The dawn of true mobile broadband, usually enabled by LTE, has brought with it a host of new network buzzwords ‘ small cells, HetNet, SuperMacro, Massive MIMO, Cloud-RAN, hyper-density. All of these involve a combination of market hype and real promise, and it is only when a reasonable number of mobile operators ‘ not just the cutting edge mavericks ‘ start to deploy LTE at scale that a real picture is obtained of that balance. Carriers formulating mobile broadband strategies have to face more key decisions than in any previous generation because the traditional ways to improve the ‘three Cs’ ‘ coverage, capacity and cost efficiency (joined in 4G by power reduction) ‘ are nearing exhaustion. But most of the ways to improve on merely adding more macrocells or more sectors are relatively unproven, so all the catchphrases listed above have, despite the fanfares of their supporters, been approached with great timidity by operators.
In its regular survey of over 150 mobile operators worldwide, Maravedis-Rethink’s RAN service has identified some of the most important decisions carriers are making, spoken to those companies which have already made them, and assessed the real world uptake of key architectures, as well as forecasting adoption to 2019. The report, Towards the hyper-dense network ‘ the shape of the HetNet 2013-2019, will be published in June, and will analyze some clear emerging trends in operator behavior.
While the ideal economics might be achieved by adopting a fully virtualized HetNet, with multiple layers of cells knitted together by SON (self-optimizing networking) in the cloud, that is a platform for the future. Some MNOs may choose to eke as much life as possible from 3G and then leapfrog straight to a truly modern platform, but they will need to be able to do without wide-scale LTE for three years or more, so many are having to make do with what is available and trusted now. The timing, and the need for those carriers to write down their current equipment investments before moving to something radical like Cloud-RAN, will be a significant brake on rapid deployment of new architectures ‘ and also a key reason why many MNOs will turn to outsourcing and cloud companies to handle their migration to a virtualized world.
First-stage LTE roll-out, then, remains conventional in most ways, outside the showcase roll-outs of NTT Docomo or SKT. In most cases it is geared to coverage first, especially where sub-1GHz spectrum has been awarded, and many operators without that spectrum will wait until they obtain it to roll out at any scale. Coverage projects have relied mainly on macrocells, often using a distributed architecture where remote radio heads and antenna-integrated radios are separated from the baseband, but with limited adoption of virtualized basebands (Cloud-RAN or Centralized-RAN) as yet.
The main decision for most operators in phase one has not been whether to deploy small cells, or a virtualized architecture, but whether to adopt an overlay or a rebuild (Single RAN) approach. The former approach has so far been more popular, as seen at Verizon Wireless, Telefonica and others. With an overlay of existing networks, carriers can roll out rapidly and keep deriving benefits from existing 2G/3G infrastructure, while introducing new vendors if required. By contrast, SingleRANs like Sprint’s come with higher initial cost and time to deploy, but leave the carrier with a fully modernized architecture, which can deliver significant efficiencies in power consumption and the number of sites required.
Despite those advantages, in 2013 the early movers were heavily focused on speed and cost efficiency and the LTE cells deployed in overlay projects outnumbered those for SingleRAN and greenfield projects by three to one. That difference will narrow over time, however, as 2G/3G technologies run out of steam, new LTE options become cheaper and as new architectures mature. The number of cells deployed for SingleRAN and LTE-only will overtake those for overlays in 2017, and by that time, many overlay adopters will also be adding capacity and future-proofing their systems by adding a new layer of small cells and WiFi, possibly managed in the cloud but interworking with the established macro network.
This is the start of a HetNet, and shows the usual shift from coverage-first, with some capacity hotspots, to capacity-driven deployments ‘ and to a level of capacity which has never been seen before.
Some of the operators which embarked on phase one early, notably Verizon Wireless, are now moving towards phase two. Whereas in the first years, areas where high density was required ‘ such as stadiums, transport hubs or business parks ‘ could often be addressed with a small number of metrocells, or with WiFi offload, but as the decade progresses, operators in developed economies will increasingly be hearing the dreaded word ‘densification’. That will see the key decision shifting from overlay vs SingleRAN, to macro-first vs small cell, in the move towards an eventually ‘hyper-dense’ network.
A dramatic rethinking of the whole network is needed to drive down cost of delivery, allocate capacity flexibly where it is most needed, and support massive bandwidth affordably. The air interface upgrade itself delivers only just over 15% of the additional capacity and efficiency that will be required by 2020, so many other techniques will have to be used in combination. Key elements of the LTE-Advanced family of upgrades (3GPP Releases 10 and 11), will be important from 2014, initially carrier aggregation, but moving on to other technologies like eICIC (important for mitigating interference between small cells) and CoMP (a motivation to adopt C-RAN).
