Ofcom. Final Report Predicting Areas of Spectrum Shortage 7 April 2009

Transcription

1 Ofcom Final Report Predicting Areas of Spectrum Shortage 7 April 2009

2 Ofcom Final Report Predicting Areas of Spectrum Shortage 7 April 2009 Private and Confidential PA Knowledge Limited 2008 Report Prepared for: Prepared by: Final Report Ofcom Phil White David Ramsbottom Simon Mardle Melanie Cook PA Consulting Group Cambridge Technology Centre Melbourn Herts SG8 6DP Tel: Fax: Issue A

3 EXECUTIVE SUMMARY The Need to Examine Potential Shortages of Spectrum Part of Ofcom s role as Telecoms regulator in the UK is to ensure that sufficient radio spectrum is available to meet the needs of the UK, both now and in the future. To inform this requirement there are a number of reports available which forecast future demand for services, although these reports generally do not take into account the implications of this changing demand on spectrum, nor the effects that new technology may have on the amount of spectrum required. In order to gain an overall perspective of demand and supply and ensure that sufficient spectrum is available in a timely manner, it is necessary to have a somewhat wider understanding of the future usage of spectrum. It is generally recognised that demand for services is complex and uncertain, and similarly that the introduction of new technologies is also to some extent unpredictable. However, even though the future requirements for spectrum cannot be known exactly, it is possible to model a set of potential future scenarios in order to identify if and where spectrum shortages are likely to arise, highlighting the high risk frequencies and services at particular points in time - this is approach that we have taken in this study. The objective of this study was: to build a model capable of considering a number of scenarios of both the growth in spectrum demand based on application growth and also the ability of networks to respond to that demand based on typical investment timelines, speed of standardisation, etc. The model can then be used to understand under which scenarios and timescales congestion occurs. Possible regulatory action to avoid these periods of congestion can then be examined. The work that we report on here enables Ofcom to: Recognise emerging technologies which will have the largest effect on spectrum Inform the market about potential future spectrum shortages in order that network operators and others can plan their businesses more effectively, to the benefit of the UK economy Determine the best position to take in international negotiations on spectrum harmonization through demand prediction Anticipate how spectrum demand may change in the next years. The Services Studied In our consideration of services that require significant amounts of spectrum, and that are likely to be of most interest in understanding potential shortages in the future we have considered the following: Cellular, Short Range Wireless, Broadcast TV, Broadcast Radio, Fixed Wireless Access and Backhaul. Of these services it is apparent that Cellular and Short Range Wireless are of most interest since it is for these services that demand is likely to change significantly over the next years. i

4 Executive Summary Spectrum demand for Broadcast TV could also increase significantly although we recognise that this can be overflowed to satellite transmission and hence is likely to be of lower priority than those services that fundamentally require mobility. We do not believe that Broadcast Radio will demand significantly more or less spectrum over the period. The Applications and Technologies Studied Whilst voice has traditionally been the main application for cellular services, this is changing rapidly with the growth of 3G data cards, video streaming and download applications. For Short Range Wireless such as WiFi the main applications are datacentric as a wireless tail from fixed lines to PCs in the home and public areas. Other innovative data-centric applications continue to be introduced and are likely to dominate traffic growth over the next few years. We have therefore studied a range of applications ranging from those available today to some which are inherently speculative, including a number of breakout options such as 3D television, growth of mobile virtual worlds etc. Our primary forecasts for application demand growth have come from previously published results, extrapolated as necessary to Sources such as earlier Ofcom reports, UMTS Forum and WiMAX Forum have been consulted and incorporated into the model. We have then amalgamated these applications into a number of scenarios and applied growth rates to them according to the scenario of interest. The demand forecasts that we have taken as inputs to this work predict that the main applications driving growth in spectrum demand for cellular will be video streaming and downloads, with increasing demand for machine-machine communication. Voice will cease to dominate as these new applications grow in popularity. The main applications that will drive spectrum demand for Short Range Wireless will be web browsing, , gaming, and video streaming, recognising that much of this traffic is in the home. In Figure A below we show the baseline Business As Usual traffic forecasts that we have used in this study. Figure A: Total Offered Traffic (Terabytes) in Business As Usual by Service (log scale) ii

