The rollout of 3G (third-generation) wireless data networks began in Japan as early as 2001, but the first 3G network didn’t appear in the United States until 2002. Major rollouts began here a few years later. With the release of Apple’s iPhone 3G in July, consumer expectations about media availability and access began running high. The appearance of competitive 3G phones, such as Google’s open-standard, no-license-fee, OS-based Android platform, will likely drive those expectations even higher (see sidebar “3G handsets”). But is the entire network infrastructure ready for simultaneous data download and upload traffic by even half of the world’s 3 billion cell-phone users if they all decide to attempt social networking, mobile-video transactions, or both at the same time? Even though new 4G (fourth-generation) standards are in the offing for improving latency and other problems in cell-access networks, what about the backhaul networks, head ends, and network cores?

Whether 3G networks are well under way or just beginning depends on your geographic location, how you define the set of features, and which part you are considering: network infrastructure, data services, or handsets. The definitions of 3 and 4G-network technologies and services and the boundaries between generations have shifted over the last few years (see sidebar “3 and 4G definitions and technologies”). To some extent, this situation is happening because 3G-network buildouts, handset development, and data services did not become available together in lock step. Originally, 3G meant data services—not just texting, e-mail, and IM (instant messaging), but also full-fledged Web browsing and data downloads and uploads over high-speed Internet-broadband links, as well as video-telephone calls. Yet, current 3G networks and services have been unable to deliver sustained data rates high enough to support video of 10 Mbps or more.

Meanwhile, even more air-interface standards have proliferated, especially if you count shorter-range wireless protocols, such as Wi-Fi and Bluetooth. A not-uncommon view of what 4G mobile devices will accomplish is the ability to handle several air interfaces from multiple base stations and negotiate with them more or less simultaneously while processing multiple types of data and services switching in real time. This ability was one early vision of 3G.

The ITU (International Telecommunications Union) has defined 3G wireless communications under the IMT-2000 standard and named five qualifying air interfaces: W-CDMA (wideband-code-division multiple access), CDMA2000, TD-CDMA/TD-SCDMA (time-division CDMA/time-division-synchronous CDMA), EDGE (enhanced-data-rates-for-global evolution), and DECT (digital-enhanced-cordless telecommunications). Last year, the ITU added mobile WiMax (worldwide interoperability for microwave access) as a sixth. Nearly every major operator has 3G today, says Allen Nogee, In-Stat’s principal analyst for wireless technology and infrastructure.

Yet, according to a new In-Stat study, says senior analyst Gemma Tedesco, through 2012, most cell subscriptions will be for GSM (global-system-for-mobile) communications, a 2G technology, and the 2.5G GPRS (general-packet-radio service) (Figure 1 and Reference 1). “Even CDMA2000 is being used only for texting and voice, not for data,” she says. HSPA (high-speed packet access) has grown in Western Europe, but, even in 2013, there will still be huge amounts of GPRS and EDGE. Most operators in the world haven’t built out 3G networks. Where they have, such as in Asia and most of South America, not all cell users have 3G-capable devices. In the United States, most operators have 3G networks, and lots of users have 3G devices without paying for 3G data service. “So, 3G is not really here yet around the world,” she says.

Meanwhile, according to recent In-Stat research, although carriers rolled out many new 3G networks worldwide in 2007, overall shipments of cell base stations in 2008 declined considerably (Reference 2). “New 3G networks are continuing to be deployed,” says In-Stat’s Nogee in a press release. “However, the worldwide economy has been faltering, subscriber-GSM growth—even in fast-growing developing areas—is starting to slow, and wireless broadband use, while growing, is not growing fast enough for operators to spawn continued base-station growth.”

Many industry observers talk about nimbleness and flexibility, so that cell phones and other access devices, such as notebook PCs, will make the most effective connection possible, given the application, the location, and the available services and networks. The picture that emerges is a patchwork of multiple broadband-wireless-access technologies and services. To some extent, this image also applies to the backhaul.

Published discussions of the increased data traffic expected in 3 and 4G networks rarely mention the effect on backhaul networks and the network core. Yet, equipment and technology in this area are at least as vital to the network as access and terminal-device technologies. Carriers have a variety of choices to consider, as well as a number of issues to manage for increased data traffic, especially as that traffic begins to move onto higher-frequency 4G networks.

As demands on the backhaul and network core increase because of more data-centric services, carriers are shifting from a leased-line model with T1/E1 lines to deploying optical or fiber technology right up to the base station to carry extra capacity, says Jagdish Rebello, iSuppli’s director and principal analyst of wireless communications. Instead of taking this traffic back to the base-station controller or switching station, equipment manufacturers are putting more intelligence into the base station. This approach will reduce traffic in the core and reduce latency, especially important for social networking, gaming, and push-to-talk applications. According to Greg Waters, executive vice president and general manager for front-end products for Skyworks Solutions, many operators are focusing on next-generation cell-base-station technology, such as femtocells, that goes beyond 3G (Figure 2).

