SuperCell, LTE-C, Super C-RAN ‘ these are some of the labels which are already being bandied around for the next generation of mobile macro network architectures, and those are just from NTT Docomo. The Japanese carrier, together with China Mobile and the two largest Korean cellcos, have established themselves firmly as the innovation leaders in the macro layer, deploying Cloud-RAN architectures before there are any standards or off-the-shelf equipment to support them; and not content with that, already looking to the next generation. They are all major supporters of small cells too, but in that area they are joined by AT&T, Vodafone and other enthusiasts from the west. In redesigning the macro network, however, they are well ahead of the pack, and making important moves towards the full HetNet (heterogeneous network), in which the small cell/Wi-Fi and macro layers will truly interoperate and form a seamless cloud of capacity.
Despite the obvious benefits this would bring in terms of flexible use of spectrum and network resources ‘ with consequent improvements in cost, user experience and power consumption ‘ there are still many hurdles to cross before HetNet becomes real (although, of course, many operators will introduce a smattering of Carrier Wi-Fi and small cells and label it a HetNet). But in the meantime, and despite the roll-out of small cells, most carriers will continue to expand and enhance their macro networks for better 3G and LTE coverage and capacity. Small cells will initially be deployed tactically within the macro network to fill notspots or deliver a targeted performance boost where required. Then they will start to be rolled out in large numbers, as a network in their own right, relying on a combination of licensed LTE and unlicensed Wi-Fi bands. And eventually that layer will interwork fully with the macro network to form a seamless HetNet.
But the economics have to work, and currently they do not. The investment in brand new architectures will outweigh the revenue gains from additional capacity unless the cost of both small cell and macro enhancements is reduced dramatically. There is a lot of work being done to deliver the promises of $50 small cells with low cost of ownership (self-provisioning, self-optimizing, self-healing). But what about a similar cost target for the macro cell site (which may, incidentally, end up being almost as small as its metrocell cousin in a different band, in dense urban environments)?
Metrocell proponents currently target about a fourfold improvement in total network cost, compared to traditional macro. Initial C-RAN designs would be lucky to deliver a twofold boost, because of the costs of fiber and servers, and the early stage lack of scale. That will improve with bigger roll-outs and the advent of LTE-Advanced, which has features that considerably simplify the deployment both of C-RAN and metrocells. But a HetNet could target savings of 10x or more.
This is where Cloud-RAN steps in, with some significant new thinking needed to bring down the cost of the macro cell site, even before it is fully integrated with metrocells. That will rely on several key changes, which Docomo, China Mobile and others are incorporating in their ‘Super C-RAN’ R&D projects ‘ some of which could turn into commercial systems, if semi-proprietary ones, as early as next year. These are:
‘ Standardization, to reduce costs and improve vendor competition
‘ Harnessing key elements of LTE-Advanced, such as CoMP, to improve efficiency and reduce the number of cell sites needed. LTE-Advanced standards will help make full interworking between different layers, cell types and frequencies more standardized, reducing cost and risk. Key features will include CoMP, which reduces the number of cells required by improving performance at the edge; and eICIC, for better interoperability between layers.
‘
‘ Increased integration of cell site equipment to reduce cost and power, aided by AAS (Active Antenna System)
‘ Introduction of a ‘Supercell’ design, where a base station’s range is extended using low cost antennas (based on DAS or Distributed Antenna System principles) or metrocells
‘ Development of methods to use high performance microwave in place of fiber for linking the radio heads to the virtualized baseband in the cloud (fronthaul). The option to use microwave would reduce cost for some carriers, and remove a key barrier to C-RAN
These changes add up to the next wave of C-RAN, dubbed Super C-RAN, though in reality it is just the next evolution of a development which has been going on for years. C-RAN itself has evolved from earlier approaches to deconstructing the traditional macrocell ‘ splitting the radio from the baseband in a remote radio head, connected by the CPRI interface; integrating the radio and antenna into a single compact unit; moving baseband processing from a box on the cell site to a remote server. And in its current form, it is still a stage along the way to the new macro network, not an end point.
Future C-RAN architecture (Source: NTT Docomo)
Perhaps the most mportant enabler of the new network economics will be antenna innovation. AAS (Active Antenna System) is in its earliest stages, but will deliver even greater integration of cell site equipment than current offerings like Ericsson’s AIR (Antenna Integrated Radio). AAS goes a step further by placing radios along the antenna itself, reducing footprint and cost, and allowing for greater flexibility in spectrum bands. NTT Docomo has been among the first carriers to conduct major field trials of AAS though various iterations of the technology have been going through tests for almost two decades. Before LTE, the cost of AAS deterred operators, except in a few niche applications (like ArrayComm in the fixed wireless area). But they have far higher capacity needs in the LTE era, even as costs have come down with more efficient signal processors and integrated semiconductors. So they are revisiting AAS, and all the main OEMs have development programs in place.