5 Executive Summary Building on these forecasts for demand growth in traffic over the period from 2008 to 2025 and the likely changes to applications and networks that may occur during this period, we have then taken into account other factors such as technology improvements as new data compression technology is introduced, and as current cellular networks are increasingly replaced by 3G and other emerging standards. We also look at changes in supply of spectrum, delays in re-farming bands, and emerging standards, to understand the pressure on spectrum that is likely to result under a number of different scenarios. Before discussing the results, it is important to note that demand for spectrum is not absolute, but depends on a wide range of factors including price. So whilst we discuss potential shortage of spectrum in this report, this needs to be interpreted as a shortage according to a particular set of assumptions. In practice, if spectrum is in relatively short supply, networks operators will tend to introduce more cell sites, and will introduce fewer advanced high traffic services (or will tariff them very highly). Similarly if spectrum is readily available, networks will tend to be less dense and operators will be more generous with their offerings. The reader should therefore not interpret potential shortage in this report as meaning that consumers will suffer poor Quality of Service, nor attempt to draw conclusions about appropriate price levels for spectrum from this report. Pressure on Spectrum Our Findings We have built a number of scenarios within which we have studied demand and supply growth for wireless spectrum. These range from a central Business as Usual view up to All You Could Want which assumes very high growth for data-centric applications and rapid progress in network technologies and roll-outs driven by this demand. On the more conservative side we have two scenarios which model the cases where demand growth starts to fall off, and network growth slows accordingly. Cellular For the three high growth scenarios of Business As Usual (baseline), Wire Free World and All You Could Want, we believe that there is likely to be significant pressure on spectrum over the next 3-4 years as take-up of data services increases ahead of new network rollout. This will tend to limit growth of new applications until new spectrum and improved network technologies become available The pressure on spectrum is likely to become less acute from around 2012 because of the shift from 2G to 3G+, and subsequently (in 2014) because of the Digital Dividend release. Beyond 2014 increases in demand for cellular services are largely matched by the availability of spectrum, except in our most aggressive growth scenario. This arises because of a combination of two factors: the growth in demand for services is paralleled by growth in spectral efficiency of advanced technologies and use of smaller cells, and there is additional spectrum made available through this period. There is also an increase in spectrum availability beyond 2010 because of the 2.6GHz band becoming available and an additional block of spectrum assumed to become available from public sector release (MOD) providing capacity from 2015 / 2016 in the 2.7-4GHz region. However, we note that where spectrum is refarmed from one service to another, the time taken to clear the band and introduce a new network introduces significant inefficiency in spectrum use. iii

6 Executive Summary A key feature in reducing pressure on spectrum for cellular services is the improved efficiency of newer standards compared with GSM. Moving from GSM to 3G and beyond we expect major improvements in efficiency arising from the use of cdma technologies and more sophisticated coding techniques. Without these improvements it would be very difficult to provide sufficient spectrum for advanced high-speed data services. In Figure B we show our forecasts for Cellular spectrum demand under the different scenarios examined. This shows the short-term pressure that we expect over the next 4-5 years, a subsequent easing as new technologies are introduced, and renewed pressure after 2017 in high growth scenarios. Figure B: Demand for Cellular Spectrum (All bands, 100MHz 4,000MHz) Short Range Wireless We are forecasting very high growth in demand for spectrum for Short Range Wireless services in several scenarios, resulting in a clear shortage of spectrum in approximately This arises as a result of very high growth in fixed line data traffic and Short Range Wireless being used as the link to the PC or other terminal in the home and other locations including work and leisure venues. The short range nature of WiFi and UWB and the very high burst rates that these technologies aim to deliver mean that this pressure on spectrum will however result in a reasonably graceful degradation of network performance as congestion starts to have an impact, reducing data rates and perhaps having a knock-on impact on the volume of offered traffic. The problems may be localised, confined to a small percentage of the country, but, because demand tends to cluster in time and space, it will affect a significant proportion of users of the technology in the very high growth scenarios. Although we believe that there is sufficient spectrum for WiFi today, this seems to be contrary to experience in dense urban areas where congestion occurs. We believe that a significant amount of this congestion is due to inefficient use of spectrum, the number of different networks and limits on user numbers rather than spectrum shortage per se. This will become an increasing issue in the short and medium term as traffic grows. Increasing introduction of 5.8GHz systems will alleviate this, as could moves to licensed bands for business critical services. iv