The deployment of a layer of small cells will be a common approach, and an important step towards the fully-blown HetNet (heterogeneous network, another overused term but the ultimate goal of most cellcos. This will create a common pool of capacity, harnessing different technologies, spectrum bands, cell sizes and backhaul links in a fully integrated way.
But the macro layer remains the primary source of coverage and capacity, and it too must be modernized to provide better total cost of ownership, cost of data delivery and overall efficiency. There are two main approaches to this. One is Cloud-RAN, in which carriers deconstruct their networks completely, virtualizing the baseband functions in base station hotels or on standard cloud servers, and leaving just low-cost, low power radio/antenna units at the cell sites (which may be increasingly small).
The other is to add more intelligence to that cell site and create what the UK’s leading cellco, EE, dubs the ‘super macro’. These approaches need not be either/or ‘ many operators will use a mixture of both, or will converge the two into a hybrid system where super macros act as ‘master cells’ within a Cloud-RAN, localizing some functions like caching and supporting a large number of more basic sub-cells, which may use a distributed antenna design.
Similarly, most operators will use a combination of Cloud-RAN and small cells on the road to HetNet, even though these two are often presented as mutually exclusive (or even confused with one another). Small cells are initially being deployed, in the public access environment, to fill gaps in the 3G network, but their real disruptive effect will be felt when they form a separate layer of capacity, interworking completely with a reconstructured macro layer.
The main carriers pointing the way to these new approaches have often been the mavericks of Japan, China and Korea, but increasingly, they are being joined by traditionally more circumspect western players such as AT&T and the UK leader, EE. The latter epitomizes the way that most mobile operators, at whatever stage they reach phase two or three of LTE, will start to harness a mixture of architectures ‘ with the balance and interworking between them as important as the technologies themselves.
EE has been an exponent of enhancing its macrocells as much as possible before investing heavily in small cells, with a mixture of advanced MIMO, carrier aggregation and other techniques. The MNO’s principal architect, Andy Sutton, told a recent conference: “Super macro is the first step toward building a HetNet. Evolving the macro is the most cost-optimized way to add capacity into our networks.” A super macro would typically support 2G, 3G and 4G in multiple bands with 3-6 base station sectors and carrier aggregation. Such a site would often be a master cell, acting as a hub for smaller microcells or distributed antennas.
Super macros use some of the techniques which were well tested in 3G for adding capacity, coverage and efficiency ‘ adding more base station sectors, vertical sectorization, attaching microcells, and so on. They will also be increasingly reliant on emerging antenna techniques such as ‘super MIMO’ (4×4 and beyond, or ‘massive MIMO’ as China Mobile is trialling), and AAS (Active Antenna System). Indeed, most ‘5G’ R&D centers almost entirely on antennas, along with software-defined networking and the effective use of many spectrum bands, including ones currently considered unsuitable for mobile broadband access, such as 60GHz. Increasingly, the radio will be integrated entirely into the antenna, allowing for sophisticated beamforming, vertical sectors and multiband, multimode support, with MIMO arrays adding increased range and bandwidth.
Despite the focus on super macros, EE will launch small cells in the next two years, according to Sutton. “We’re going to be introducing heterogeneous networks, and we’ll be putting public access small cells into the network where appropriate,” Sutton told FierceWireless but he added that the operators immediate priorities were to add capacity by building out in its 2.6GHz spectrum and develop its ‘super macros’ to improve coverage. That means “we don’t need small cells today. So we’re working on trials, working on development today’. He expects roll-out to ramp up quickly from 2016, indicating a common pattern when it comes to public access small cells ‘ caution and a limited hotspot approach in phase one of 4G, followed by a full-scale, dense layer under most of the network in subsequent phases.
Beyond the current generation of 4G technologies, the potential for multilayered networks and small cells becomes greater, thanks to virtualization and other advances. Sutton said: ‘We’re looking at opportunities maybe to focus a lot of the control plane in the macrocell, and then really focus the user plane on the small cell layer in the urban areas where we’ve got high traffic density.’ This echoes an architecture demonstrated by NTT Docomo, which uses ‘phantom cells’ in the user plane, interworking tightly with a Cloud-RAN.
In the nearer term, carrier aggregation is the most important enhancement to EE’s mobile broadband capacity and data rates. It is currently planning the aggregation of a 15MHz third carrier, which would boost the network’s theoretical peak to over 400Mbps and is likely to be introduced around the turn of the year. EE announced a pilot of dual-band CA in London’s Tech City last November, using 20MHz of 1.8GHz and 20MHz of 2.6GHz spectrum. It is not clear which band the third carrier will be in. EE only has 2x5MHz in the 800MHz band, while owning 2x45MHz in 1.8GHz and 2z35MHz in 2.6GHz (and its 2x20MHz of 2.1GHz 3G frequencies). The only carrier to have announced triband CA to date is SKT in Korea.
