Radio interfaces make the difference in 3G cellular systems

Third-generation cellular telephony is on its way--not, unfortunately, as a single worldwide system, but as three incompatible ones. The main difference between the three lies in their choice of radio interface technology. This fact is crucial for several reasons, since the radio interface determines not only the fundamental capacity of a mobile radio network, but also how it deals with such issues as interference, multipath distortion, and handing off calls from one base station to another as users move around. Consequently, as might be expected, the choice of radio interface has a dramatic effect on the complexity of the system and its cost. Also, global travellers will need more than one phone with which to communicate, at least until trimode phones reach the market.

To understand what is being developed, and why, let's begin with one of the stated goals of third-generation (3G) systems, namely to support variable user data rates as high as 2 Mb/s. In one way or another, all three approaches provide for adaptive bandwidth-on-demand. Two of the systems use wideband code-division multiple access (WCDMA) for the radio interface. The other (of which more later) uses two variations of time-division multiple access (TDMA).

With WCDMA, a user's information bits are spread over an artificially broadened bandwidth. The job is done by multiplying them with a pseudorandom bit stream running several times as fast. The bits in the pseudorandom bit stream are referred to as chips, so the stream is known as a chipping, or spreading, code. It increases the bit-rate of the signal (and the amount of bandwidth it occupies) by a ratio known as the spreading factor, namely, the ratio of the chip rate to the original information rate.

The key devices in any CDMA system are its correlation receivers, which store exact copies of all of the system's chipping codes. These codes the receivers use to multiply a received data stream, selecting the same chipping code as was used in the transmitter. The devices also perform whatever other mathematical operations are needed to restore the original user data. The result is that at the receiver output, the amplitude of the de-spread signal is increased by the spreading factor relative to interfering signals. In the process, those interfering signals are diminished and simply add to the background noise level. In other words, correlation detection uses the spreading factor to raise the desired user signal from the interference. The effect is called processing gain.

Note that spectrum spreading by itself does not confer any benefits. It is the combination of spreading and de-spreading that works the magic in CDMA, allowing all the base stations in a network to use the same carrier frequency because every conversation on it is assigned a separate spreading code. The scheme can also resolve different propagation paths, turning multipath distortion from a destructive nuisance into a helpful ally.

To help understand this complex idea, remember that frequency-division multiple-access systems keep conversations from interfering with each other by assigning them to different frequency bands, whereas time-division systems do so by assigning them to different time slots. With CDMA, however, the conversations occupy the same frequency bands at the same time. But each interaction is multiplied by a different chipping code, and when the signals are de-spread, the only one that comes through intelligibly is the one whose code was used by the de-spreader. The others, as stated above, simply add to the background noise level (which ultimately limits the number of users that can share a channel).

For the system to work, two factors are key. First, only soft handovers may be employed, since with them mobile terminals can maintain simultaneous connections to different base stations as they move among them. Second, transmitter powers must be strictly controlled so that signals from all mobile terminals arrive at the base station with about the same strength, despite their differing distances from the base station. Strict power control is maintained with multiple real-time power control channels, plus control loops with different resolutions (coarse and fine). The control channels operate at power command rates between 800 Hz and 1.5 kHz. That is, base station equipment measures the power received from each mobile unit as much as 1500 times a second and issues commands to the mobiles at that rate to raise or lower their output power.

Many users can be accommodated. The maximum WCDMA chip rate is 3.84 megachips per second (Mch/s) and yields a modulated carrier about 5 MHz wide. System operators can deploy multiple carriers, each of which occupies 5 MHz. Moreover, in a WCDMA system, multiple end-users can share each 5-MHz channel.

Photo: John Boyd

[Fig. 1] Unlike second-generation narrowband code-division multiple-access systems, wideband CDMA, or WCDMA, allows network operators to dynamically reassign channel bandwidth [vertical axis] in response to user needs. These bandwidth allocations are updated every 10 ms, as shown. The bandwidth allocations in each of a system's channels can be changed independently of one another.
Click on the image for a larger view.

In Fig. 1, for example, five users share a single channel. Three operate at fixed data rates, while the data rates of the remaining two are variable. The WCDMA system deals with this situation by continually changing the way it distributes the channel's bandwidth among the five users, adjusting the spreading factors of each of the users every 10 ms.

In sum, experience with narrowband CDMA (the current IS-95 second-generation systems), not to mention advanced experiments, field trials, and general diligence in wireless practice, has borne fruit. WCDMA is the favored radio interface for 3G in those situations where its increased bandwidth can be tolerated. But wideband options may not suit every situation, which is one reason why 3G will take more than one form. Each 3G option accommodates a specific second-generation predecessor, and uses that legacy protocol as one of its operating modes in 3G [Fig. 2].