7 Executive Summary A complicating factor for Short Range Wireless spectrum demand will be the extent to which UWB is successful. There is significant UWB spectrum available today that is not used (it is shared spectrum and cannot be used for other applications). If widely adopted it could take some of the very high bandwidth short range traffic from WiFi, although range and propagation through walls will be a limiting factor. In Figure C we show our forecasts for Short Range Wireless spectrum demand under the different scenarios examined. Broadcast TV and Radio Figure C: Demand for Short Range Wireless Spectrum The high growth scenarios that we have modelled do allow for a dramatic increase in the number of broadcast TV channels. This means that spectrum demand for Broadcast TV channels could also increase significantly. However, we assume that this can be provided over satellite and therefore do not consider that it will create significant pressure on spectrum competing with applications that are fundamentally mobile and which have no alternative means of delivery. Additionally, increasing moves to HDTV coupled with increasing subscriptions to premium satellite channels could mean that Digital Terrestrial TV increasingly caters for fewer subscribers. Fixed Wireless Access For Fixed Wireless Access the majority of demand is in rural areas and we are currently forecasting significant spectrum surplus for the foreseeable future. Only beyond 2019 do we see a significant issue, and only then under highest growth scenario. Should the demand for communications take off to this extent, it is likely that we would see a much wider roll-out of fibre to the home, which in turn would reduce the pressure on Fixed Wireless Access as a technology Technology Improvements It is clear from the results that improvements in technology are critical to meeting the demand for spectrum over the coming decade. Previously published results for growth in traffic demand show factors of ten or more compared with today; in the higher growth scenarios this can be much higher as use of wireless data become ubiquitous. v

8 Executive Summary Although it is possible to increase supply of spectrum by a significant amount, it is difficult to achieve much more than a factor of 2-3, partly because of the short ranges exhibited in the higher frequency bands. Similarly it is difficult to envisage all this additional capacity in the wide area being provided by additional macrocells the cost, logistics and acceptability of these are already partially limiting factors. This leaves a very large gap to be made up by the introduction of new technologies that improve spectral efficiency, such as newer modulation formats, improved reuse etc. Specifically we assume that improvements in cellular technology will continue to be introduced, improving both data rates and spectral efficiency. Additionally, use of improved on-air protocols will be important for Short Range Wireless if the bands are to be used more efficiently. We have assumed in this work that these improvements come into being over the next 5-10 years, and are adopted by the industry and users alike. However, it is clear that if this does not happen then lack of capacity could form a barrier to growth. An additional concern is that it is difficult to drive improvements in technologies for unlicensed bands, since the market here is uncontrolled and very cost sensitive. Low cost WiFi systems, security systems and audio / video repeaters use these bands and continue to be introduced in high volumes. Ultra Dense Urban Areas The majority of this study has considered Dense Urban as the neighbourhood type ( geotype ) of main interest, with Dense Rural being used for Fixed Wireless Access since that is where the bulk of FWA subscriptions and traffic is expected to occur. Ultra-dense Urban is however the limiting case, and is exemplified by certain areas of the City of London, transport termini, sporting venues etc. However, it should not generally be taken as the driver for spectrum demand since population and traffic densities in these areas are sufficiently high to justify taking special measures to deliver services. Cellular provision in Ultra-Dense Urban areas is today provided by significantly increasing the number of cell sites, either through use of microcells, or through remote radio heads (equivalent to picocells). Cell spacing can be as little as a few hundred metres, and buildings will have indoor coverage specifically designed to provide very high capacity through use of low power cells. We believe that similar techniques will continue to be used in future, with increased emphasis on more micro cells and pico cells, coupled with femtocells when these become widely available to provide the very high capacity, very short range, coverage that data-centric applications may require. The widespread use of WiFi that we see today will continue and grow in urban areas. This poses specific difficulties, since there is no management of quality between the very many independent service providers (which include various cafes, venues, corporate offices as well as the public service providers such as Openzone and T-Mobile). We expect continued pressure on traffic within the 2.4GHz band, with congestion and Quality of Service issues growing over the next few years. vi

