How do we cut the energy consumption of mobile basestations? We follow the researchers around the world who are working on it.
What are we going to do about mobile basestations, the most energy intensive part of a 3G cellular network?
The UK is already committed to reducing the amount of CO2 it emits in 2050 to 20 per cent of the levels seen in 1990. Although mobile cellular communication accounts for only 2-3 per cent of these emissions, we're feeing enormous growth in mobile subscriber numbers and in the amount that subscribers use their devices. Further analysis shows that, over the 10 to 15-year lifetime of a radio basestation, its major contribution to emissions is from operating energy, rather than, say, manufacture or installation.
Let's look at some numbers. A typical 3G basestation uses about 500W of input power to produce about 40W of output RF power. This makes the average annual energy consumption of a 3G basestation around 4.5MWh (an improvement on the consumption of GSM basestation designs). A 3G mobile network, such as those installed in the UK, with 12,000 basestations will therefore consume more than 50GWh a year. This causes a large amount of CO2 emission as well as contributing to the network's operating costs. Network energy use may be much greater in developing markets with their large geographic areas and large populations. China Mobile already has 580 million mobile subscribers, serviced by 500,000 GSM and 200,000 3G CDMA basestations.
Traffic on HSDPA and other data networks is growing exponentially (See E&T vol 5 #13), due to the increased use of smartphones and dongles. This growth, which has been up to 400 per cent per year, is far outstripping operator revenue growth, which trails at 23 per cent per year. Servicing this traffic growth implies rapidly increasing network energy consumption as the basestation infrastructure grows. This means there is an urgent need to reduce the energy used per data bit carried, to minimise the operating power per cell. 'Green radio' research being carried out within Mobile VCE, and elsewhere, is trying to develop more energy-efficient mobile networks that can also accommodate exponential traffic growth.
There are two types of client that need more efficient mobile networks. In the industrialised world, the infrastructure is in place and the markets are saturated in terms of subscriber numbers, but traffic is growing exponentially and so quality of service is becoming a key issue. The main drive from the operators is to cut costs. Things are different in the rapidly growing emerging markets: there's less established infrastructure, often large geographical areas to cover, uncertain mains power and so a lot of interest in using solar and other renewable energy sources. Vodafone, for example, uses one million litres of diesel per day to power its remote basestations worldwide.
One of the key issues in green basestation design is the trade-off between cell size and achievable data rates. When basestations are being used to create physically large macrocells, the signal to interference and noise ratio (SINR) close to the basestation is relatively high, which allows the use of advanced modulation schemes, such as 64-way quadrature amplitude modulation (QAM), to achieve relatively high data rates. However, due to signal propagation loss, SINR is reduced at the more distant cell edge, permitting less sophisticated quadrature phase-shift keying (QPSK) modulation, with a consequent reduction in achievable data rates. Smaller cells, with shorter distances to the cell edge, counter this problem - although it takes many such small cells to achieve the same coverage as a macrocell.
The introduction of femtocells, such as the Vodafone Sure Signal box, is making it easy for homeowners and small businesses to install low-power basestations, using their broadband connections for backhaul to the operator networks. Some forecasts say that 45 million femtocells will be deployed worldwide by 2014.
There are other techniques that can provide quick wins or immediate energy reductions in basestations: optimising the power-supply rectifier; upgrading older 2G systems to more efficient transmission schemes, for example from GPRS to Evolved EDGE; and employing more free-air cooling.
Mobile VCE is trying to improve the efficiency of heavily loaded basestations to achieve a combination of high spectral efficiency and good coverage (that is, high bit/s/Hz/km2 figures) throughout the network, at low power drain. The graph above shows the results of a series of simulations of a macrocell set-up, in which an increasing fraction of the customers are serviced by femtocells. The graph plots the system power consumption per user, for different densities of active users in the macrocell, given that any unused femtocells are switched off. The left-hand side of the graph shows what happens when no femtocells are in use. The right-hand side shows what happens when all the femtocells are used and the macrocell is off. Overall system power consumption falls as femtocells are deployed, but starts rising again when penetration reaches 60 per cent.
This system power consumption measure includes both the operational power, and the embodied energy of extracting and transporting raw materials, the manufacture, assembly, and installation of both the basestation and the femtocells, and their final disassembly, deconstruction and decomposition. At lower user densities, of only 30 to 60 users per macrocell, we achieve an increase in power saving. In these simulations an overall energy reduction of between 3 and 40 per cent is achieved.
The traffic load on cellular networks can vary by a factor of two daily, between a low at 7am and a peak at 9pm. This means we also need to consider low-load situations in our efforts to improve basestation efficiency, including the use of sleep modes to better share scarce spectral resources.
There are two closely related low-load techniques. One reduces the number of active basestations and reallocates their users, saving power by switching off parts of the network at selected frequencies, or by reducing the numbers of basestation sectors in use. Switching active basestations to operate at 900MHz, rather than at 1.8GHz, helps reduce propagation losses and so saves RF transmission power. It is possible to model the load in the low-frequency band against the high-frequency band and show that, particularly at low loads, operators could achieve a 70 per cent to 80 per cent energy saving by shifting to lower frequency operation.
We can also improve network efficiency in situations, such as handling email, which can tolerate long delays, by using store-and-carry-forward relaying techniques. The figure at right shows how the technique can be implemented in vehicles. A user's data packet is picked up over a short distance by a passing car, and then delayed (stored) until the car is closer to the basestation, so the data can be carried over shorter distances with less propagation loss. The technique only applies to data that is not time critical, but it may provide another way to achieve energy reductions.
There are other ways to achieve energy savings in basestation design, often through more efficient components or greater integration. Using remote radio heads can save cable transmission losses, as well improving the coupling of the power amplifier to the antenna.
Advanced power amplifier techniques, such as Class J designs that use a reactive component at the fundamental terminal impedance, can offer efficiencies like those of Class B designs with useful output power levels. The figure below shows a Class J power amplifier design which offers 60 per cent to 70 per cent efficiency, at full output power level, with 140MHz of RF bandwidth at its 2.14GHz nominal operating frequency. These designs are not a panacea – it still takes care to handle the high peak to average power ratio of the orthogonal frequency-division multiplex (OFDM) signals now widely used in cellular mobile systems. Done well, though, it may be possible to combine a Class J power amplifier with techniques in which the supply voltage tracks the RF envelope, to offer a considerable overall increase in power amplifier efficiency.
The efficiency of radio basestations, defined as the amount of RF energy they radiate divided by the AC power supplied to do so, is increasing year-on-year, from 3 per cent in 2003 to 12 per cent in 2009 and possibly to 25 per cent in 2015. But we still need to establish the optimum cell size (macro vs pico vs femto) taking into account backhaul power requirements, and to ascertain whether femtocells or relays or a mixture of these will prove the most energy-efficient way to extend cell coverage.
For the future, the next generation of the LTE standard, LTE-Advanced, may implement coordinated multi-point strategies, in which transmissions from several basestations are coordinated to cancel interfering signals, particularly at cell boundaries. This opens up the possibility that the improved knowledge of the propagation channel and interference signal that the standard implies could be used to reduce energy consumption as well as increasing traffic throughput.
Basestation designers can also consider further hardware optimisations and greater component integration as a way of shrinking power yet further. Applying all these techniques in concert should help us keep control of network power consumption and the resultant CO2 emissions, even as the usage of mobile networks sky rockets.
We would like to thank the academic researchers from our partner universities for contributing their results and figures.
Peter Grant is an Emeritus Professor at the University of Edinburgh. Simon Fletcher works at NEC Telecom Modus.