This is the submission that I made to the UK National Infrastructure Commission in response to its Call for Evidence on 5G
1 of 12
Download to read offline
More Related Content
5G submission by Simon Pike to the NIC
1. 1
5G
Supporting the competitiveness of the UK economy
Submission by Simon Pike C.Eng MIET
1 Introduction
I am grateful for the opportunity to make a submission to the National Infrastructure Commission
study on how the UK can maximise the benefit from 5G. Almost all of the discussion and analysis of
5G to date has been done by industry stakeholders, researchers and bodies with interests in the
future of mobile; it is therefore welcome that the Chancellor has requested that this study will be
undertaken by a body that carries out independent and unbiased assessments.
The Chancellor has requested the NIC to “consider what the UK needs to do to become a world
leader in 5G deployment, and to ensure that the UK can take early advantage of the potential
applications of 5G services”. The second half of this request is the more important, because 5G
deployment is a means to achieve the objective of increasing competitiveness of the UK. In the
second half of the request, the key words are “early advantage” and “applications of 5G services”:
- “Early” is not the same as being first; for previous generations, the best position has been
fifth or later to launch – the first couple of networks to launch have used pre-standard
equipment, which has subsequently needed to be re-engineered for the finalised standards,
and the few after have borne the brunt of ironing out teething troubles.
- “Advantage” implies providing benefits to consumers and business users and ‘vertical’
industries (not just a status symbol).
- “Applications of 5G services” again indicates that the focus should be on the benefits to
users; obviously, the network deployment must support these applications, but the
deployment should not be an end in itself.
2 Some fundamental considerations
Before discussing the specifics of 5G, it is worth highlighting a few ‘home truths’. These may seem
obvious, but they often overlooked in discussions on 5G:
1) Historically, telecommunication services have grown exponentially – until they stop growing.
- It is therefore important to consider the factors that generate this growth, and their limits of
validity.
2) All services (current or envisaged for the future) that require a very high bit rate involve, in
one way or another, images – and almost always moving images.
3) The typical resolution of a human eye is very unlikely to change by 2030.
- This sets an upper bound on the useful video resolution, and therefore the highest bit rate for
video that there is consumer benefit in transmitting (see section 4 of this response).
4) Almost all video is consumed when the user is inside (either indoors or while travelling as a
passenger in a car or train).
- Therefore, the ability to serve indoor users will be critical for the success of 5G networks.
- If the base stations are outdoors, then the ability of the signal to penetrate indoors will be
crucial.
2. 2
- The high mm-wave frequency bands that are being considered for 5G have a very poor ability
to penetrate through windows and walls.
5) All services that require low latency involve a servo loop in one way or another.
- This servo loop might involve control of machinery or the relative position of vehicles.
- It might also involve humans, such as in interactive computer games.
3 What is 5G?
3.1 The definition of 5G
Before assessing how the UK can maximise its benefit from 5G, it is first necessary to consider what
5G actually is - or will be. There are many different visions of 5G, but they are mainly - explicitly or
implicitly - different combinations of aspects of growth and expansion:
- Growth in data traffic
- A more consistent user experience
- Higher bit rates
- Expansion into new market sectors (vertical applications)
- The resulting growth in operator revenue
- Growth in ARPU
- As a consequence of these two, expansion in the ability to invest in new infrastructure.
- Expansion of the ecosystem built around licensed spectrum
- Growth of the national share of the global equipment market (national industrial policy)
- Expansion and refreshing of IPR (patent) portfolios and licence fees.
- Growth in the sales margin for equipment supply, and expansion in the number of vendors
making a positive return on investment.
- A new radio interface and network architecture to deliver these.
- Re-defining the term 5G for other purposes (e.g. fixed wireless access)
Some of these aspects are appropriate for Government intervention; others are best left to industry.
Several might be negatively impacted by the unintended consequences of Government or regulatory
interventions in other areas of telecommunications or industrial policy.
It is therefore important for the NIC study to define clearly the definition of 5G that will be used in its
analysis.
3.2 Use Cases for 5G
A key driver for 5G in Europe is likely to be the support of new
applications in 'vertical' market sectors, many of which are listed in the
table. Most attention is focussed on a few of these, particularly
driverless cars and massive M2M, which are sectors with large global
players with long term research and strategy capabilities. However,
other verticals could also provide substantial opportunities for UK
industry.