9 Executive Summary Overall Conclusions and Recommendations Our conclusions are as follows: There is a significant risk of short-term increases in spectrum scarcity arising from transitions between different technologies, particular in the cellular domain. These occur under moderate to high growth assumptions. In the medium term these pressures ease as new network technologies are introduced, and as additional spectrum is released for cellular services. In the longer term it appears that improvements in spectral efficiency and the move to higher density network architectures will provide sufficient capacity to handle most high-end predictions of future demand, although there could still be limitations due to pressure on spectrum in the 2020 timeframe.. The use of Short Range Wireless Systems will increasingly lead to congestion in the unlicensed bands, with negative effects both on consumers and on organisations that use WiFi and similar systems for business-critical applications. Recommendations We include a number of recommendations to Ofcom in this report, the main ones being: Ofcom should examine closely the likely pressure on cellular spectrum that we predict may occur over the next 3-5 years. Depending on the rate of growth of cellular applications this may require more rapid rollout of 3G than has previously been envisaged. Ofcom should consider the extent to which additional spectrum may be released for use in Dense Urban and Ultra-Dense Urban areas. This might be from Defence use 1 or from existing civil users in the bands above 1GHz. We note that there is less pressure on spectrum in rural areas and that regional release covering major towns and cities would suffice to significantly improve capacity of networks. Within the scenarios presented here, femtocells are not delivering a significant capacity gain for 3G network operators, primarily because a large fraction of cellular traffic is unsuited to femtocells, and they require a dedicated (5MHz) channel which detracts from the operator s overall network capacity. We recommend that Ofcom consider this issue to see whether more should be done to stimulate use of 3G femtocells through alternative spectrum allocations (e.g. shared channels for low power 3G applications). We also recommend that Ofcom repeat this work in , to track progress against this report. By doing this Ofcom will be in a position to understand which of the scenarios we have presented is most likely to come to fruition, and the extent of pressure on spectrum over the period. 1 PA Consulting Group recently completed a report for UK Ministry of Defence, reviewing the current and future use of Defence spectrum: nications/publicconsultations/ukdefencespectrummanagement htm vii

10 TABLE OF CONTENTS Executive Summary i 1. Introduction Predicting areas of spectrum shortage in context The complexity of predicting areas of spectrum shortage Effects of pressure on spectrum Organisation of this report Scope of the Study The types of networks and services covered by the study Services, applications and technologies Frequency bands Spectrum use today Our Approach The approach taken to quantifying predictions of spectrum demand and spectrum shortage Forecasting the demand for services Forecasting the relationship between demand for services and demand for spectrum Pressure on spectrum Alternative Scenarios That May Unfold in the Next Years Drivers of change Defining the future scenarios Implications of scenarios on supply and demand of capacity A comparison of the scenarios and their implications Assumptions for Forecasting the Demand for Services Cellular Short Range Wireless Broadcast TV and Radio Fixed Wireless Access (FWA) The Business As Usual Baseline Scenario Demand for services and supply of spectrum Business As Usual - results Demand Growth and Spectrum Impact for Individual Services Introduction Understanding the main sources of the gains in efficiency in the networks Cellular Growth Growth in Short Range Wireless Growth in Broadcasting - TV Growth in Broadcasting - Radio Growth in Fixed Wireless Access Demand for Spectrum All Services and All Scenarios Ultra-Dense Urban and Other Neighbourhood Types Spectrum Demand in Ultra-Dense Urban areas Conclusions and Recommendations Conclusions Recommendations 10-4 viii

11 TABLE OF CONTENTS APPENDIX A: Assumptions for Network Architecture and Spectrum Use for Forecasting the Demand for Spectrum APPENDIX B: APPENDIX C: APPENDIX D: APPENDIX E: APPENDIX F: APPENDIX G: Demand Forecasts Sources Spectrum Demand in Wire Free World The All you Could Want Scenario Scenarios Paper Stakeholder Meeting Notes Model Specification ix

12 1. INTRODUCTION 1.1 PREDICTING AREAS OF SPECTRUM SHORTAGE IN CONTEXT Part of Ofcom s role as Telecoms regulator in the UK is to ensure that sufficient radio spectrum is available to meet the needs of the UK, both now and in the future. To inform this requirement there are a number of reports available which forecast future demand for services; however, these reports generally do not take into account the implications of this changing demand on spectrum, nor the effects that new technology may have on the amount of spectrum required. Although the future requirements for spectrum cannot be known exactly, it is possible to model a set of potential future scenarios in order to identify if and where spectrum shortages are likely to arise, highlighting the high risk frequencies and services at particular points in time. The objective of this study as set out by Ofcom was therefore: to build a model capable of considering a number of scenarios of both the growth in spectrum demand based on application growth and also the ability of networks to respond to that demand based on typical investment timelines, speed of standardisation, etc. The model can then be used to understand under which scenarios and timescales congestion occurs. Possible regulatory action to avoid these periods of congestion can then be examined. The work that we report on here enables Ofcom to: Recognise emerging technologies which will have the largest effect on spectrum Inform the market about potential future spectrum shortages in order that network operators and others can plan their businesses more effectively, to the benefit of the UK economy Determine the best position to take in international negotiations on spectrum harmonization through demand prediction Anticipate how spectrum demand may change in the next years. 1-1

