Experiments in so-called 5G mobile communications have begun, but early indications suggest that it's going to be a fundamentally different kind of cellular network that emerges.
The cellular industry's 'generation game' could now be coming to an end. At each point in its evolution cellular radio has found itself leaning increasingly on previous generations. When the first digital systems were introduced they effectively replaced the older, less-efficient analogue cellular systems.
When telecom operators spent big on spectrum licences that would allow them to deploy 3G networks, 2G systems - such as GSM - were pushed into the background, but remain even today important parts of the network because they have such wide coverage. The 3G parts of the network, meanwhile, remain focused on population centres.
For years to come, the 4G networks now being installed based on the Long-Term Evolution (LTE) protocol will lean on the older 2G and 3G networks to support voice calls. LTE will be reserved for high-bandwidth data and video. It will almost certainly take a long time for 4G to extend beyond major conurbations, but it is a system that is meant to cover both urban and rural environments, so could, in principle, eventually push 2G and 3G out of the mobile communications picture altogether.
So now operators are thinking about the next step forward - but this will be a different kind of network evolution. Rather than being an enhanced replacement for 4G, operators see it as a merger of many different technologies. A new radio standard is only part of the picture. And yet some of the changes being proposed could reshape the way cellular networks operate.
Although Samsung claimed to have demonstrated the first 5G-capable systems in Q2/2013, any network that legitimately lays claim to that name is some way away. Nonetheless, Professor Rahmi Tafazolli, who heads the Centre for Communication Systems Research at the University of Surrey, says it will evolve: "It will be at least seven or eight years before we have a complete specification," he believes.
"We are looking at new systems coming in around 2020," predicts Lauri Oksanen, head of research and technology at equipment maker Nokia Siemens Networks, "but we do not talk about 5G as the overall network evolution. We don't want to label everything '5G'. Our view is that 5G will really be about better local-area performance, with lower latency and higher bandwidth in high-density hetnets."
Oksanen refers here to the industry's contraction of the term 'heterogeneous networks', in which different types of radio technology co-operate and interoperate. "Further development to LTE is the most likely way to go for macrocell, wide-area coverage," he adds. "A new 5G radio would be more of a complement to LTE evolution."
Prof Tafazolli also maintains that 4G is a very good technology: "It provides good speed, and when it offers national coverage people will have a much better experience of Internet usage on the move... The problem is that the way that we use the available radio spectrum and the way we have developed the standards is not efficient," Prof Tafazolli adds. "We are running out of radio spectrum. Before 2020, with all the spectrum that we have, most Western Europe capitals and cities such as New York, Los Angeles, and Tokyo will run out of capacity. In short, we have to come up with revolutionary ways of using the spectrum."
In 1948 the American mathematician and electronic engineer Claude Shannon (1916-2001) developed a key piece of communications theory that asserted, for a given level of noise, there is a limit to how much data a channel can send. Nick Johnson, CTO of basestation maker Ip.access, says that existing radio technologies are "operating as close as makes no difference to their Shannon limits".
The University of Surrey's Prof Tafazolli agrees with this assertion, but adds: "It's wrong to compare everything to the Shannon capacity limit - because that is defined for point-to-point connections. The metric that really applies to cellular communications is capacity per metre squared... and that's what we are going to be doing with 5G."
Increasing the data-communications density will mean finding new spectrum and being as smart as possible about using existing frequency bands. The mobile industry and its confederates has already started down that road with 4G, by introducing small-cell basestations or 'femtocells' that work alongside 'macrocells', which cover much wider areas ['Feeding time', E&T, April 2013]. These very short-range basestations are designed to be packed into city streets in dense meshes, possibly hanging from street lamps or even deployed in users' homes where they double up as Wi-Fi access points.
The FON network, now owned by BT, provides an indication of how private access points can be used to provide a high degree of public wireless coverage - the service was rolled into the BTOpenzone service several years ago. Standard Wi-Fi itself provides the possibility to offload traffic from the 3G and 4G services.
One option is to move these basestations into hitherto unused parts of the radio spectrum. "As part of the 5G innovation work, we will look at new frequency bands," confirms Prof Tafazolli.
Frequencies above 20GHz - ten times higher than those used for 3G and Wi-Fi communications - offer massive potential data-rates, because the bands themselves are much wider. "If you go up a couple of orders in frequency you can go up a couple of orders of magnitude in bandwidth," explains Professor Ted Rappaport, director of the NYU Wireless research centre at the Polytechnic Institute of New York University.
The millimetre-wave bands - they range from around 3mm to 30mm in wavelength - are also practically unused for commercial wireless communication. There is a reason for that. Absorption by rainfall climbs rapidly from 2GHz to 100GHz, making this region of the spectrum unattractive for long-distance radio communication. It is also a difficult region of spectrum to serve. Only recently have low-cost silicon processes reached the level of development where they can be used in handsets that support such high frequencies.
