Faster data rates for mobile broadband are being made possible by piling one technology upon another, according to E&T.
When the first 2G all-digital GSM system was launched in 1991 it offered data rates of 14.4kbit/s. Extensions of GSM, with GPRS and multi-slot EDGE technology, pushed up the data rates and became known as 2.5 and 2.75G systems, respectively. The wideband code-division multiple access (WCDMA) system - also known as the universal mobile telecommunications system (UMTS) and cdma2000 - is now described as a 3G technology, the first with data rates high enough to support mobile Web browsing, email, video and mobile TV using the DVB-H standard. High speed packet access (HSPA) extensions to WCDMA in the up and down links, which achieve 3.6 and 7.2Mbit/s data rates, respectively, are now defined as 3.5G systems.
Users of next-generation mobile systems, which are being called 3.75G, 3.9G and even 4G, will benefit from a substantial increase in mobile data rates compared to current approaches, even when moving at high speeds, in cars on a motorway or in a train. This next-generation mobile broadband network will support data rates of up to 100Mbit/s on the move, and up to about 1Gbit/s when used in relatively static situations, such as for local wireless access. These rates will be essential to enable many users to access large downloads at once.
Mobile systems developers have combined three key techniques to achieve the high data rates offered by current mobile broadband connections. They will layer on even more complexity to achieve the Gbit/s rates forecast for future 4G communications systems.
Modulation and more
The first of the techniques used in current 3G systems, known as quadrature amplitude modulation (QAM), has long been used in fixed point-to-point microwave line-of-sight (LoS) communications to extract the greatest data throughput from a limited bandwidth allocation.
Replacing simple binary phase-shift keyed transmissions (in which a data bit is represented by a 180° change in the signal's phase) with QAM transmissions using 16 symbols (16-QAM) quadruples the spectral efficiency, or data capacity of the available spectrum, from 1 to 4 bit/s/Hz. In 16-QAM systems each data-carrying symbol is transmitted as one of 16 possible combinations of amplitude and phase states, so that it represents four information bits. This decouples the achievable data rate from the occupied spectrum (which is governed by the symbol rate), since each symbol sent now represents more than one data bit. The QAM approach can be taken further, although the signal becomes more fragile as the number of symbols used increases. For example, engineers have designed 1024-QAM LoS transmission systems that offer 10 bit/s for every Hz of allocated bandwidth, given relatively noise-free channels.
The second, more recent, innovation is to use multiple-input, multiple-output (MIMO) antenna arrays with two, four or eight separate antennas at both the basestation access point and the mobile terminal. These can be useful in 'rich' multi-path environments, where a terminal receives many versions of a transmitted signal as it is reflected from buildings and moving vehicles.
The first diagram on p69 shows the variance in the time and angle of arrival of the multiple versions of an indoor signal that has been reflected off walls and ceilings. By using multiple receiver antennas it is possible to resolve and collect these components to extract the best overall signal. Typically, there are between five and 25 significant individual components to indoor signals, shown as the stronger red, yellow and light blue signals in the diagram. MIMO techniques add multiple transmission antennas to further improve the system data throughput, dependent on the number of antennas used, as shown in the second diagram.
Researchers have shown that 4X4 MIMO systems can offer 4bit/s transmissions in every Hz of available bandwidth and industrial test-beds have combined MIMO and 16-QAM to deliver 1Gbit/s transmissions over practical wireless channels, including overcoming Doppler shift issues in moving vehicles. The mobile test-bed built by Japanese telecoms giant DoCoMo can transmit 32 HDTV channels at once in a 100MHz bandwidth. So it achieves a real-world spectral efficiency of 10bit/s/Hz, from a theoretical spectral efficiency of 16bit/s/Hz defined by the combination of 4X4 MIMO and 16-QAM techniques.
The test-bed uses modulation and channel-coding schemes that adapt to changing channel conditions, so it can get the most out of the available spectrum. It also layers on the third key technique for high data-rate mobile transmissions - multiple-carrier orthogonal frequency-division multiplexing (OFDM). This converts a single high data-rate transmission into many parallel low data-rate channels, each at a slightly different but non-interfering (hence 'orthogonal') frequency, which makes it easier for the receiver to recover the multiple versions of the transmitted signal generated by reflections.
OFDM transmission and reception has been made possible by the ready avail-ability of fast Fourier transform DSP chip designs and intellectual property cores. Why not stick to increasingly complex QAM approaches, such as the 1024-QAM signal constellations used in microwave LoS systems, to increase data rates? The problem is that cellular wireless systems experience much more interference and noise than systems that beam signals from point to point. So lower-complexity adaptive QAM, in which the number of transmitted symbols changes dependent on the quality of the available radio channel and the signal to noise ratio at the receiver, combined with MIMO and OFDM techniques, now form the heart of the 3G mobile standards in use today.
