A year ago 5G appeared to consist of a ragbag of aspirations and disparate use cases with such different performance criteria attached it seemed impossible that any one standard could encompass them all.
At one end are high-performance, high-throughput, low-latency, ultra-reliable applications to support the likes of autonomous driving. At the other end is a wide range of low-power, low-throughput Internet of Things (IoT) applications such as smart meters, which only have to transmit data occasionally.
But a lot has happened in a year, as Dr. Håkan Andersson, Head of Mobile Broadband, Radio Business Unit at Ericsson, says: ‘I think there has been a lot more consensus as to what 5G is over the past year and when it comes to the use cases a fairly large common understanding of what will be required from the standard.’
David Hutton, Head of Networks at the GSMA, agrees: ‘I think there had been a lot more clarity added to the discussion. Standardisation work has started now, so it is not just about aspirational views. Previously, the generational step focus was on the mobile broadband experience and 5G will encompass that, but 5G is more of a whole ecosystem approach.’
Key 5G requirements
The vision for that 5G ecosystem looks to have coalesced around several key requirements:
• Much greater throughput (download speeds of up to 20 Gbps – an entire HD film in a second)
• Much lower latency (sub-1ms end-to-end round trip delay) – 50 times faster than 4G, and required for applications such as mission-critical SCADA systems and vehicle-to-vehicle communication
• Ultra-high reliability, availability and security – a network that does not ‘break’ and is therefore suitable for mission-critical applications
• Much higher connectivity density – ability to support massive IoT deployments
• Higher mobility range – seamless, uninterrupted connectivity – in a high speed train, for example
• A 90% reduction in network energy use.
‘There is an evolutionary and a revolutionary part of 5G,’ explains Andersson. ‘There’s a lot of things that are precursors to 5G that are also being discussed as components to 5G.’
These precursors include the likes of network function virtualisation, (NFV), software defined networking (SDN), mobile edge computing (MEC), and narrowband low power, wide area networks for Internet of Things (IoT).
‘Concepts like narrowband Internet of Things (NB-IoT) for massive IoT deployments is one of these evolutionary proof points,’ says Andersson, ‘and the standard is already being worked on by 3GPP, so that’s something that spans the generational shift.
‘In previous generations, the focus was on the technology and modulation scheme and this is how they defined the generation. But now it is more about what use cases are we trying to solve?
‘I see 5G as a vehicle to enable us to go after these use cases. So, with 5G there is not so much focus on labelling a new generation or wave modulation, but on establishing a much wider range of technologies and use cases,’ argues Andersson.
‘5G will not be just one big bang introduction,’ he continues. ‘We need to establish something that has forward compatibility.’ There is no shortage of views as to how this might be achieved, with numerous bodies, operators and equipment vendors working on different aspects of 5G research.
A new report published in January 2016 by the 5G Innovation Centre (5GIC) at the University of Surrey in the UK entitled, 5G Whitepaper: The Flat Distributed Cloud (FDC) 5G Architecture Revolution, provides a good summary of the main thrust of what many people think the 5G requirements will look like.
The paper states: ‘The foremost requirement is that the 5G infrastructure should be far more demand/user/device centric with the agility to marshal network/spectrum resources to deliver “always sufficient” data rate and minimal user plan (UP) latency (subject to use-case) so as to give the end-user the perception of an infinite capacity environment.
‘Thus a new architecture is expected to address enhancements in terms of:
• Flexibility: it should be easy to introduce new services, software upgrades and change traffic management policies and systems
• Complexity: should be reduced in terms of implementation, deployment and costs structures
• Performance: should be scalable, routing unlimited, UP and CP latency according to use case and traffic management made simple to set, monitor and adjust.
5GIC notes that a key difference of approach to 3G and 4G is that it envisages that ‘text generation mobile networks should be able to present multiple slices of the network to different users and/or different services for the same user depending on their usage context’.
The 5GIC has indentified three key tenets that a 5G network should provide: the perception of infinite bandwidth; always connected capability; and tailored context awareness. In defining its proposed flat distributed cloud (FDC) network architecture, the 5GIC sets out seven main requirements:
• Provision of distributed cloud-based services and architecture flexibility/evolution in a timely manner
• Support for content centre networking
• Support for integrated IoT across a number of different IoT system types
• Support for Context Aware Networking supporting both user and network 5W’s to optimise use of available network resources and the service experience delivered to the user
• Support for low latency services and download times significantly reduced in order to support Cloud computing and fast access IoT devices
• Provision the user with a universal network capability over as large a coverage area as possible
• User plan control to be more efficient than the GTP/ESM approach of LTE.
The end result of these requirements is that the ‘proposed architecture is designed to be able to adapt to user service requests according to user and network context, and be able to respond to demand such that user perception is always managed to best meet each request with the available resources’.