It is generally agreed that the use cases for 5G fall into three broad
groups of capabilities: enhanced mobile broadband, massive machine-
type communications and ultra-reliable communications. These are
illustrated in the 'triangle diagrami
; the NIC will doubtless receive many
Potential vertical market
applications for 5G
Automotive
- Driverless cars
- Traffic management
Smart Cities
Energy
Massive M2M
Agriculture
eHealth
Railway (main line & light)
Factory automation
Logistics, retail & advertising
Multimedia & Broadcast
Public Safety
Education
3. 3
such diagrams in the responses
to its consultation.
Most M2M and ultra-reliable
applications will require very high
geographic coverage, which will
not be available in the early years
of 5G deployment. Many of them
could also be supported by 4G
technologies, as they will have
evolved by 2020. Most devices
for these applications will have a
substantially longer lifetime than
a consumer smartphone
(especially if they are embedded
in products). There is already demand for these applications, and many of them can be supported by
LTE or other technologies that are available now, or will be by 2020.
Therefore, by 2020, it is likely that there will already be thriving ecosystems for M2M and ultra-
reliable applications, using LTE and other technologies. It is unclear whether 5G will offer sufficient
benefits over these ecosystems, or enable sufficient new applications, to justify the substantial
investment needed.
The network capabilities needed for different types of application are very different, and the type of
network needed to support them will also be different. This is discussed in more detail in the annex
to this response.
3.3 Technologies for 5G
The ITU is currently developing a process for submission and evaluation of 5G technologies (which it
terms IMT-2020). It is likely that a technology, to be accepted as IMT-2020, will need to meet
performance objectives in a range of user environments (such as urban, suburban, rural and indoor).
It is expected that the 5G standards submitted for IMT-2020 evaluation will be ready in time for
commercial launches in 2020.
Three are currently two technology and standards ecosystems from which 5G technologies are most
likely to emerge, 3GPP (LTE) and IEEE 802.11 (WiFi and WiGIG). The 3GPP ecosystem is based around
a value chain of licensed spectrum and a small number of network operators (with customers having
a contract with one of them) - although some licence-exempt elements are now being added to 3GPP
standards. Most content is consumed by consumers on smartphones, and they can do so in a wide
range of environments.
On the other hand, the 802.11 ecosystem is based around a value chain of licence-exempt spectrum
and free access, although there are many providers of commercial WiFi access points. WiFi modems
are embedded in a wide range of consumer products, and are mainly used indoors and low mobility
environments. WiFi will soon be complemented by a new technology called WiGIG which operates in
the mm-wave bands that are also envisaged for 5G. WiFi and WiGIG are likely to have one or two
phases of technology enhancement by 2020.
The 802.11 family of technologies may not meet the criteria to be designated IMT-2020 by ITU,
because they are not designed to operate in a wide enough range of user environments, but they will
M.2083-02
Gigabytes in a second
Smart home/building
Voice
Smart city
3D video, UHD screens
Work and play in the cloud
Augmented reality
Industry automation
Mission critical application
Self driving car
Massive machine type
communications
Ultra-reliable and low latency
communications
Enhanced mobile broadband
Future IMT
4. 4
be very capable in providing connectivity in the environments and for the applications for which they
are intended.
Many of the developments in radio interface technology and network architecture that are being
researched and discussed for 5G can also be applied to 4G.
As discussed below, video is expected to form a large proportion of mobile traffic, and much of this
will be consumed indoors – the environment for which WiFi and WiGIG are primarily intended.
4 The importance of video
It is generally accepted that the majority of future mobile broadband traffic will be video, with some
studies predicting up to 85% by 2030. The trends in multimedia content production and distribution
therefore give valuable insight into the likely future demands on mobile broadband.
Most video content is currently consumed in standard or high definition format. Most large screen
displays now support Ultra High Definition (UHD – 4096 horizontal pixels), and an increasing
proportion of programme content is being produced in UHD format – for future-proofing of archive
libraries, if not for immediate transmission.
