Smartphones designed to exploit the emerging 4G networks are ravenous for data, and that's good news for service operators who need ways to maximise usage-based revenue models. But the transition from 3G to 4G is likely to throw up several contentious issues that will take time and effort to reconcile.
When work started on Long-Term Evolution (LTE), the cellular technology designed to replace 3G, the plan was to provide data rates that could keep pace with handset technology. Now demand threatens to outstrip supply, and 4G is mutating into a new kind of wireless network.
It will be a network which, instead of relying mostly on large 'macrocell' basestations, will give way to an extended fabric of much smaller 'femtocell' transceivers sitting in people's homes or 'metrocells' that hang off street lamp posts.
For the wireless industry, "this transition is what the PC was for the computer industry", says Rupert Baines, vice president of strategic marketing at semiconductor firm Mindspeed. "It's the same philosophy of small, cheap things pushed out to the edge, rather than the focus on big iron tended by teams of experts that we saw with mainframes."
Driving the transition is the ravenous hunger for data that the smartphone has brought with it. A study by wireless-networks usage specialist Arieso of wireless usage on a major European operator's network found that each new generation of smartphone is much more hungry for data than the last.
"There is a device war going on; each new device constitutes a new front line," says Michael Flanagan, CTO at mobile solutions provider Arieso. "There has been a 50 per cent increase in usage with each generation."
As an example, iPhone 5 users typically transfer four times more data than people with the original iPhone 3. The reason for the explosion in smartphone data usage seems to be down to their always-on nature, Flanagan reckons. People keep their phones active to be able to take calls - "And data will continue'in the background," Flanagan explains. When not being actively prodded and stroked by users, tablets are far less likely to have persistent data connections and are used more often in Wi-Fi zones, so feature less in cellular data-usage measurements.
"We are moving towards 1GB per day per consumer by 2020," notes Kai Sahala, head of mobile broadband marketing at cellular equipment vendor Nokia Siemens Networks. "We see that being consumed by some users already with some LTE subscriptions."
Simply shifting from 3G to 4G technology can handle some of the necessary increase in capacity but with all its data-coding techniques LTE is now close to the Shannon limit - the maximum theoretical data rate that can be achieved for a given slice of radio spectrum.
"There are only a few ways to get more capacity out of your spectrum," says Nick Johnson, CTO of picocell/femtocell vendor Ip.access. One option is to use the higher frequencies that are less congested, "but the propagation of radio above a few gigahertz becomes very difficult. You can't get more than a few tens of metres", Johnson adds. "The other option is to split the cells into smaller chunks. Cells have to get smaller to reuse the same spectrum over and over again."
Adding cells sounds easy but is far from trivial. "Most tier-one operators are running out of places to put traditional macrocell basestations," says Arieso's Flanagan. One answer is to scale down, using much smaller femtocells and metrocells. This is where the heterogeneous network or 'het-net' comes in. "The idea is to manage small cells within a macrocell network," says Johnson.
Sahala says operators can optimise how they split radio frequencies across the basestations in a het-net. Lower frequencies can propagate more easily - those released by the switching off of analogue TV are much better than those above 1GHz that were allocated for 3G, and which EE is using for its initial LTE network in the UK.
Those higher frequencies are likely to end up being allocated to small cells. The small cells do not need to be based purely on LTE. Pockets of Wi-Fi-based access are already up and running. They are used by operators to try to offload traffic from their 3G networks although at the present time, users have to manage the switch between Wi-Fi and cellular themselves.
"Then you find the Wi-Fi network congests while there is capacity on 3G," says Johnson, because people naturally assume that Wi-Fi will deliver better performance, which is fine until too many people are on that part of the network. In principle, the self-organising network (SON) software that controls the various basestations and Wi-Fi nodes can'look at where the network has capacity, and then switch users on to it. The first step is session-based control in which the SON allocates a user to Wi-Fi or cellular when they first appear and that is where they stay until they move out of range or stop using the service. Another user arriving just seconds later may be given a different allocation.
"When you want a more sophisticated approach, you want to be able to switch networks during a session. If LTE gets busy, you can move a user to Wi-Fi," says Johnson, "but you can optimise occupancy to within 5 per cent using just session-based control."
Flanagan argues that the importance of Wi-Fi is receding. "Offload means to me 'please offload me off Wi-Fi'." He argues that the proliferation of Wi-Fi access leads to interference issues that the more precisely planned cellular networks do not experience.
Baines says Wi-Fi still has a chance, especially with more recent versions of the standard. "When you've got 5GHz Wi-Fi the range isn't great but if you're sitting in a coffee shop that's fine, and the bandwidth is whopping. Most phones have built-in support for 802.11n that will handle higher data rates."
Wi-Fi offloading benefits
Ip-access's Nick Johnson says operators are moving back to seriously considering Wi-Fi for offload purposes: "The momentum is coming. Operators used to see it as a threat but now see the opportunity. They see a way to manage it, using technology from the cellular side." As the intelligence of the het-net evolves, it will make more strategic decisions. As the user moves around they could quickly move in and out of range of many small cells.
Although cellular standards have robust handover protocols, each handover increases the risk of a dropped connection. The SON can be used to detect users moving at faster than walking pace - which indicates they are in a vehicle that could quickly move out of range - and avoid connecting them to small-cell basestations even if they would get better data rates for brief periods. "The software may switch you to a macrocell if you get on a bus," says Mindspeed's Baines.
Because they will usually lie within the confines of another basestation that is operating on a different set of frequencies to avoid mutual interference, macrocell and metrocell basestations can be used in more complex ways. The two types of basestation have very different levels of output power.
You can hang a metrocell off a lamp post because its radio transmissions are much weaker and so have a much smaller coverage area than those from a macrocell transmitter that has to sit at the top of a tall tower.
Devices will often be in a zone just beyond the edge of the metrocell or femtocell's nominal range, which should put it into the zone of the macrocell; but, Anders Furusk'r, technologist at Ericsson Research, points out that the phone has to transmit data to the basestation as well and has to use more power to reach a macrocell that may be further away than the femtocell.
The best answer is to make the device receive data from one and send data to the other; that takes a lot of coordination in the 'backhaul' network that links the two basestations together. ABI Research's principal analyst Aditya Kaul maintains that Ericsson is pushing for these advanced features. There is a catch: they only work well where the equipment vendor has tight control over the software and hardware used right across the het-net.
Although metrocells will be much cheaper to buy than macrocell basestations, there are still obstacles to overcome, such as wiring them up to the core network through backhaul links, and civic red-tape: convincing local authorities to host them on their publicly-owned street furniture.
"The question is how cost-effective it is for operators to deploy," says ABI Research's Aditya Kaul "That will determine what proportion of LTE will have small cells. The overall cost of small cells comes from not just the access point itself but the backhaul, management and site acquisition."
He adds: "All of that does add up to a lot of money. It will take time for some of that cost to reduce, especially the backhaul cost, site acquisition and management. Given the rate at which bandwidth needs are increasing, the rise of the small cell as part of the radio access network (RAN) seems inevitable. Operators will not be too fussy about which particular protocols they then use. Nick Johnson at Ip.access calls it the "Shannon-limit RAN": "Those technologies are all as close as makes no difference to their Shannon limits. The trick then is to make a network from multiple radio standards that all behave the same way: Wi-Fi, 3G, 4G integrated in one box with much of the intelligence moving to the edge."