Li-Ke Huang, Research & Technology Director at the test side of Cobham Wireless, observes: ‘It is very important that the 5G infrastructure is able to create many more services than traditional networks such as mobile broadband, IoT, high reliability, low latency applications and so on. So, a different core technology is being proposed to meet all these requirements.
‘There is some divergence of opinion in terms of what core technology goes into this, although there is a lot of common ground here too,’ notes Huang. ‘But it must support a diverse level of services and so it needs to be very flexible. The underlying infrastructure must be very programmable and to make that feasible we need software defined networks.’
The 5G network will need to harness all the resources available to become a multi-radio access architecture, so it can intelligently deploy the most appropriate network resources in whatever part of the network and in the most efficient way possible to meet the demands of the required service.
In a blog posted on 14 October 2015, Matt Branda, Staff Manager, Technical Marketing at Qualcomm, said that new 5G multi-connectivity technologies will support simultaneous connectivity and aggregation across 5G, 4G LTE and Wi-Fi technologies with a multi-access 5G core network.
‘The new architecture will deliver mobility-on-demand by distributing network functions at the core or edge of the networks based on the service requirements and device context, while taking advantage of emerging virtualisation technologies to generate network slices optimised for the target services or deployment types,’ said Branda.
Network slicing involves partitioning a single physical network up into many virtual networks to enable the mobile operator to offer the best support available for different types of services and for different types of customers or industry verticals – ‘networks on an as-a-service basis’, as Ericsson puts it.
Pointing out the advantages of this for mobile operators, Ericsson’s Andersson says: ‘These software defined virtual networks allow operators to try new use cases without a heavy investment in hardware, because these virtual networks, or network slices, run in parallel on top of the common network.
‘It is a lot more agile therefore, and shortens the time needed to go to market. You could even create a network for a temporary purpose by defining one for a particular service and then get rid of it once it has served its purpose.’
This will certainly appeal to mobile operators and others, especially when coupled with mobile edge computing (MEC) – adding more processing power and storage at the edge of the network, thereby cutting latency. And by opening up their APIs, operators can enable third parties to develop their own potentially revenue-generating applications at the edge (for more on MEC see article on P.12).
A standard for MEC is being worked on by 3GPP and Hutton observes: ‘Mobile operators do not want to have to wait for the 5G standard to be finished to open up these revenue-generating opportunities. MEC can be implemented sooner than the 5G standard will be ready, for example.’
Mansoor Hanif, Director of RAN Development & Programmes at UK mobile operator EE, agrees that MEC will open up a lot of revenue generating possibilities. What he is looking for from 5G is a network that is ‘flexible, programmable and ultra-reliable’.
The latter requirement is particular pertinent to EE, as by mid-2017 it will be the first commercial mobile operator in the world to run a country’s nationwide emergency services communication network – and for that, reliability is an absolute must.
‘We are pragmatic,’ he says. ‘We like to talk about what we can do in the next two to three years, rather than in 10-15 years’ time. We need to cut the industry innovation life cycle, which is about 10 years at the moment, but the likes of Instagram and Facebook move in three- to five-year cycles. The challenge is how can we move at that speed too? By making 5G flexible, virtualisable and programmable we may be able to achieve this and that is what we are pushing for.’
Hanif sees NFV and SDN as the key building blocks, arguing that if you want full SDN you need to make it programmable. ‘We need to break it down into elements and make things as COTS (commercial off the shelf) as possible, and commoditise the hardware, but support virtualisation, especially at the edge of the network.
‘NFV has focused mostly on typical network functions, but how far can we break that down and separate it even in the radio part of the network, so it is not just network functions, but radio characteristics too?’ he asks.
‘By making 5G completely programmable and virtualisable we can split the network into different SLAs and service guarantees for our customers in a very organised, but totally standardised way that also enables us to build things ourselves,’ says Hanif.
A new air interface
Judging by all the above, some existing air interfaces will be able to handle some of the proposed 5G use cases, but it is clear that some will require a new air interface (including signalling, modulation schemes and other software-driven innovations).
A new interface will certainly be needed if, as expected, millimetre wave frequencies up to the 70 GHz band are pressed into service to meet the additional capacity requirement created by the vast amounts of data traffic 5G is expected to generate.
But what that new air interface will be is still one of the great unknowns. Alcatel-Lucent has proposed a modification of the OFDM (orthogonal frequency-division multiplexing) LTE interface it calls universal filtered OFDM (UF-OFDM). Others have pushed for a new modulation called Orthogonal Time Frequency and Space (OTFS).
Huawei has proposed its version of F-OFDM, while Samsung is touting a variation of a waveform called FBMC (filter bank multi-carrier). Ericsson has something it calls
NX, which supports multipoint connectivity with distributed MIMO.
Instead of the device just connecting to one cell at a time and using various techniques to ensure a seamless handover to the next cell as the user moves around, a 5G mobile device connects to several 5G cells at the same time – multipoint connectivity in other words. What this does is ensure a more resilient high-quality connection.