Over the pond, the big US carriers are also adopting complex combinations of technologies to maximize their capacity, data rates and coverage, while keeping costs low. While Sprint has created a new infrastructure, with Network Vision, to host all its networks and spectrum bands, Verizon and AT&T have been more focused on overlay, and now on adding smaller cells. Ben Moreland, CEO of tower operator Crown Castle, gleefully told the recent PCIA infrastructure conference that he expected the two leaders to build “several thousand” more cell sites in the next few years, especially once they acquire new spectrum in the broadcast incentive auctions in 2015. In particular, he believes it is still more effective for carriers to add new antennas for new spectrum bands, to maximize the spectral efficiency of their latest holdings. Single-antenna arrays supporting multiple frequencies are evolving rapidly, but Moreland insists these remain
‘sub-optimal, depending upon which frequency you are talking about’.
Whether the antenna makers would agree, the tower companies are certainly looking for a 4G bonanza, and turning their efforts to lower level sites such as rooftops as well. They are especially excited about VoLTE, which could require operators to increase the density of their LTE networks by 20% to 30%, by some estimates, in order to ensure full coverage and avoid fallbacks to 2G/3G during voice call.
AT&T may be one of the great cheerleaders for small cells, but it also expects to add between 1,500 and 3,000 macro sites a year for the ‘foreseeable future’, to improve coverage and capacity and support new spectrum such as 2.3GHz WCS or AWS-3, according to Bill Smith, who heads up network operations at the carrier’s mobile arm. He told the PCIA conference that AT&T is also on track with its plan to deploy 40,000 small cells by the end of 2015, and it is introducing new waves of DAS for indoor coverage, as well as tower-mounted radio heads to prevent power loss to its antennas and so boost signal strength.
Like EE, he also sets great store by the potential of carrier aggregation to make better use of existing spectrum and network assets. He commented that CA was “probably one of the single most important” things AT&T could do to improve the performance of its networks, though the operator will not reveal its LTE-Advanced plans, including aggregation, until later this year, and start to implement them in 2015.
For its part, Sprint has suffered delays in its ambitious network modernization program, but recently completed lab testing of new TD-LTE radios, with eight transmitters and eight receivers (8T8R) per box, which will give its network a significant speed boost . Field testing of the 8T8R gear has now begun with its three vendors, Ericsson, Alcatel-Lucent and Samsung, with commercial implementation expected, as an overlay to existing base stations, in a couple of months’ time.
When used at the cell site in conjunction with MIMO antenna arrays, 8T8R can give Sprint greater range in the 2.5GHz band, whose main drawback is limited reach. The overlay will coincide with the completion of Sprint’s replacement of its 3G network and LTE will cover 250m POPs, while it will also add HD Voice nationwide (to enhance voice quality rather than racing towards VoLTE).
Such advances show how the macrocell is being stretched to new limits via MIMO and new radios, and especially the increased use of TDD spectrum, which has many advantages in the HetNet, including the ability to reduce interference between layers. However, while the 2.5GHz TDD spectrum Sprint acquired with Clearwire is plentiful, it does have shorter range than its rival’s primary bands, such as 700MHz and AWS, or its own 1.9GHz. Jim Taiclet, CEO of American Tower, said on a recent earnings call that Sprint would need to build out 30,000 to 40,000 new sites if it wanted to reach the same coverage in 2.5GHz as it has in 1.9GHz. Sprint said last summer that it would deploy 2.5GHz, and its multiband Spark offering, on all 38,000 of its planned Network Vision cell sites, rather than keeping the high frequency spectrum just for urban areas, as many had expected. Sprint also said it would likely have to deploy additional small cells beyond Network Vision because of the lower propagation of 2.5GHz, though it has not specified any numbers.
Last summer, Jonathan Chaplin, an analyst at New Street Research, estimated that Sprint would have to increase its total site count, including small cells to between 50,000 and 60,000.
All this mirrors the findings of Maravedis-Rethink’s new report, which points to increasing density, not only through small cells, but in all areas of the network. Macro layer coverage and the improved efficiency and economics of LTE have certainly delivered benefits to carriers in terms of performance and cost, but those will be quickly overtaken by the growth in mobile data demand if they do not enhance their networks on an ongoing basis. Densification will be the next priority, along with virtualization – not so much to reduce the number of sites, but to manage them more efficiently and improve total cost of ownership. Eventually, SON and virtualization will help create a full-blown HetNet, a huge pool of capacity which can be allocated flexibly as required regardless of frequency, air interface or cell size. That remains a long way off, and may not be fully developed until ‘5G’ appears around the end of the decade, but the steps which operators will take to get there are becoming far clearer, which will accelerate the process of creating common industry platforms and the reassurance of a shared approach.
For more information about the upcoming report, ‘Towards the hyper-dense network ‘ the shape of the HetNet 2013-2019’, please email Adlane Fellah on [email protected]