The development of the next step in video resolution – 8k – is being led by the Japanese broadcaster
NHK. It will produce content in 8K format at the Brazil Olympics, and plans to launch a commercial
service in around 2020. Major developments in broadcasting occur around once a decade, so this can
be assumed to be the upper bound in resolution of entertainment broadcast content up to around
2030. This requires a bit rate of around 120Mbit/s (8K, 100 frames per second sequential, HEVC
coding), which might drop by 50% over the next decade due to improvements in video coding
technology.
However, it is unclear how widely 8K video production will be adopted in Europe, due to the far
higher cost of production, and the benefit can only be seen on very large screens (bigger than around
60 inches in a domestic viewing environment). In any case, for many types of content, the actual
resolution will be substantially less due to external factors such as depth of field and motion blurring.
It is also unclear why 8K will need to be delivered using 5G, since the large displays will not be
portable (though there may be wireless connection between the display and other devices in the
same room).
If the highest resolution that will be viewed on a wireless device is UHD, then the bit rate needed will
be in the order of 10-25Mbit/s. When the video is viewed indoors – which is likely to be the majority
of the time - then they could equally receive content from terrestrial or satellite broadcasting, or
from a fixed or cable TV network.
5 The critical importance of spectrum
The availability of suitable spectrum is absolutely fundamental to the successful deployment of any
mobile network. By 2020, mobile network operators will be fully using their spectrum in all current
frequency bands. By this time, LTE carrier aggregation will probably be supported over three
frequency bands, which means that the user experience will equate to 2 X 30MHz or more of
available spectrum.
The 700MHz band will be released in UK in a similar timeframe to the probable launch of 5G services.
This has 2 X 30MHz of bandwidth, which means that the four UK MNOs will probably each acquire
2 X 5MHz or 2 X 10MHz. The performance of 5G is likely to be better than 5G, but this will not be
sufficient to offset a disparity in bandwidth of 3:1 or greater. Therefore, to offer a 5G service that
5. 5
matches its existing 4G service in performance, an operator would need to take some of its heavily
used LTE spectrum and refarm it for a 5G network that would initially not have many customers.
The ITU is currently studying a number of potential new bands for 5G, with the aim of some of them
being identified for IMT-2020 at WRC-19. They would then become available in the UK in the early
years of the next decade. However, these bands are all above 24.5GHz ("mm-wave bands"). This is
ten times or more higher than the current mobile bands, which corresponds to a wide gulf in
performance - the link budget would be many tens of dB less favourable than an existing mobile
band, corresponding to a cell radius of perhaps between 3% and 20% of a current small cell,
depending on assumptionsii
. A cell this small would contain only a small number of active users in
most environments, which would lead to typically low utilisation. mm-waves are therefore not suited
to providing the reliable ubiquitous coverage that many vertical applications need. It would lead to a
high deployment cost for network roll-out.
Therefore, there is no clear roadmap for the availability of spectrum needed for the successful launch
of 5G, either for the ubiquitous coverage for vertical applications or for enhanced mobile broadband
to serve indoor users in urban areas. The potential consequences of these are:
1) 5G technologies for existing frequency bands might be evolutionary from 4G, so that they can
share the same spectrum as LTE networks during the launch period (with LTE and 5G interleaved in
time and frequency).
2) WiFi and WiGIG might prove just as effective as 5G in delivering services to indoor users
(which is likely to be the majority of the consumer use).
6 The risks of headline target numbers
6.1 How long will exponential growth in traffic continue?
Most of the respected forecasts of data traffic, such as the Cisco Visual Networking Index, look
forward around five years. They use a model for different types of data traffic which grow at different
rates, but the aggregate traffic is generally found to grow roughly exponentially. It is therefore often
(but incorrectly) assumed that data traffic will continue indefinitely to grow exponentially.
For example, the ITU Reportiii
M.2370 on “IMT traffic estimates for the years 2020 to 2030” contains
an estimate of an average of 257.1Gbyte data traffic per device per month (4394 Exabytes/month
global traffic), of which 85% would be video. However, this study also assumes that the number of
devices will be twice the global population, and it is reasonable to assume that consumers will only
watch video on one device at a time. This is equivalent to every person on the planet watching
around 3-5 hours of individually streamed high definition video content per dayiv
delivered via mobile
networks (i.e. this number of hours excludes any viewing of content delivered by fixed or broadcast
networks, or multicast over mobile networks, or cached in the device).