AAS adds to the efficiency of the macro network in various ways, including active steering, which helps avoid interference between base stations ‘ vital as cells get closer together, and potentially applicable to metrocells in future, according to some innovators. China Mobile is already using active antenna steering in its TD-SCDMA and TD-LTE networks (in the latter, it uses an eight-way antenna array to steer the beams and reduce interference which can be a risk to unpaired spectrum system performance).
While interference management is AAS’s main contribution to making TDD networks more efficient and cost-effective, while in an FDD system, the big benefit comes from capacity enhancement. Features such as active beamsteering and vertical sectorization can deliver capacity improvements of 75% or more, at lower cost than alternative capacity solutions like metrocells or new urban towers. All these techniques will be used in combination in dense environments ‘ as Joe Madden, principal of Mobile Experts, puts it: ‘The laws of physics dictate that any new ‘5G’ air interface will not improve greatly on LTE’s spectral efficiency. Instead, the entire industry needs to turn toward spatial efficiency, using directional antennas, small cells, and other techniques to reuse spectrum more wisely.’ And deployment in China will start to introduce economies of scale to AAS, even as operators’ capacity pressures make them more willing to spend the $20,000 or so currently required.
Another aspect of antenna technology which could find a new lease of life in C-RAN is DAS, or at least new implementations of that approach for the new macro environment. Usually associated with coverage projects for stadiums or large enterprises, DAS techniques are being applied to C-RAN to extend the range of cell sites. Combined with CoMP, DAS could help the operator to reduce the number of base stations required to deliver 4G to a given area. A ‘master base station’ or supercell, supported by the cloud servers, would in turn have its range extended by distributed antenna units. (In China Mobile’s demonstration, a similar topology replaces Docomo’s antennas with small cell units, which would increase the intelligence at the edge. In either case, the aim is to extend coverage and capacity with the lowest cost equipment available).
This topology is being driven by the emergence of DAS equipment that uses the CPRI interface, the standard for connecting radio heads to the baseband. That makes it easier and cheaper to integrate DAS with LTE (currently, base station power levels have to be lowered to integrate with DAS, reducing overall efficiency). It also reduces the need for fiber, and that huge cost could be driven down further if microwave can be chosen for some fronthaul links too. Again, CPRI extensions are important, especially the Ericsson-backed implementation of the interface standard over microwave.
However, this would not always be appropriate in the most traffic-heavy areas. The data transported between the baseband and a remote radio is measured in Gbps, requiring far higher bandwidth than that between a small cell and the core network. While the data backhauled from a small cell consists of the user data plus some overhead for control and management, the data between the baseband and RRH consists of oversampled I/Q streams for different MIMO antennas. The higher the number of antennas, the higher the data rate and bandwidth, which will be a significant factor as operators look to deploy another key LTE-A feature, MIMO with 4×4 arrays or larger. It is doubtful whether microwave will have the capacity to handle the most advanced CPRI interfaces, designed for very high bandwidth in applications like C-RAN (the same goes for the other standard base station interface, OBSAI, whose release 3.01 defines a 6Gbps rate).
All that means that the basic equation when evaluating C-RAN is the balance of reduced processing cost versus increased transport cost, and the result will vary according to an individual operator’s circumstances. Lower cost architectures for fiber, and the use of microwave where appropriate, will reduce C-RAN transport costs per site, but this will often be outweighed by the increase in the sheer number of radio heads to support. Yet without huge numbers, the economies of scale of centralized processing do not work out.
With limited savings to be made in the fronthaul, the pressure is on to usher in a radical new cost structure for the infrastructure itself. The major contributors will be:
‘ Extremely low cost, low power cell site equipment, with radio/antenna units that are driven down to the ultimate $10-per-sector level.
‘ Commoditized hardware at the back end, based on massive cloud server farms, with similar economics to the servers of Google or Facebook. These may be operator-owned or provided ‘as a service’ but they will rely on standardized Intel or ARM processors with commonplace languages such as C. Dedicated processors will remain for some specialized activities (for instance, deep packet inspection in sensitive application), but many see the digital signal processor, with its rarefied programming skills, disappearing from the wireless network and finding its growth in other applications like video and gesture recognition.
China Mobile has said its platform will only make sense if a server is supporting hundreds of LTE sectors. An Intel server blade costing thousands of dollars cannot compete against a BTS chip costing a few hundred dollars, until there is huge scaling ‘ but at that point, the cost of the BTS chip needs to come down below $50, which will be impossible using traditional designs.
C-RAN will not stop at virtualizing the baseband processing of a few cell sites ‘ it will gradually virtualize all the network’s activities so that just about all the cost is centered on the servers and their software, delivering economies of scale. Emerging standards for virtualization and carrier SDN, notably NFV, will be important drivers, as will the ability to program in C rather than specialized languages.