[Fig. 2] Three main second-generation cellular technologies dominate the industry today: narrowband CDMA, GSM (a narrowband time-division multiple access, or TDMA, system as far as its radio interface is concerned), and a different narrowband TDMA system known as IS-136. Each has its own path mapped out for migrating to the third generation [right], meaning that there will be three incompatible versions.
Click on image for a larger view.

GSM adopts a CDMA interface

The community abiding by the Global System for Mobile Communications (GSM)--by far the largest and most acronym-happy of the second-generation cellular variants--extends itself into 3G under the general rubric of Universal Mobile Telecommunications Services (UMTS). Their radio interface uses WCDMA radio techniques and is called UMTS Terrestrial Radio Access (UTRA).

WCDMA has two forms, distinguished by how they separate the two directions of communication. Frequency-division duplexing (FDD) employs separate uplink and downlink frequency bands with a constant frequency offset between them. The other form, time-division duplexing (TDD), puts the uplink and the downlink in the same band, and then time-shares transmissions in each direction. This mode may be useful for indoor applications or for operators with spectrum restrictions. The FDD version of UTRA is so well known that discussions about any of the other 3G radio interfaces are usually framed as comparisons with it.

The WCDMA physical layer includes the variable-bit-rate transport channels required for bandwidth-on-demand user applications. These can multiplex several services onto a single connection between the fixed infrastructure and a mobile terminal. Some of the physical channels do not carry transport channels; in fact, they do not carry user information of any kind. They serve the physical layer itself, and include such resources as some pilot channels (that assist in modulation recovery), a synchronization channel (that lets mobile terminals synchronize to the network), and an acquisition channel (that establishes initial connections to mobile terminals).

WCDMA resembles all CDMA systems currently deployed in that it applies the spreading function in two phases. An initial channelization code spreading is followed by a scrambling code spreading. The initial channelization code spreading alone determines the occupied bandwidth of the radio signal. As for the scrambling code, it is used to distinguish different mobile terminals at the base station's receiver and to distinguish multiple cell sites in the mobile terminal's receiver.

The second-generation IS-95 CDMA system uses a single pseudonoise code common to all base stations, but applied by each base station with a different time offset. Consequently, the uniqueness of a base station lies not in the code itself but rather in its time offset. The CDMA mobile, then, sees several base stations just as it would multipath products from a single base station, and it can resolve each of them in a bank of correlation receivers. Different user data streams from the base station are separated by one of 64 channelization codes (known as Walsh codes) before the final scrambling spread.

WCDMA elaborates on this scheme to allow for multiple connections to a single mobile terminal as well as variable spreading factors at the channelization spreading stage. Low user data rates get lots of coding gain with high spreading ratios, while high user data rates get less coding gain because of their lower spreading ratios.

The spreading details differ in the downlink, or base-to-mobile, and the uplink, or mobile-to-base, directions. In the downlink direction, the channelization codes separate different users (mobile terminals) in a cell. In the uplink direction, they separate different physical channels (parallel connections) in a single mobile--for example, they distinguish user data from signaling. The WCDMA channelization codes have a variable length of 4 to 512 chips in the downlink and 4 to 256 chips in the uplink. All the spreading occurs on 10-ms frames at a constant chip rate of 3.84 Mch/s.

The scrambling spreads also differ. In the downlink, the spread is performed with any one of 512 different 38 400-chip codes on a 10-ms frame (3.84 Mch/s). The uplink from the mobile terminal has two possibilities. One possibility is to use any of millions of different 38 400-chip scrambling codes. Another is a family of shorter 256-chip codes within a 66.7-µs frame (also 3.84 Mch/s), but this possibility is reserved for advanced base station receiver designs that use a technique called multiuser detection.

To digress a little, multiuser detection can use a bank of correlation receivers, which bank is also called a rake receiver, because its block diagram resembles a garden rake. Broadly speaking, rake views the contribution of radio emissions from other terminals in a cell as a special kind of interference, which can be recognized as coming from other mobile terminals in the same cell because of the latters' use of known spreading codes. The detection schemes attempt to make symbol decisions based on the whole received waveform from all the interfering users; they de-spread each of the emissions from mobiles and remove their interfering influence directly. Multiuser detection is a highly specialized subject far beyond the scope of this article [for more information, see To Probe Further].

Wireless standards normally seek to minimize power consumption in the mobile unit when specifying uplink modulation schemes. Obviously, the longer the battery can go between charges, the better. Chief among the techniques used to prolong battery life are measures that reduce the peak-to-average power ratio in the mobile's transmitter. WCDMA specifications call for a complex scrambling process immediately following the channelization and scrambling spreads.