While this is not physically impossible, it is implausible. It certainly would require a change in
consumer behaviour that goes well beyond what can be assumed through exponential growth.
6.2 The most meaningful parameter for speed is “user experienced” bit rate
Whatever the service that they are using, mobile consumers want a consistent user experience
wherever they go - i.e. a bit rate that is sufficient for the service that they are using. This is often
described as the "user experienced bite rate" or "sufficient bit rate". Radio propagation is subject to
6. 6
many variables, so this cannot be guaranteed but needs to be associated with a percentile (close to
100%) of availability.
However, mobile networks are often characterised by peak bit rate, which is only available close to
the base station (or perhaps only in the lab or in theory). This has a very poor correlation to what
users want. Also, the very high peak bit rates targets/requirements in some visions for 5G (10Gbit/s
to 50Gbit/s) would require a very high bandwidth of spectrum to support - bandwidth that would not
be justified by user demand.
Cell peak bit rate is not a meaningful parameter to characterise mobile networks, and should not be
used by NIC.
6.3 The myth of one millisecond latency requirement
It appears that the figure of one millisecond (1ms) latency originated as an objective for research, and
has morphed into a target or even a requirement for 5G standards without any proper analysis. I have
been unable to find any significant application for 5G that needs such low latency; here are a couple
of examples that are often quoted:
Car-to-car communications: At 70 miles per hour, one millisecond latency corresponds to 31mm
distance of travel; the braking distance according to the Highway Code is 75 metres (2400 times
larger, ignoring ‘thinking distance’).
Human interaction: It has been suggested that a very low latency for some human activities such as
interactive video games. However, the highest available refresh rate for computer monitors is 240
frames per second, which by itself implies a latency of 4ms (the total delay, from trigger to video
image, is likely to be more than 10ms). It is therefore unclear how this subjective requirement could
have been assessed.
It is worth noting that 1ms delay corresponds to the round trip over optical fibre of less than 100km
without any other delay. If a safety critical system requires a latency of 1ms, then the maximum
permissible duration of interruption of signal under any circumstance (including radio interference)
would be need to be comparable. The impact of interference has not, to my knowledge, been
considered at all in the development of 5G requirements.
7 Energy Efficiency
Telecommunications networks currently consume several percent of global electricity generation. It is
widely predicted that the traffic over these networks will increase by a factor of a thousand by 2030.
For the UK to meet its green energy targets, it will be important that the overall energy consumption
of networks does not increase, and hopefully decreases - i.e. a thousand-fold or more improvement
in energy efficiency.
Much of this improvement will come from technology development of the network hardware
elements and the type of network deployment, but this may not be enough by itself. As discussed in
the previous section, the size of 5G mm-wave cells is likely to be small if very high bit rates are to be
supported. These cells will, on average, be relatively lightly loaded with traffic, and therefore
relatively inefficient in energy consumption. Two networks serving the same area will have a lower
overall energy efficiency.
7. 7
Therefore, in the next decade, there is likely to be a tension between telecoms competition policy
(which favours multiple networks) and energy policy (which favours a single shared network).
8 Conclusions
The roadmap for the deployment of 5G is not clear at the present time, in terms of:
- how many applications that will not be supported by 4G, as it has evolved by 2020.
- the revenue streams that would fund the substantial investment needed for network roll-out.
- the spectrum in which it would be deployed.
A key reason for this is that the technical requirements for 5G were largely defined before there had
been any rigorous analysis of the business case or the likely applications1
. There has also not been
any significant analysis of the potential of mm-waves to provide services that would be attractive to
consumers. There has been little consideration of which applications would provide greatest
consumer or social value, and which would just provide an alternative delivery mechanism for
services that will also be provided by other means.
I would therefore suggest that a key aspect of the NIC study should be to undertake analysis in these
three areas - to identify the worthwhile use cases for 5G, and to separate them from the 'useless
cases' and the 'basket cases'.
The high envisaged peak bit rates for some 5G services could lead to a network with a very large
number of very small cells that are not efficiently used. This would be compounded if multiple
networks are deployed in the same geographic area. The NIC should therefore consider the extent to
which the current tenet of competition regulation, that multiple networks are desirable, is optimal
(or even viable) in the world of 5G services - especially given the tension with energy policy and the
green agenda.
