Verizon may have driven the first wave of LTE roll-outs, but the three South Korean cellcos have claimed the undisputed crown when it comes to pushing ahead from basic 4G coverage roll-out. While initial LTE deployments have generally followed conventional lines in terms of network design, as capacity shortages loom in some regions, carriers need to start implementing the features which make LTE a truly new architecture, not just an air interface upgrade. Here, the Korean operators are leading the way with critical new elements of LTE and LTE-Advanced (3GPP releases 8-11) ‘ public access small cells, carrier aggregation, higher grade MIMO and so on. But the return on investment remains uncertain until these new capabilities are fully standardized ‘ and so far, there is significant operator caution about whether the standards alone can deliver what they need.
This week, the third Korean cellco, LG U+, emulated SKT by offering 150Mbps download speeds in some parts of its LTE network by harnessing carrier aggregation across its 800MHz and 2.1GHz spectrum. Samsung has supported this capability in its Galaxy S4 LTE-A and LG is likely to follow suit with the upcoming Optimus G2. (Some carriers, like EE, are achieving the 150Mbps peak speed without being able to make the LTE-A boast, by using techniques present in early 3GPP releases.)
So far SKT has implemented carrier aggregation to boost peak download rates to 150Mbps, and in future will combine two 20MHz channels to double that speed again (the current standards allow for up to five 20MHz channels to be aggregated). The other key element of SKT’s initial LTE-A roll-out is CoMP and in 2014 it will add eICIC (enhanced intercell interference coordination), which is important for HetNets as it mitigates interference between macro and small cell layers. However, this comes at the expense of reducing macrocell capacity, so eICIC only makes sense once a carrier has a significant number of small cells ‘ carrying between 60% and 80% of traffic, at Ericsson’s estimate.
The Korean mobile pioneers are being rewarded with massive uptake of their 4G services – SKT said earlier this month that its 150Mbps LTE-A network had attracted more than 150,000 customers in its first two weeks of operation. However, their relatively small and saturated markets, and the constant pressure to invest in leapfrogging one another, is not yet turning that uptake into significant profit growth.
That is just one of the warning signs that can be seen in the Korean experiment. Even the most advanced 4G operators in the world, with their sophisticated and affluent mobile populations, and their intense focus on new kinds of applications and services, are not transforming their results. Increased spending on infrastructure and marketing resulted in a 55% year-on-year profit drop for KT in the last quarter of 2012, while LG U+ recorded a loss in that period, though it returned to profit in Q113. And even market leader SKT has seen only marginal profit gains attributable to LTE ‘ its 15% year-on-year leap in Q113 was mainly down to its chip affiliate SK Hynix.
Operators round the world, then, face a lengthy period when investment in mobile broadband coverage and capacity will be a squeeze on profits, not a boost, and offsetting that will rely on two tough challenges ‘ reducing operating cost of the networks, and supporting new revenue streams. There are plenty of players which aim to harness at least one or two LTE-A features in the coming year. Those planning to make their initial moves this year include Sprint, Verizon Wireless, Telstra, Optimus in Portugal, Yota in Russia, and Tele2 and Telenor in Sweden. The other major US cellcos will follow early next year, with many others, with device availability, small cell deployment and mobile data traffic levels being key deciders of their timing.
But that leaves the huge bulk of operators holding back. To make the roll-out of a brand new architecture, not just a macro layer upgrade, justifiable, they are relying on new approaches to building networks, notably the small cells which promise to deliver targeted, intelligent capacity with low overheads; and on availability of tried and tested standards to address the problems inherent in the new platforms. So while the Korean majors have highly customized ‘ and hence expensive – infrastructure and devices, most lesser mortals will be waiting until currently cutting edge elements of LTE become mainstream and standardized, and until off-the-shelf devices are widely available.
Standards may be written, but they take a long time to be implemented in mainstream equipment and devices, and then tested for capability and interoperability. And there are also question marks over whether the standards can deliver the increasingly demanding requirements of the new mobile broadband networks. Standards-based kit delivers cost savings because of economies of scale and vendor competition, but it can also slow down roll-out, and support a lowest common denominator approach.
Operators therefore face two critical trade-offs which will affect the economics of their LTE strategies. One, whether to deploy early, which brings first mover advantage but also imposes semi-customized deployment, increasing upfront costs, risking interoperability problems later on and adopting first wave (and possibly immature) kit. And two, whether to go for a vanilla standards-based approach, which will reduce cost and risk, but could also delay roll-out and land the carrier with sub-optimal performance.
For those wanting to wait for availability of more affordable, fully standard products ‘ and that will be the majority of the world’s cellcos, outside the major hotspots of mobile capacity shortfall ‘ the key LTE and LTE-A specifications may be agreed, but they are not yet ready to deliver the goods in terms of performance and profit.
So which items from the LTE standards menu ‘ whether ‘Advanced’ or plain old Release 8/9 ‘ are carriers urgently wanting to deploy, and will the specifications, and the vendors, deliver what they need?