13 1. Introduction 1.2 THE COMPLEXITY OF PREDICTING AREAS OF SPECTRUM SHORTAGE The purpose of this study has been to design and build a model capable of considering a number of potential future scenarios covering both the growth in spectrum demand based on new and/or existing application growth, and also the developments in technologies, networks, and spectrum availability, to provide the additional capacity needed to meet that demand. The model provides an analysis framework for each of the potential future scenarios over the next years, to enable an understanding of which of these future scenario(s) could result in a shortage of spectrum and in what timeframe it is likely to occur. The model takes account of factors identified in previous studies relating to spectrum demand in the UK together with the issues identified during this study. These factors can broadly be considered in four categories. Demand side -The demand side considers growth of traffic demand for different applications, including mobile voice, data, multimedia, video-on-demand to handsets, fixed wireless access, broadcasting, etc. Spectrum usage - Whilst spectrum is a fixed resource, and overall supply doesn t change, the important parameter is the amount of spectrum that is usable for an application in a given area, and the data carrying capabilities of that spectrum. This is affected by changes in technology such as improved modulation techniques, smaller cell sizes, use of cognitive radio techniques, higher frequency capabilities, etc. Balancing supply and demand - In a perfect market, supply and demand are always in equilibrium in the long term. However, since spectrum is a fixed resource this may push spectrum prices up to the level where users are denied access to new services, and does not imply that there is not shortage. The price that network operators are prepared to pay for spectrum also depends on the costs they incur elsewhere in the network; pricing that needs to be taken into account therefore includes: price of spectrum, price of new technologies (small cells are more expensive per user than large cells, particularly where new backhaul is needed), and the price of services being offered. Timing factors It is possible that short-term market distortions may occur due to lags in investment cycles, effects of short-term economic fluctuations, and the ability of manufacturers to introduce new technologies etc. Additionally, standardization activity necessarily takes a considerable amount of time. All these factors introduce time delays which create instability in supply / demand equalization. 1-2

14 1. Introduction There are additional factors that also need to taken into account in the model, such as: Current and forthcoming changes in services - For example, TV broadcasting is partly moving towards high-definition TV, increasing the sector s need for spectrum. Changes in demand for services, and hence their demand for spectrum - Cellular operators serve as an example as the demand for mobile data accelerates, their appetite for spectrum (and their preferences for particular spectrum bands) will alter. Current and forthcoming technology advances - For example: Digital TV provides broadcasters with greater spectral efficiency, and hence they can deliver more channels in the same amount of spectrum. The latest generation of radar equipment should provide considerable advances in spectral efficiency when implemented in 5-10 years time. These changes will affect the requirement for spectrum. Technology neutrality This affects the supply of spectrum, and hence the availability of suitable bands to meet requirements. However, this may not only ameliorate the problem by making more spectrum available, but could make things worse if it increases uncertainty about future usage of particular bands. One example is the potential use of WiMAX in cellular bands, which, if WiMAX services are not successful, could reduce the amount of spectrum available to incumbent operators whilst not satisfying the demand for advanced wireless services. Other extrinsic changes - Changes to WRC and Europe-wide harmonization plans have a significant effect on demand for specific bands because the unit cost of equipment for a band falls dramatically if its use is harmonized across a large international community of operators and users. Additional benefits may also accrue to operators and users e.g. in the case of cellular mobile, international roaming delivers benefits to users, operators, and also to other parties. Cost of network rollout The choice of technology will affect network cost. For example, more recent cellular technologies will typically use IP networks which have a lower cost of ownership than traditional ones. Evolution in the behaviour of operators when faced with spectrum scarcity e.g. mobile operators are now starting to share some of their access network infrastructure this type of response is a natural response to shortage of resource. The model incorporates all of these factors in a manner that allows Ofcom to determine whether the supply of spectrum will fall short of demand in terms of: Projections of future spectrum demand per service Projections of future excess supply / demand for spectrum, by band / by service An indication of bands where alternative capacity might be available. In addition the model also enables Ofcom to investigate the reasons behind the results: Allowing sensitivity analysis so that the effect of changing each input can be explored and the consequences can be understood Auditability - the user can follow the calculation logic both forwards and backwards through the model, and can explore the formulae used and the sources of the data Metadata to support each input or assumption i.e. identifying the source and (where appropriate) the derivation of the input value. 1-3