All this could bring carriers closer to the dream of the $50 cell site ‘ a price point which would enable them to deploy vast numbers of cells in their macro layers, not just their small cell layers, at affordable cost. And that scale is essential to make the C-RAN equation balance ‘ the investment in the cloud servers is not justified if they are supporting only a few sites, and gets more attractive the more the network is scaled up.
Fuller analysis of C-RAN economics is available from Maravedis-Rethink’s RAN Service, including an upcoming major report on LTE RAN trends, and a new research note on C-RAN costings. For more information, please email Adlane Fellah on [email protected]
ASOCS:
A start-up called ASOCS has been vocal in pointing out these new directions. It licenses its IP (silicon cores for modems) to vendors and chip suppliers in the C-RAN environment and has worked with Intel and China Mobile on the latter’s trials.
There are two main elements to its solution. One is MPL (Modem Processor Language), which allows developers to call C libraries to run jobs on processors and ports DSP-based code into the C-RAN environment. It aims to take a similar approach to that of OpenCL in graphics, shielding developers from the complexities of the hardware below in order to ease migrations to C-RAN.
The other aspect of its solution is the MPU (modem processing unit), which provides base station functions on a low cost board. The MPU board is the point of entry for signals from the remote radio head and acts as a bridge to connect to other base stations in the C-RAN architecture, which may be only a few inches away. ‘We are the watchdog, liaison officer and hardware accelerator for the system,’ says CEO Gilad Garon. The approach reduces the cost per LTE sector to under $30 in a large roll-out, says ASOCS.
In the second round of China Mobile C-RAN trials, starting this fall, the operator will use MPL as well as servers with PCI Express cards for the baseband processing. Each card will run on four FPGA (field programmable gate array) chips with Asocs C2100 cores, and each FPGA will be able to handle 12 LTE sectors, according to Garon. The core link to Intel Xeon processors, though in future, ARM-based servers are also likely to take a role in carrier SDN. That will be the next big move to bring new price competition into play and disrupt the C-RAN cost base yet again.
The latest LTE-A trials and upgrades:
Carriers are pushing the limits of the LTE platform, especially in Asia, even before many have moved ahead from 3G. A growing band of operators is doubling peak LTE speed to 150Mbps, harnessing wider bands (like EE in the UK) or carrier aggregation (like SK Telecom). But others are boosting the speeds even further, in trials at this stage, but with commercial support on the horizon. Some recent highlights include:
‘ Japan’s eAccess has achieved downlink transmission speeds of 291Mbps using carrier aggregation and 4×4 MIMO, in recent field trials conducted in the city of Takamatsu, in the 1.7GHz band. The regulator had granted a temporary licence to trial the network from August until October this year.
‘ Another Japanese player, Softbank, has completed the world’s first LTE-Advanced trial in the 3.5GHz band, which many operators are eyeing for future capacity enhancements and as a possible small cell roaming frequency. This spectrum should be allocated in Japan by 2015. It worked with Huawei on a pre-commercial evaluation in the Tokyo district of Ginza and achieved peak download speeds of 770Mbps, and an average of 500Mbps. This used carrier aggregation, CoMP and Cloud-RAN technologies to boost performance and spectral efficiency. Softbank also demonstrated a ‘five-carrier carrier aggregation technology for TD-LTE in 3.5GHz using a prototype system that delivered maximum download speeds of 1.2Gbps on Huawei’s ultra-wideband remote radio unit. These RRUs can support more than 100MHz of bandwidth.
‘ Poland’s fourth cellco Play, which operates under the P4 brand, will roll out LTE to 13 cities by year end and will immediately skip to the 150Mbps speed. It will deploy in the 1.8GHz licences it won earlier this year. Polkomtel, which runs the Plus brand, already offers LTE in the 1.8GHz band.
‘ SK Telecom was the first cellco in the world to deploy features of the LTE-A standard, notably carrier aggregation, which it introduced to parts of its network this summer. It says this technology is now delivering average downlink speeds of 62Mbps, according to benchtests run by Signals Research. All three Korean operators have now deployed LTE-A carrier aggregation, moving towards bandwidth of up to 100MHz, with potential peak downlink of 150Mbps (higher if other LTE-A enhancements like 4×4 MIMO are also deployed). Last week, SK Telecom gained access to 35MHz of 1.8GHz spectrum, which it will split into 20MHz downlink and 15MHz uplink. It plans to extend 150Mbps CA-enabled LTE-A nationwide by next July in 800MHz, and then it will combine the new 1.8GHz frequencies too, boosting peak download speed to 225Mbps by late 2014.
‘ Hong Kong operator CSL has demonstrated an LTE network which is ready to support 300Mbps peak downlink speeds, working with ZTE. The speeds harness carrier aggregation. CSL is the only Hong Kong cellco with 2x20MHz in both the 1.8GHz and 2.6GHz bands for LTE.