Prof Rappaport contends it is a matter of distance. If you restrict the use of 20GHz-plus signals to relatively short distances, some of the problems go away. "It is a common myth that rainfall and oxygen absorption will attenuate these frequencies too much," says Prof Rappaport. "We've performed measurements to show it in one of the toughest radio environments we have: New York City."
Over distances of a few hundred metres, there is some loss - but far from enough to wreck the technology's chances. Says Oksanen: "In densely populated areas, that is already a long distance. Even macrocells are less than 400m apart in urban environments." There are also sweet spots in the spectrum, such as 28GHz and 38GHz, where Prof Rappaport's group has conducted experiments. "We will be measuring 72GHz this summer," he says.
Although there is a steady rise in absorption towards 100GHz, a number of the candidate frequencies lie in troughs between very strong peaks. One area that is badly affected is around 60GHz, a frequency now earmarked for automotive radar and indoor wireless networks. But either side of that range are frequencies that are far less affected by air absorption.
One issue with higher-frequency transmissions is that they are highly directional and work best where the handset has a clear line of sight to the basestation; but Prof Rappaport's group found the waves bounce off buildings providing multiple paths to a user even if they cannot 'see' the transmitter. "We've done research that shows you can get range extension to 400m by combining antenna paths," he reports.
To steer radio transmissions towards a receiver, Prof Rappaport envisages the use of beam-forming with multiple antennas, which are already being introduced on handsets on much lower frequencies to improve reception quality.
More spectrum required
As wavelength is inversely proportional to frequency, higher frequencies will make it easier to pack more antennas into the handset. Although designers are struggling to squeeze multiple antennas for sub-2GHz bands into extant designs because of the need to use structures appropriate for them, the wavelengths above 20GHz are at least ten times smaller. So-called 'massively MIMO' antennas, such as the 64-element structure used by Samsung in its 1Gbit/s transmission over 2km at 28GHz experiment, have become realistic. As they introduce higher frequencies for small cells, the industry will have the opportunity to reallocate spectrum to make best use of existing bands. In general, the lower the frequency, the further it tends to propagate.
"It's clear that the industry's direction is to have macrocells at as low frequencies as possible," says Oksanen at Nokia Siemens Networks. The lower end, however, is the most precious area of the radio spectrum and reallocation will not completely fix the problem.
"We are working on the spectrum front with customers and regulators and other industry stakeholders to find new spectrum in the low bands," Oksanen adds. "That is one of the important things about the future. It's not just about higher spectrum. We need new low-band spectrum."
Prof Tafazolli says: "We are really short of spectrum. We should not be limited to licensed, we could also use unlicensed spectrum." One way for cellular operators to use unlicensed spectrum, which allows anyone who keeps within power limits access to a band, is to use cognitive-radio techniques, in which transmitters constantly monitor other active radios and attempt to use the spaces between them, hence the term 'white space' radio.
The Weightless Special Interest Group is promoting this use of unlicensed spectrum for machine-to-machine communications, offering long-distance communications at low data rates ('Standard's net gains', E&T, June 2013). Such spectrum may not suit cellular operators, Oksanen points out: "It's difficult to invest in and use a band where you cannot guarantee quality of service to the end user."
There may be a middle way between dedicated and unlicensed spectrum. Oksanen says that he is "already working with industry stakeholders on how we can maximise low-band spectrum," and adds that "there are current users who have spectrum who don't use it all the time".
There are bands allocated to radar and wireless microphones, as well as other bands reserved by governments, that are not in use for 90 per cent of the time, according to Oksanen. "We are working on a regulatory regime and developing a mechanism whereby operators have guarantees that when they use it they can use it in the same way they use licensed spectrum, but the primary user can claim it back when they need it."
For sub-10GHz urban radio, as well as adding extra bands there could be changes to the way the data is transmitted. 4G uses orthogonal frequency division multiplexing (OFDM) - already used in Wi-Fi and wired broadband - to spread multiple data bits over a single band. One way to improve data rates is to have multiple basestations communicate with a single handset on the same frequency band - but synchronising them is tricky.
"OFDM requires a lot of management - but there are other potential solutions," says University of Surrey's Prof Tafazolli. "We don't have that technology yet, but I'm more in favour in other types of waveform that do not require strict timing and frequency synchronisation because they would reduce the management load."
Oksanen says: "There are some proposals for new coding methods' But when we look at whether we can do better than OFDM, there doesn't seem to be any significant improvement with these new methods. You find you can improve power efficiency, for example, but the spectrum efficiency goes down. We believe that OFDM is the best way to go forward - and it looks to be the most promising way for local-area 5G radio."
The industry has a while before it has to make a decision on what 5G means, but individual radio standards are only going to be part of the picture. "Cellular architecture needs to change. The legacy structures that we defined in 2G need to be revised," Prof Tafazolli concludes. "We need to have a better way of structuring communications between the basestation and devices."