1
An obvious example of this is the disconnect between the characteristics of multimedia consumption and the
predictions of future mobile broadband traffic.
8. 8
Annex 1 - 5G Capabilities
This annex provides more detailed information on the capabilities that 5G would need, in order to
support different types of potential application. The diagrams were developed by a group in the UK
Spectrum Policy Forum; they were approved as a UK contribution to ITU WP5D and are now included
in ITU Report M.2370iii
. The text description is my own.
The first four diagrams correspond to the three broad categories of application - enhanced mobile
broadband, massive machine-type communications and ultra-reliable communications. However,
mobile broadband has been separated into two parts, because multimedia is expected to form a
large proportion of mobile broadband traffic and it has different characteristics to other types of
data. The fifth diagram illustrates the capabilities envisaged in various industry visions for 5G.
In each diagram. the red envelope shows the capabilities generally needed by applications in that
category, and the coloured symbols show capabilities for specific applications and use cases.
A key message from these diagrams is that the red envelopes are very different in shape - which
means that the network needed to deliver them would also be substantially different.
A.1 Data and Voice
This category includes the capabilities that are likely to be needed to support data applications apart
from multimedia. It also includes some machine-to-machine applications (Massive M2M is included in
'sensor and actuator applications' and high performance applications are included in Mission critical
and low latency).
Moving platforms include trains and buses, where a radio link provides a connection from the vehicle
to the network, with a separate radio link providing the connection to the passengers in the vehicle.
9. 9
A.2 Multimedia
This category includes broadcast, on-demand and over-the-top video content and multimedia.
It is assumed that portable devices will have smaller screens, and therefore need lower bit rates. The
highest mobility is for trains, which would probably be connected with internal hotspots and an
external link to the network.
It does not include:
- 8k video, which will be predominantly viewed on fixed screens
- Telepresence and interactive gaming, which require low latency and therefore use video
codecs with lower bit rate compression.
- Download for later viewing, which does not require a constant bit rate.
10. 10
A.3 Internet of Things - sensor and actuator applications
This category includes sensor and actuator applications, including smart metering, home automation
and many smart city applications. They will communicate small amounts of data at infrequent
intervals (typically a few times an hour to a few times a day). Many of these devices will not have
external power supply, so they need to operate for many years from an internal battery. For tehse
devices to become ubiquitous, the unit cost will need to be low (a few pounds, or less for some
applications).
11. 11
A.4 Mission critical and low latency applications
This category includes a wide range of applications, mainly for industrial, business and professional
use.
The exception to this is interactive video gaming, which requires low latency for natural interactions.
This low latency prevents the use of the most efficient video codecs, because these use algorithms
that compress across a number of video frames - and the inherent delay is therefore at least the
duration of these frames. This results in a required bit rate that is substantially higher than for
programme content of the same resolution.
A low latency requirement is generally associated with the radio link forming part of a servo loop
between entities in close proximity. These are generally peer-to-peer applications with continuous
transmission, which do not need to involve a network. They are generally niche applications, which
may be better served by a dedicated technology.
12. 12
A5 New capabilities envisaged for 2020 and beyond
This diagram reflects a number of visions within the mobile sector and is different in nature to the
first four, which reflect applications for 5G that can currently be identified, and their likely growth in
the period up to 2030. This diagram speculates on the capabilities that may be needed beyond what
can currently be justified by analysis.
These visions have often started life as research objectives and then morphed into targets or even
requirements for 5G standards. It is therefore not possible to say with any confidence what these
capabilities will be used for, when they might be needed, and what economic or social value they
might generate.
i
Recommendation ITU-R M.2083-0; IMT Vision – Framework and overall objectives of the future development
of IMT for 2020 and beyond; Figure 2 - Usage scenarios of IMT for 2020 and beyond.
ii
The link budget would be less favourable by around 20dB due to higher transmission bit rate, 20dB due to
building penetration loss, 10dB due to lower performance of technology at mm-wave frequencies, and more
favourable by around 10dB due to higher effective base station antenna gain (rough figures to give an order-of-
magnitude estimate of the difference).
iii
ITU Report M.2370; “IMT traffic estimates for the years 2020 to 2030”
iv
This assumes a bit rate of round 6-10Mbit/s for HD video, taking into account the likely improvement in video
coding performance by 2030.