To go into more detail, the mobile terminal transmits two parallel channels: a data channel and a control channel. The complex scrambling codes are carefully fashioned so that the phase rotations during a symbol period never exceed 90 degrees. This technique causes the modulation to (so to speak) look like a rotating quadrature phase-shift-keying signal of constant level, regardless of the differences between the data and signal channel power level settings.

CDMA2000--the multicarrier mode

Besides UTRA, one other technology in the new generation is based on CDMA--multicarrier (MC) mode CDMA, usually called cdma2000. This mode is intended to provide 3G services over mobile radio networks that conform to the ANSI document "TIA/EIA-41," which includes existing (2G) IS-95 and TDMA (IS-136) nets. In the future, it is expected that the MC mode will be extended to allow connection to core networks based on GSM's Mobile Application Part (MAP).

The multicarrier mode is very similar to the frequency-division multiplexing form of WCDMA. The dissimilarities stem from the need to allow the IS-95 mode to work in the enhanced 3G network just as GSM handsets can be accommodated in the UTRA extension. The chief differences are: a 20-ms framing structure instead of 10 ms, and a slightly different spreading rate--3.6864 Mch/s, which is exactly three times the IS-95 rate of 1.2288 Mch/s. The legacy IS-95 mode is one operating mode in the multicarrier mode. Multiple parallel connections can be established in up to three CDMA carriers, in what is called the 3X operating mode.

UWC-136--the TDMA modes

How does an operator with limited spectrum deploy third-generation services? This is an important question because personal communications services (PCS) in North America are already deployed in the part of the spectrum reserved for 3G. Since some of the spectrum allocations are as narrow as 5 MHz, an alternative to the CDMA schemes with very high spectral efficiency is attractive to PCS operators who want to enhance their networks with 3G services. The TDMA networks (ANSI-136), which are largely confined to the Western Hemisphere, live on as one of the operating modes in a family of radio techniques that employ no CDMA techniques at all--the whole family of TDMA operating modes designated by the Universal Wireless Communication Consortium as UWC-136.

One of the modes, 136HS Outdoor (also called EGPRS-136), is almost identical to the GSM packet radio scheme called Enhanced Data rate for GSM Evolution (EDGE). Packet radio techniques are coupled with adaptive modulation--Gaussian minimum-shift keying (GMSK) and eight-phase phase-shift keying (8-PSK)--to give EDGE all of the 3G features except for its 2-Mb/s indoor data rate. With unsynchronized base stations, a second-generation GSM network can be upgraded to EDGE within only 2.4 MHz of spectrum. Synchronizing the base stations reduces the spectrum requirement to only 600 kHz. Synchronization here refers to a means for preventing base stations from transmitting while a designated beacon base station is signaling its assigned mobile terminals.

The TDMA and GSM communities enjoy a common radio interface in EDGE for outdoor and typical mobile applications. Indoor office applications up to 2 Mb/s are accommodated by two radio techniques: the TDD mode in the WCDMA interface on the GSM side, and 136HS Indoor on the TDMA side. 136HS Indoor is sometimes called WTDMA, being a TDMA-based concept employing 1.6-MHz carriers with two types of bursts: a sixteenth and a sixty-fourth of a 4.615-ms TDMA frame. The high user data rates are accommodated in the 1/16-length burst, while the intermediate user rates can be accommodated in the 1/64-length burst. Adaptive modulation in both kinds of bursts lets 136HS Indoor adapt itself to a wide variety of user applications, just as all the new-generation systems do.

Packet radio networks

Cellular radio systems are deployed on only two kinds of incompatible core (wired) networks: those based on ANSI-41 and GSM-MAP. These standards refer to extensions to Signaling System No. 7 (SS7)--the packet network that supports signaling within the worldwide public switched telephone network. The extensions are required to support the mobility of wireless devices. The 3G networks add a new packet-switched network to the existing circuit-switched infrastructure.

The GSM community is making the transition to 3G in three phases. In the first phase, a packet-switched network is added to GSM while the base stations and their controllers are retained virtually unchanged. Of course, to use the General Packet Radio Service (GPRS), as it is called, a subscriber must get a new GPRS-capable handset. The second phase (UMTS) replaces the base stations and their controllers with a new sub-network, the UTRA network. In the third phase, transition to UMTS is completed with the introduction of UMTS handsets and their corresponding UMTS Subscriber Identity Modules [Fig. 3].

>[Fig. 3] GSM [top diagram] is making the transition to the next generation [bottom diagram] in three steps, the first of which is already happening. In the first step [middle diagram], called General Packet Radio Service (GPRS), a packet radio network is added to the GSM system. In the second, the older base stations and their controllers are replaced by the UMTS terrestrial radio access network (UTRAN). In step three, the transition to third-generation Universal Mobile Telecommunications Services (UMTS) is completed with the addition of new handsets, along with new UMTS subscriber identity modules.
Click on the image for a larger view.