15 1. Introduction 1.3 EFFECTS OF PRESSURE ON SPECTRUM Throughout this report we consider demand for spectrum as an absolute, that depends on demand growth of new applications and on the extent to which new technologies deliver greater efficiency and hence capacity within any frequency band. Our results are therefore informed by historical growth, with the various scenarios we consider forming future variants. Effects of shortage of spectrum However, in practice demand for cellular spectrum is not independent of supply; if operators do not have sufficient spectrum to meet potential demand and cannot solve the problem cost-effectively, they will tend to delay roll-out of high bandwidth applications. Similarly other competing operators will probably be in the same situation so there will be no competitive disadvantage. Customers may not notice since they will still see the Quality of Service that they are accustomed to, and will not have come to expect more. So there exists a complex balancing mechanism not only around the price that network operators are prepared to pay for spectrum, but around the applications that they promote in the marketplace. Additionally, because of the extent to which alternative mechanisms may be applied as an alternative to using more spectrum (e.g. installing additional cell sites), and the uncertainty associated with many of the parameters of the forecasts, it is not possible to state that there will be a shortage of spectrum. In this report we therefore use the terms pressure on spectrum and potential shortage to indicate scenarios where we predict a risk that demand may exceed supply. Effects of surplus of spectrum Similarly, if the operators are able to access more spectrum than they might need for a high-density cellular rollout, they will tend to relax the number of sites that they might otherwise install (subject to achieving geographic coverage). This will reduce their costs and enable them to offer lower prices to customers. However, it does not mean that spectrum will be unused, or even that it will be used inefficiently. 1-4

16 1. Introduction 1.4 ORGANISATION OF THIS REPORT This report includes the following sections: Section 1 Introduction This section. Section 2 Scope of Study Discussion of the services, applications and frequency ranges considered. Section 3 Our Approach Outline of the methodology used. Section 4 Scenarios Details of the demand / supply scenarios modeled. Section 5 Assumptions and Forecasting Methodology Discussion of the growth models used, the services offered by each type of network, and the starting point for the forecasts. Section 6 Baseline Scenario Business As Usual Detailed results for the baseline scenario, showing growth of all services and spectrum supply / usage over the period to Section 7 Demand Growth for Services This is the main results section, showing our traffic demand, spectrum demand, spectrum supply, and pressure on spectrum forecasts for different scenarios. It is organized into sub-sections each covering one service, e.g. cellular, short range wireless, broadcast etc. Section 8 Demand for spectrum all bands and services Having presented all the results by service, in this section we bring together the total picture of pressure on spectrum, for all services in all bands. Section 9 Population density issues In this section we consider the implications of ultra-dense urban and the implications of the very high population densities on spectrum requirements. Section 10 Conclusions, Recommendations and Next Steps In this final section we discuss the results and their implications. Appendices Additional detail is included in Appendices, including: Assumptions for network architecture and spectrum use Scenarios used Model specification Demand Forecasts sources Additional detailed results shown by service and scenario Stakeholder Interview notes. 1-5

17 2. SCOPE OF THE STUDY 2.1 THE TYPES OF NETWORKS AND SERVICES COVERED BY THE STUDY The scope of the study covers five broad types of network, agreed with Ofcom as the main services of interest: Cellular (mobile device to/from fixed transmitter/receiver) Short Range Wireless (device to another nearby device) Broadcast TV (fixed transmitter to multiple devices, one way, video and audio) Broadcast Radio (fixed transmitter to multiple devices, one way, audio only) Fixed Wireless Access (nomadic or fixed device to/from a fixed transmitter/receiver). In addition, spectrum use for Backhaul technologies was studied. Since the results showed that most Backhaul spectrum usage was above the range of interest, the results have not been included in detail in this report, although they are shown in some of the charts of spectrum usage. The current services are as follows: Cellular. Each of five operators operates a fully mobile network with of the order of 20,000 base stations. Each of these networks offers a mix of different cellular technologies currently covering 2G, 2.5G, 3G, and likely to extend to others. The networks provide a mix of applications, currently dominated by voice traffic and (in value terms) messaging, with some data applications such as and web browsing. The mix of applications is expected to shift considerably, away from voice towards data, possibly with some take-up of quasi-broadcasting applications / technologies such as real-time video to the handset. Traffic is carried from the fixed base station to the end user s device and vice versa. Requirements for spectrum are affected by a range of factors including the amount of traffic offered, the peakiness of demand, the required quality of service to be offered, the distribution of users across the country, the availability of useful sites and the ease with which the same frequency can be reused across many different sites within the network. Short Range Wireless. This is in truth a population of devices and users, rather than a large-scale network or set of networks: DECT and Analogue cordless technologies provide short range voice communications 2 between one fixed device and (usually) one nomadic device. WiFi is a wireless LAN technology focused on providing data links between one fixed point (base station) to one or several nomadic devices. Bluetooth and UWB are wireless Personal Area Network (PAN) technologies to connect fixed or nomadic devices within close range of each other. Requirements for spectrum are sensitive to a similar set of factors as for cellular. 2 DECT is also able to support other applications, but there is little usage of this capability. 2-1

