Barcelona, Spain -- With deployment of third-generation networks and the GSM Association's High-Speed Downlink Packet Access (HSDPA) technology just getting under way, mobile-network operators are being forced to respond to an alternative technology offering direct mobile Internet access: WiMax.

As evidenced at the 3GSM World Congress here last week, chip designers, network equipment vendors, makers of test-and-measurement gear and network operators are already plotting the next stage of the wideband-CDMA (W-CDMA)-based air interface. In some quarters, that next phase is referred to as 3G Plus or, incorrectly, as 4G, but the standards setters at the Third Generation Partnership Project (3GPP) have dubbed it Long-Term Evolution (LTE).

During a keynote session at the 3GSM World Congress, one spectator posed a question that went to the heart of the matter: Wasn't WiMax the "the elephant in the room" at the conclave?

Arun Sarin, CEO of U.K.-based Vodafone Group, responded that while WiMax "is an interesting technology," it is "not really ready for prime time the way that we build our networks today. . . . To get all puffy and excited about it now is too much, too soon." HSDPA, the GSM Association's newest high-speed upgrade, is on schedule to compete with WiMax, he said.

GSM Association chairman Craig Ehrlich added that GSMA is "gearing up on our evolutionary path" to HSDPA and LTE.

Nevertheless, operators realize that the time to act is now. "We need to move faster" along the envisioned evolutionary route, Sarin said, "or the other guys will eat our lunch."

The 3GPP is not alone in pushing toward mobile broadband at data rates of a theoretical 100 Mbits/second for downlink and 50 Mbits/s for uplink--well beyond what HSPA (the combination of HSDPA and its uplink counterpart, HSUPA) will be able to offer in a year or so. Indeed, the pace at which the standards body is pushing LTE--with initial specifications expected this September--is driven by challenges from both mobile WiMax and a parallel effort by the CDMA community called ultramobile broadband.

LTE deployment is expected between late 2009 and 2011. The motive behind the LTE push is clear, said Kaivan Karimi, director of corporate strategy for the wireless market at Freescale Semiconductor Inc. "LTE will, once and for all, draw the line" between the evolutionary path of the existing network infrastructure owners and the WiMax-based mobile broadband Internet connectivity promoted by new entrants.

Where's the value?
Some in the industry remain ambivalent about LTE. "What's in it for consumers?" asked Marty Schroeder, associate vice president and general manager at Marvell. "My concern with LTE is what its value is to them, and at what price it will be offered." Sanjay Jha, COO of Qualcomm, put it another way: "What is the fundamental business model for this among the operators? Could HSPA be enough?"

Advances in LTE include the use of orthogonal frequency-division multiplexing (OFDM) transmission with frequency domain adaptation, along with multiple-input, multiple-output (MIMO) smart-antenna technologies.

Other alternatives use the same techniques, however. And chances are slim that the industry will coalesce around a single air interface for emerging all-Internet Protocol networks, technology suppliers caution.

Unlike W-CDMA's single 5-MHz carrier transmission scheme, OFDM supports spectrum allocations ranging from 1.25 to 20 MHz, varying by subcarrier. And because OFDM supports both time-division duplex and frequency-division duplex access, operation is possible in either paired or unpaired spectrum.

While OFDM's 20-MHz capability is significant from a technology standpoint, it could create a regulatory bottleneck, since network providers will need to secure additional spectrum.

LTE is targeted at allowing operators spectrum flexibility and efficiency, for a smooth migration path into legacy bands and low latency. Japanese carrier NTT Docomo is leading the way toward that evolution--just as it spearheaded the shift from 2G--with imminent deployments of both HSDPA and HSPA. In July, Docomo started requesting bids from infrastructure suppliers and handset manufacturers for LTE, which it calls Super 3G.