LTE and beyond
The next step for cellular broadband is the LTE (long-term evolution) standard, which was ratified by ETSI a year ago. LTE is designed to bring together all the current 3G cellular standards, including WCDMA as used in the US by Verizon and as further modified in China (TD-SCDMA). Verizon plans to launch LTE in 35 markets before the end of 2010 and AT&T has also said it will adopt LTE. Forecasts suggest 440 million subscribers could be using LTE by 2015.
LTE will be followed by 'LTE Advanced', for which the standardisation discussions began late last summer. China is pushing to include a TD-LTE variant in the standard, to provide a way forward for its home-grown, and less successful, TD-SCDMA variant of the more widely used WCDMA approach. LTE Advanced will use the techniques described above, and layer on enhancements to minimise interference between cells, in order to squeeze even more data capacity from the available spectrum.
The 'single-user' MIMO antenna techniques described above will be combined into 'networked MIMO' or 'coordinated multi-point' schemes that enable self-organising networks. These networks will be able to coordinate their activities to minimise interference between users in adjacent cells. This will involve the integration of several techniques such as closed-loop transmit diversity (in which the transmitter's behaviour adapts to the channel conditions), and beam-forming (which 'points' the transmitted beam at the receiver). LTE Advanced will also demand more accurate tracking and estimation of the state of the channel for all the radio channels used in the MIMO array. It will also need to be able to synchronise transmissions between cells, to enable more effective interference cancellation. But there should be a worthwhile payback: the third diagram shows the improvement in data-rate performance expected from a networked MIMO system.
Combining QAM, MIMO and OFDM techniques brought us 3G standards. LTE and LTE Advanced may layer on transmit diversity, beam-forming, and coordinated MIMO techniques to increase mobile data rates by three- to five-fold over current single-user MIMO systems. But although these enhancements are technically possible, they will make severe demands on the receiver DSP. The large amount of channel-state and other operational information that has to be shared between transmitter and receiver to optimise the system will eat into the available throughput. Applying all these techniques is also expected to require a 20-fold increase in handset processing capability, within the same power budget that is available to current designs.
In practice, the highest download rates (of 30bit/s/Hz) are likely to only be available close to basestations or access points, where the signal to noise ratio is high. At the cell edge, due to the longer propagation distance and higher noise and interference levels, data rates are likely to be one hundredth to one thousandth of those possible close to the transmitters. So researchers are looking closely at how best to share the available data rates for cell-edge users, to ensure acceptable quality of service across all the users of the LTE Advanced standard.
One approach would be to use more, smaller cells, so that everyone is closer to a transmitter. But this demands an increase in the number of cells, associated basestations, and backhaul connections, which makes for a more costly system. It also increases the power consumption of the overall system, destroying its green credentials.
In addition to this,the femtocells used to build such dense 4G networks will have to be much more complex than today's 3G offerings, incorporating closed-loop transmit diversity, beam-forming, accurate channel-state tracking or estimation and cell synchronisation to enable the interference cancellation necessary between the larger number of cells in use.
Analogue TV spectrum or WiMax alternative?
Another approach to increasing mobile bandwidth, in the UK at least, would be to redevelop the spectrum that will be released when analogue TV transmissions end. In comparison to current cellular spectrum, the analogue TV spectral allocation is enormous. The TV spectrum is also at a lower frequency than cellular transmissions, with radio propagation characteristics that would make it easy to build large cells in rural areas.
The other current concern for mobile operators and equipment makers is whether they should back LTE, an evolution of current cellular standards, or shift to the mobile WiMax standard. WiMax, as promoted by the WiMax Forum, is emerging as the mobile data connection technology of choice for computing devices such as notebooks and netbooks. Although 2.75G EDGE and fixed WiMax basestations have quite modest installation costs, 3G UMTS, mobile WiMax and LTE basestations are more expensive to build and install, due to the sophisticated DSPs necessary to achieve the multi-user, high data-rate broadband capability.
Given the separate investments made to date by each of the mobile operators in cellular basestations, a major roll-out of WiMax to compete with existing and upgraded cellular systems seems unlikely. However, in places that lack any mobile infrastructure, WiMax may be an attractive option, although it will require a handset design that won't work with LTE systems. It may be, though, that future software-defined radios, using programmable multi-mode receiver designs, could make it possible to build handsets that offer WiMax and LTE connections from one chipset.
Professor Peter Grant is a director of the Mobile VCE and a co-investigator on the China Science Bridges Award. He holds the Regius chair of engineering at the University of Edinburgh and is a Fellow of the IET.
Find out more
Communications: telecoms, fixed, mobile and satellite [new window]