The most popular technologies for early adoption, starting this year in some areas, are those which enable carriers to make more efficient use of their spectrum, or to ease small cell deployment. In the former case, that mainly centers on carrier aggregation, implemented commercially for the first time in the world by SK Telecom recently. In the latter, the important enablers will include eICIC, to help mitigate interference; enhancements to standards-based SON (self-optimizing networks), to deliver metrocells’ automation promises without proprietary, and non-interoperable, SON products: as well as development of the X2 interface for interoperability with the macro layer.
The macro layer itself also needs to deliver more quality of experience within the same spectrum, since this is all most operators will have in LTE for several years. That puts the focus on CoMP and on higher grade MIMO. Operators, a recent study by Maravedis-Rethink found, are generally cautious about MIMO, especially beyond 4×4 arrangements, because of device cost, availability and battery life. However, some are coming round to the speed and quality benefits more antennas can deliver. An example is T-Mobile USA, which will deploy 4×2 MIMO in some of its new LTE network, to double the number of transmission paths compared to standard 2×2 LTE designs. That will improve performance at the cell edge since devices will be receiving more signals ‘Petri Hautakangas, north American chief for one of TMo’s vendors, NSN, told GigaOM that lab tests had delivered speed gains of up to 100% on the uplink at the cell edge, and between 50% and 60% on the downlink. Since 4×2 is included in the current LTE release, most devices will support it unchanged, whereas 4×4 and beyond will need handset modifications, with more antennas being squeezed into the phone.
Another important LTE-Advanced element, also deployed uniquely by SKT, is CoMP (Coordinated Multipoint Transmission/Reception), an interference management technology which improves performance at the cell edge. It works by coordinating and combining signals from multiple antennas, to support a more consistent user experience at the edge and to smooth handovers to new cells. It aims to harness the interference at the edge of a cell (large or small) and turn it into useful signals.
However, like other standards, it may not deliver the level of improvements operators will need to meet their users’ demands while keeping their own costs low. At the recent LTE World Summit, the director of ZTE’s European CTO group, Lei Xue, warned that the problem of LTE intercell interference is far greater than anticipated, even before carriers get on to dense metrocell designs. That means they may not be able to rely on CoMP alone to solve their congestion and user experience problems.
Another problem with CoMP, according to Lei, is that its relies on significant backhaul capacity being in place ‘ far more than most carriers will guarantee at every cell (the specs were devised with fiber presence assumed). Of course, ZTE has a solution, called Cloud Radio, a scheduling approach which creates a ‘supercell’, able to ‘follow’ the user as he or she moves around. by allocating resources dynamically between base stations. This enables it to use existing DSL or wireless backhaul links because cells are tapping into a pooled resource. ZTE says it reduces handovers by 50% as well as delivering an enhanced QoS at the edge.
This is a key trend in carrier thinking, also seen in Nokia Siemens’ Liquid Net platform, Cisco’s Intucell and other ways to make cells more ‘elastic’. But all of these perpetuate the cellcos’ eternal problem of being stuck with one supplier’s proprietary solution, increasing cost and potentially cutting them off from a rival vendor’s innovations.
The same dilemma is clearly seen in another important LTE-A feature, especially for small cells ‘ automatic neighbour relations (ANR) enabled by the X2 standard. Some operators and suppliers have serious doubts about X2’s capabilities and a year ago, the Small Cell Forum set up a taskgroup to examine the additional elements which would be needed to make X2 fit for purpose. This sharply illustrates the challenges for standards bodies to keep up with real world network requirements, both in terms of timeliness (the 3GPP’s processes move far more slowly than mobile business models), and relevance (the LTE ‘A releases contain huge numbers of elements, many of which will never be implemented).
The X2 interface supports communication between base stations to support various functions such as handover, load balancing, CoMP and cell activation and organization. It standardizes these control plane procedures to make different vendors’ eNodeBs interoperable, and to improve interworking between macro and small cell layers. However, not only does X2 in its vanilla form add to bandwidth and latency, but it is inadequate for complex networks of many cells, say its critics, raising the risk that vendors will step in with proprietary alternatives.
X2 suffers the fate of many standards ‘ because it was devised to try to meet any possible future use case, it is very vague. For instance, one technique to avoid interference is for macro signals to use ABSF (almost blank subframes) to transmit to small cells. This means that one layer does not transmit L1 control signalling within given subframes. However, the standard does not define which blank subframes to use, so two small cells can clash while trying to transmit in the same ones.
Such weaknesses mean that vendors are already adding proprietary extensions or interpretations. Cisco’s lead architect for mobile systems, Mark Grayson, commented: ‘Standards can define the X2 protocol but vendors can deploy it with different algorithms so it isn’t interoperable. CoMP and other features are moving back towards proprietary.’ He added: ‘There is a pushback from the tier one OEMs not to go beyond X2.’ Nick Johnson, CTO at small cell vendor ip.access, set out the cellcos’ trade-off more starkly, saying: ‘Operators can accept a de facto RAN monopoly in their networks, or support these interoperability standards.’