Six other global mobile operators joined Docomo in forming the Next Generation Mobile Networks initiative last summer to lobby collectively for "a coherent view of what the operator community is going to require in the decade beyond 2010." The co-founders are Vodafone, Sprint-Nextel, T-Mobile International, Orange SA, KPN Mobile and China Mobile. Last week, 12 more companies joined the initiative. Half are carriers, and the remainder are big players from the equipment makers' side: Samsung, Huawei Technologies, LG Electronics, NEC Corp., Starent Networks and ZTE Corp.

The group will also press for an "improved" licensing regime for whatever comes after HSPA and 1xEV-DO (Evolution-Data Optimized). They want next- generation technology with reasonable, nondiscriminatory intellectual-property rights, leading to more-predictable fees from IP-rights holders. A group within the European Telecommunications Standards Institute has been reviewing ways to ensure lower royalties. Still, many technology suppliers dread the emergence of an IP-rights logjam around OFDM, given its status as an underlying technology for both LTE and WiMax.

"A patent arms race has already begun," said Marvell's Schroeder.

Technology challenges
Many senior executives and technologists interviewed agreed that multiple antenna techniques under the MIMO umbrella are key to success, as is adoption of a hierarchical network of basestations. Optimizing the air interface for multiple users demanding different data applications will also be important.

Ericsson, the biggest supplier of wireless infrastructure equipment, demonstrated an LTE test bed at the 3GSM event, showing live transmission of data at more than 100 Mbits/s. Hakan Djuphammar, vice president of system architectures, said the FPGA used in the demo was based on lower-layer specifications frozen in the standard over a year ago. Those will be converted to ASICs by the time Ericsson starts interoperability testing, expected in late 2008, he said.

Other companies demonstrating LTE test beds at the event included Alcatel-Lucent and NEC.

"We have made significant contributions to the standards-setting process and possibly have the largest amount of [IP rights] in LTE," Qualcomm's Jha said. "Thus, we have been able to simulate much of the specifications in the emerging standard. We have used this model in four generations of development for W-CDMA and CDMA."

The next step is to develop an FPGA- and DSP-based platform that will then be ported to an ASIC, said Jha.

"All this feels to me like UMTS [the Universal Mobile Telecommunications System] and W-CDMA did in 2001 and 2002. If [LTE] proceeds in the same way, we are looking at the second half of 2010 as the most realistic time for commencement of deployment," said Jha.

Freescale's Karimi claimed his company is "best positioned" to lead the LTE charge. The five key ingredients for success, he said, are a "baseband, A/D and D/A wideband converter, integrated front-end transceivers, software chain and power amp efficiency closely linked to RF." Freescale, Karimi claimed, is "the only chip company today active in all five areas."

In developing an OFDM-based modem for LTE, Freescale is working on microcoded engines, rather than a software solution, for both the media-access control and physical layers. Freescale will leverage OFDM IP acquired from CommASIC in 2005, as well as IP from another, undisclosed company.

Bill Krenik, manager of advanced architectures at Texas Instruments Inc., estimated a need for more than 1 million gates of logic in the baseband portion to implement LTE. "But we can already envision very powerful DSPs to cope with this type of complexity," he said. Over time, LTE could enable development of software-defined radio, he said, though initially "we need really sophisticated digitally controlled radios that in the strict sense of the definition will not be SDR."

Jim Tran, vice president of Broadcom Corp.'s mobile communications business unit, said operators will want to see some return from big investments in HSDPA and HSPA before committing to another upgrade. "So I feel we are two and a half to three years away from any kind of volume deployment" of LTE, he said.

Rupert Baines, vice president of marketing at picoChip, took issue with the "evolution" part of LTE. "For some carriers," he said, "the shift to MIMO and OFDM will be a significant step."

Baines said picoChip designed its 203- series multicore DSPs for OFDM operation. It plans a programmable platform for mid-2008 based on a version that will target LTE, mobile WiMax and MIMO. -- Additional reporting by Junko Yoshida

 See related image