Throughout the transition, as Fig. 3 shows, the various components of the GSM core network remain in place, so that mobile terminals can be located, authenticated, and attached to the network. The GPRS phase allows the Internet to invade the mobile network through routers (the SGSN and the GGSN support nodes [see sidebar, "Annotated Acronyms"]. Note that being connected to the network (transporting packets to a terminal, for instance) is not the same as being attached to the network (in which case the core network also knows the terminal's location and has confirmed its right to use the services).

This distinction takes on new significance in a packet radio network. Many nonvoice applications--Web browsing, for example--can give the appearance of a permanent virtual connection, which may affect charges and how the user perceives the network. Except for the addition of a packet control unit (PCU) in the base station controller and new handsets, no other changes are required. Incidentally, there is also the option of upgrading the base transceiver station to support EDGE (8-PSK) modulation as a step on the way to 3G.

The third and final phase is the installation of the UTRA network (UTRAN), which supports the highly adaptive WCDMA radio technology with its soft handoffs, connections to multiple base stations (Node Bs), and sophisticated power controls. The UTRAN handles all of the WCDMA aspects by itself.

The standardization processes for 3G are moving beyond the national and regional bodies of the second generation and into international venues, which reduce the wasteful, parallel processes of the past. Participation in the meetings is a real burden for the participating companies, which welcome anything that can increase the efficiency of the process. Two important groups--3G Partnership Project (3GPP) and 3GPP2--help with these efforts. The first formulates the GSM extensions, while the second formulates the cdma2000 extension to IS-95. It is often helpful to regard these two groups from their network perspectives: 3GPP are the GSM-MAP supporters and 3GPP2 are the ANSI-41 supporters [again, see sidebar, "Annotated Acronyms"].

Quality of service

Taken together, all the operating modes (frequency- and time-division duplexing, multicarrier, and the TDMA modes) provide plenty of resources for provisioning subscribers with a wide variety of services and applications and accommodating new applications yet to be imagined. Provisioning multimedia services in a wireless network presents an enormous new challenge. How are the services to be offered without hurting the efficiency of the network or diminishing its capacity? First- and second-generation cellular systems provided their circuit-switched resources to subscribers with a fixed quality of service (QoS), defined as the best effort it could muster.

Networks in the latest generation, on the other hand, will assign default quality-of-service profiles to users and their applications. A subscriber's application negotiates a suitable quality-of-service profile with the network, which allots its resources according to the default profile, or with some alternative profile, depending on the load on the network, the propagation conditions, and the quality-of-service profiles authorized by the user's subscription. At the subscriber level, four kinds of traffic can be distinguished: conversational, streaming, interactive, and background.

Conversational traffic is the most familiar type of traffic. Fairly tolerant of errors, conversational and videoconferencing traffic have different throughput requirements, but they demand a constant and rather short end-to-end delay.

Streaming traffic applies to applications that can start processing the traffic for presentation to the user before the whole file is transmitted to the subscriber. It can work within a small range of delays and throughput rates.

Interactive traffic is used by on-line applications, in which a subscriber is allowed to interact with a server of some kind: Web browsing, e-commerce, games, and location-based services. Interactive traffic can work acceptably over an intermediate range of delays and throughput rates.

Finally, background traffic is very tolerant of delays, works within a wide range of throughput rates, but is relatively intolerant of errors. These include applications such as e-mail, short messaging services, and file downloads. Background traffic is the most naturally compatible with packet data networks.

Each of the traffic types will draw on the network's resources in its own way. The network itself will have to deal with an application's priority (high, medium, or low); tolerance for delay (highly tolerant all the way to minimal tolerance); mean and peak throughput (measured, for example, in bits per second); and reliability (acknowledged or unacknowledged exchanges, protected or unprotected, with different bit and word error rate criteria). Different applications imply a variety of terminals and wireless appliances. These imply complicated and highly dynamic subscriptions with changing authorized applications, and accompanying quality-of-service profiles.

Operators have lots of work ahead as they try to optimize their networks and find applications to attract new revenue. All the systems will be content limited: they will have more capacity than applications to fill it. The whole point of 3G is to sell more air time to current subscribers and to attract new subscribers by offering new features and applications of specific interest to mobile users. Mobile Internet connections alone will not do the trick. The need is for services like maps and driving directions, bus schedules and route maps, interactive airline schedules and gate assignments, and automatic toll payments.

Getting such services deployed may be harder than it sounds. The recent battle between the Recording Industry Association of America and the MP3 community over music downloads over the Internet may have been a preview of how difficult it can be to integrate the creative aspects of content into networks. For 3G, the problem is not the technology, but finding the content.