18 2. Scope of the Study Radio and TV Broadcasting. National networks typically of some 1,000 transmitters 3 (of which 50 to 100 provide the bulk of the coverage). All of the traffic is on the downlink, ie from broadcaster to consumer. Each content stream may be broadcast over one or several broadcast technologies, eg a radio station may appear on Analogue FM, DAB, and DVB-T. In addition to the broadcast content itself, there is some use of white space within the existing spectrum allocations for Programme Making and Special events (PMSE), Fixed Wireless Access. The networks comprise point to point links between one fixed location and another - generally as an alternative to a cabled connection. Requirements for spectrum are affected by the amount of traffic offered and the degree to which the same frequencies can be reused across the country. Backhaul. These networks comprise a set of directional microwave links between fixed points. In the range of spectrum in scope for this study 100Mz to 10GHz the main use for backhaul is the distribution of broadcast content to transmitter sites. The majority of backhaul spectrum is used for cellular networks, but this is generally at frequencies of 30GHz and above. For this reason it is not considered further in this report. 2.2 SERVICES, APPLICATIONS AND TECHNOLOGIES Within each of these services, the other two important concepts are Technologies and Applications, defined as follows: Technology a technology is a set of hardware, software, RF, protocols etc that together define how an air interface functions within one of the six services. In this project some 20 Technologies are covered, including: 2G, 2.5G, 3G, Mobile WiMAX, WiFi, UWB, DAB, DVB-T, DVB-H, and microwave backhaul. Application a body of functionality that delivers value to users, e.g. voice, messaging, , gaming, streaming video, P2P, M2M, TV, Radio. Not every combination of technologies and applications is viable. In Table 1 below we show the relationship between services, technologies, and applications, and which combinations work and which ones do not. 3 With the exception of DVB-S, for which the content is broadcast to the whole of the UK from a single satellite. 2-2

19 2. Scope of the Study Table 1: Service, application, and technology matrix 2-3

20 2. Scope of the Study 2.3 FREQUENCY BANDS We have focussed on the spectrum in the range 100MHz to 10GHz, with particular focus on that needed by the Cellular and Broadcast services, on the grounds that these services are regarded as delivering the largest share of the value and hence place the greatest value on having sufficient spectrum available 4. In terms of frequency ranges, this places the greatest emphasis on the range MHz to about 4GHz. When analysing the demand for spectrum, this range of frequencies is too large to allow meaningful comparisons and inferences, and accordingly it is split into several sub-bands when presenting the results. These, along with the main services that use each of them, are as follows: MHz 5 : Broadcast Radio 300-1,000 MHz: Broadcast TV, Cellular 1,000-4,000 MHz: Cellular, Short Range Wireless 4,000-10,000 MHz: Backhaul, Satellite Broadcasting 2.4 SPECTRUM USE TODAY Table 2 and Table 3 show the current allocations of spectrum by Technology, Service and Band. 4 5 Ofcom report c value of wireless services Strictly, this is extended to cover the whole of the FM radio band, down to 88MHz. 2-4

