firmware to Motorola for use with the latter's 56300 DSPs, which
are used in parallel banks for programmable basestations.
In fact, the rollout of programmable "soft" basestations capable of handling
multiple frequency bands and multiple air-interface modes helps drive the sale
of high-end parallel DSPs. And that trend will accelerate as basestation
controllers have to support 3G services. Only high-end architectures like the
TI TMS320C6x, the Lucent 16000 Sabre, the Analog Devices Sharc and
the Motorola 56300 can play in this game. Bob Rango, general manager of
modem and multimedia products at Lucent, said that a common core of
multiple banks of Sabres will form the heart of future soft basestation
controllers and central-office equipment that performs access concentration
for wire-line services.
But does the work on GSM and CDMA have a practical impact on 3G?
Brian Rodrigues, marketing manager for GSM products at LSI Logic, said
that design disciplines that emphasize programmability, downloadable
algorithms and a move to single-chip baseband/IF systems will be universally
applicable in GSM, CDMA and future 3G systems, whatever they look like.
LSI's CoreWare design methodology emphasizes reuse of modules stemming
from TinyRisc to MIPS. Nevertheless, Rockwell, which uses a more
partitioned approach for its baseband-to-RF designs, sees core designs as
helping to feed 3G concepts. For example, a member of the Rockwell set
called the RF136 transceiver combines an IF receiver, variable-gain amp,
image-reject mixer, VCO operating up to 1 GHz and a transmitter I/Q
modulator. Product manager Reiner Klement said that this partitioning was
chosen to allow one transceiver block to handle multiple RF chains for the
different frequencies GSM systems accommodate.
For the near future, Rockwell will use this partitioning to support the three
frequency bands of GSM: the original 900 MHz, the DCS 1800 system and
the North American upbanded 1.9-GHz system. But in the future, Klement
said, "this same concept could help simplify handling multiple RF paths in 3G
designs."
Qualcomm still must swallow a bit of pride if it is to adopt the NTT DoCoMo
version of W-CDMA in order to play in markets where that standard is
mandated for 3G. But Lodenius and Schrock say that the important thing is to
design for packet-data processing and multiple-channel processing today.
Even if wideband data is still a capability looking for a killer application, its
potential legitimizes developing such designs now, whether for IS-95B or true
3G. Wallace of Nokia suggests that the standards-body experts working on
IMT-2000 may discover that, as a practical matter, a move to a system
dubbed 3G may be less important than the incremental adding of features to
existing digital cellular and PCS systems. From an end-user perspective, this
may mean that developers should stop looking for the mysterious killer app in
wideband data. From a baseband design perspective, there may never be a
unified air interface applicable the world over, making it more important to
work on removable SIM modules for handsets and on global "beacon"
systems for basestations than on a one-size-fits-all DSP/RF chip set.
"Arguing over an air interface may be like arguing over the type of copper
wire you use for a wire-line service," Wallace said. "Look at the ETSI
decision specifying both W-CDMA and TD-CDMA and you realize there
are sound technical reasons for considering both. It wasn't just a
two-hump-camel type of compromise. What this suggests is at the end of the
day, there will be no winner-take-all for 3G."
Copyright (c) 1998 CMP Media Inc. |