It also allows the transmission of different sets of data signals – MIMO (multiple input, multiple output) to the mobile device over the same radio frequency channel – distributed MIMO.
Andersson says Ericsson has not gone for filtering in the frequency domain, but is suggesting manipulation in the time domain, as it is a less process demanding technology. ‘But the end result is the same,’ he says.
Cobham’s Huang reports that China Mobile is proposing a programmable interface where you use different waveforms, but one management system. Andersson agrees that all the different terminologies being bandied about for the new air interface are confusing,
However, he notes: ‘It is more about trying to put different labels on much the same technologies. It is the same basic idea, but using different ways to achieve the same result – companies have a need to put their own label on things.’
As already noted, 5G will bring with it an even greater demand for more capacity. This can be met through finding more spectrum, using existing spectrum resources more efficiently, or adding more infrastructure – generally in the shape of small cells.
Finding globally harmonised spectrum for 5G roaming is going to be challenging, and the GSMA is keeping a keen eye on this. ‘We are also looking at the definition of roaming and interconnect of 5G – does it need to be changed to meet commercial requirements?’ says Hutton. ‘From a GSMA perspective, we are trying to gain an insight into what the roaming interconnect for 5G will look like, and feed that into the standardisation process.’
New spectrum is most likely to come from above 6 GHz, possibly microwave bands such as 6 GHz, 28 GHz and 38 GHz, or perhaps from within the millimetre wave bands (60-80 GHz) – although which millimetre bands might be allocated to mobile services will not be looked at until the ITU World Radio Communications Congress in 2019, so they will not form part of early 5G – which isn’t to say there is not plenty of research going on into millimetre wave channel models and propagation characteristics.
Millimetre wave frequencies have been largely ignored by the mobile industry up until now (except by some small cell wireless backhaul vendors and 802.11ad Wi-Fi) as the perception was that the bands suffer from excess oxygen absorption and/or attenuation in rain, reducing their already short propagation characteristics.
But recent studies have shown that while these losses are present the attenuation is relatively slight for the distances required; i.e. between one base station and another (less than 1km generally). Nonetheless, Huang says: ‘There are a lot of challenges associated with making millimetre wave technology work for mobile handsets.’
As already mentioned, 5G also brings the prospect of multi-user MIMO where instead of a single antenna in the receiver and transmitter hundreds of mini-antennas could be deployed to boost data rates and increase spectral efficiency, making more efficient use of existing spectrum holdings, as well as being central to millimetre wave technology.
‘Multi user-MIMO and massive MIMO use many more antennas than now to achieve better spectral use of the frequency. And then there is beamforming, where the beam is narrowed down to just be directed to a single device,’ says Huang.
One completely new development with 5G is that other industries are feeding ideas into the mix, along with the standards bodies and vendor community, so that the end users of 5G are actually helping to influence the design of the new network standard they will end up using.
‘There is a lot of engagement with industry to help build proof of concepts,’ reports Andersson. Ericsson has been working with the mining and construction industries, and ports looking at remote operations, as well as surveying and inspection on land and sea using remotely controlled drones, robots and vehicles. Other areas include the oil and gas industry and remote surgery.
The testing industry is also being put on its mettle. ‘With 5G we need to provide a good-enough test system to meet the demands of a very complicated system, and that is a big challenge,’ says Cobham Wireless’ Huang. ‘We have to enable the industry to prove its proof of concepts, so we have to understand the technology early to know what tests we need to develop to support the industry.’
There has clearly been a lot of progress in 2015 in developing what 5G will look like, at least in terms of determining the main outlines of the standard. 2016 will no doubt see many of the concepts reaching a closer definition, but there is still quite a way to go yet before the promises of 5G can be fulfilled – and we are some way off knowing whether they all can be.
There is general agreement on the standardisation timeline now following the International Telecommunication Union (ITU) meeting in San Diego, California in June 2015. The ITU refers to 5G as IMT-2020 and defines it as capable of transmitting data at up to 20 Gbps.
The 2018 PyeongChang Winter Olympic Games in South Korea are expected to see the first commercial demonstration of 5G, or something approaching it, anyway. A first real commercial deployment is expected in 2020.
The standards body 3GPP will do much of the technical standard writing focusing on four main areas: radio and core network; enhancing mobile broadband; critical communications; and network operation.
It held a 5G workshop in Phoenix, Arizona in September 2015. According to an accompanying 3GPP release: ‘There is an emerging consensus that there will be a new, non-backward compatible, radio access technology as part of 5G, supported by the need for LTE-Advanced evolution in parallel.’
The Workshop Summary stressed the need for ‘forward compatibility to be a design requirement for the new radio from the get-go’, with the study to ‘include careful investigation of design options to ensure forward compatibility for all use cases’.
5G specification work will be done in two phases with Phase 1 to be completed by H2 2018 (End of 3GPP Release 15); and Phase 2 to be completed by December 2019 for the IMT-2020 submission and to address all identified use cases and requirements (End of 3GPP Release 16).