Fulcrum Microsystems’ OEM customers require low latency for voice traffic, synchronous Ethernet, and mechanisms for congestion management of voice, video, and data—that is, QOS (quality of service), says Gary Lee, director of product marketing. Voice traffic requires maximum-latency guarantees, and video needs minimum-bandwidth guarantees and somewhat-lower latency, making it difficult to combine them on the same network. Solving QOS issues by throwing extra bandwidth at the problem may have worked in the early cell-network days, he explains, “but now we have thousands of video streams, so QOS mechanisms must be built-in.” In silicon, this situation calls for network processors, multicore CPUs, and switch fabrics that can provide minimum-bandwidth guarantees and congestion-management features.

“There’s been a significant increase in data utilization in mobile networks, but they aren’t necessarily designed to handle QOS,” says Todd Mersch, Continuous Computing’s senior product-line manager. The problems iPhone 3G users experienced when millions of them first attempted to access AT&T’s wireless network exemplify this issue. Deep packet inspection, deployed on wire-line networks to prioritize certain types of traffic for QOS, is also seeing more use on mobile networks.

If carriers are to leverage cost-effective IP (Internet Protocol) networks to handle the increased bandwidth expected from 3 and 4G services without undue packet loss and call dropping during handoff to the base station, the most critical issue is time synchronization, says Sameer Vuyyuru, vice president of marketing for Semtech’s advanced-communications division. One mobile-wireless-network-synchronization scheme, FDD (frequency-division duplex), uses the same technology as wired networks. The other, TDD (time-division duplex), requires finely tuned frequency accuracy and phase alignments among all base stations in the cell network. The current GPS-synchronization system, comprising a master GPS clock and hundreds of GPS slaves, works inconsistently inside buildings and can be expensive to implement. Competing methods that transmit clock data over packet-switched networks include the IEEE 1588 Version 2 Precision Time Protocol standard for network-based timing and synchronization, which the IEEE finalized last March. “The master still needs a GPS module, but now you’re amortizing that cost over hundreds of less-expensive slaves,” says Vuyyuru.

One backhaul alternative may be the use of Wi-Fi in the 5.8-MHz band as a separate radio structured mesh network with intelligent nodes and dynamic channel management, says Francis daCosta, founder and chief technology officer of Mesh Dynamics. “The regular 2.4-MHz Wi-Fi band can get pretty crowded,” he says. “The 5.8-MHz band or any other used solely for this purpose can act like a diamond lane that removes a lot of frequency-interference and congestion issues.” In highly congested areas, even if a carrier offloads only some of the data to a complementary network before it hits cell towers, it will improve cell service for all (Figure 3). Municipal Wi-Fi failed for multiple reasons, but commercial mesh networks remove one of its main obstacles, which was competition with cell carriers (Reference 3). “Carriers can put down a Wi-Fi node where needed, it’s unobtrusive, and there’s no licensing involved,” says daCosta. Dual-mode-cell/Wi-Fi terminals that can take advantage of this situation, such as the iPhone, are becoming more common.

“In networks built from scratch, a key element of the business plan is the high cost of the backhaul,” says John Baker, vice president of technical marketing for Andrew’s wireless-networks-solutions group. Primarily in locations without land-line connections, carriers are deploying microwave-backhaul networks. “Of course, you can also use IP over land lines or over microwave, but, even though its benefits [offer] higher bandwidth and [the lowest] cost, there are timing challenges with synchronizing base stations to the core of the net,” he says.

Many silicon vendors are now offering front-end silicon for WiMax, as well as other access technologies. These chips include RF stages, mixers, ADCs, DACs, and digital filtering. Others are offering baseband SOCs (systems on chips) or embedded multicore processors. Wireless-networking-equipment manufacturers are leveraging multicore chips to create equipment that can keep up with the vast amounts of data that 3G and other technologies require, says Stephen Turnbull, high-performance-embedded-processor-portfolio manager for Freescale Semiconductor.

Designers of several major base stations for 3G W-CDMA networks are using multicore DSPs, says Ramesh Kumar, manager of wireless-base-station-infrastructure products for Texas Instruments’ communications-infrastructure group. As cost pressures rise on 3G networks because of their expansion into developing markets, multicore chips can help increase channel-card densities and maintain relatively lower cost and power consumption. “In the past, each channel card performed L1/L2 processing for one cell, or base station,” he says. “Now, we’re looking at three cells per card, moving to six per card within the next three to four years.”

Yet, the cost of cell sites has stayed about the same, in part because the electronics cost has come down year over year. “Electronics are becoming less of a percentage of base-station cost because of increased integration and terrific volume growth,” says Andrew’s Baker.

By the end of this year, Intel’s Centrino 2 mobile processors will offer optional support for not only Wi-Fi but also WiMax through a combined Wi-Fi/WiMax module, enabling the development of additional dual-mode PCs and other mobile devices. “We expect that customers may elect to use dual-mode devices for the coverage and service of [3G] EVDO (evolution data optimized) and the added boost of WiMax in metro areas, where it is available,” says Kathy Walker, Sprint’s chief information and network officer.