21 2. Scope of the Study Service Technology Spectrum Band Cellular Short Range Wireless 2G/2.5G 900MHz GSM/GPRS/EDGE 2G/2.5G 1800MHz GSM/GPRS/EDGE Comments re which band Comments re 2008 allocation Spectrum allocation 100Mhz to 10GHz MHz Shared with 2.5G 2 Operators x 17.2MHz paired each 70MHz 1,800MHz Shared with 2G 2 Operators x 30MHz paired each, 2 Operators x 5.8MHz paired each 3G UMTS 2,100MHz Shared with 3.5G 5 Operators, 10-15MHz paired each, ignore TDD 3.5G (HSPA) 2,100MHz Possibly 2,600MHz as well Included in 3G --- 4G LTE 2,600MHz Possibly 1800MHz, 2600MHz, 3000+MHz Mobile WiMAX 2,600MHz Can use other bands e.g. 2100MHz MBMS 2,100MHz Assume this eventually takes the TDD spectrum at 2,100MHz 3G/3.5G/4G Expansion 140MHz 120MHz None in None in MHz TDD at 2100MHz (used eventually) 900MHz And/or 700MHz None in Analogue Cordless 30MHz All of the allocation is below 100MHz MHz DECT 1,880MHz 20MHz 20MHz Bluetooth 2,400MHz Same spectrum as WiFi 2,400MHz --- UWB 5,200MHz 3,400MHz to 10,600MHz Assume MHz as per UK FAT ~500MHz 6 (effective/shared) WiFi 2,400MHz 2400MHz to 2483MHz 83MHz 5,800MHz ~500MHz Table 2: Allocations of spectrum in 2008: (i) Mobile Technologies 6 Note that UWB is a very high bandwidth system that transmits below the noise floor of other systems operating in the same band. It operates over very wide bandwidths, but with an effective bandwidth of around 500MHz. In quantifying the supply of spectrum, we have taken the somewhat conservative approach of not double counting this with the 500MHz of Short-Range Wireless spectrum allocated to Wifi at 5800MHz. 2-5

22 2. Scope of the Study Service Technology Spectrum Band Comments re which band Comments re 2008 allocation Spectrum allocation 100Mhz to 10GHz 2008 Broadcast Radio Broadcast Television Fixed Wireless Access Analogue (FM) 100MHz 88MHz to 108MHz 88 to 108MHz 20MHz DAB 200MHz 174MHz to 230MHz 20MHz allocated for radio: 12 x 1.7MHz multiplexes, 7 of which are currently used 20MHz DRM 100MHz FM (and AM) band Nil --- Analogue TV 700MHz 470MHz to 860MHz 4.5 channels x 8MHz x reuse of 6 216MHz DVB-T 700MHz 470MHz to 860MHz 6 Muxes x 8MHz x reuse of 3-5 depending on coverage assumption (5 for % pop coverage) DVB-H 700MHz 470MHz to 860MHz Nil, other than some testing --- DVB-T2 700MHz 470MHz to 860MHz Nil --- DVB-S 11,000MHz Satcoms bands 11-12GHz All available spectrum (potentially 500+MHz) is above 10 GHz Fixed WiMAX TC/CDMA Proprietary As Proprietary below As Proprietary below 3,600MHz, 5,800MHz As Proprietary below Some or all may conceivably shift to mobile, or to other FWA technologies 168MHz MHz As Proprietary below Nil --- Or others e.g. 2100MHz / 2600MHz / 3400MHz Nil --- Table 3: Allocations of spectrum in 2008: (ii) Broadcast and Fixed Technologies 2-6

23 3. OUR APPROACH 3.1 THE APPROACH TAKEN TO QUANTIFYING PREDICTIONS OF SPECTRUM DEMAND AND SPECTRUM SHORTAGE In this section we provide an overview of how the raw data on demand, technology changes, and spectrum availability and usage has been taken and used to generate forecasts of spectrum shortages and surpluses over the next years. Figure 1 below shows the overall methodology that we have used during this study. Service Demand forecasting Developing projections of traffic demand, separately for each service. Converting these projections of demand for Services into a common language and quantified in an appropriate set of units. Spectrum Demand forecasting Defining the technical parameters governing how each type of network carries traffic, and establishing how these vary between Scenarios. Converting the Demand for Services into Demand for Spectrum. Quantifying Spectrum Shortages / Surpluses Establishing the quantity and type of spectrum available, and hence determining the net surpluses and shortages. Figure 1: Modelling demand for services and the net surpluses / shortages of spectrum 3-1

24 3. Our Approach 3.2 FORECASTING THE DEMAND FOR SERVICES Develop Baseline Demand Forecasts ( Business As Usual ) Data and forecasts used are from various published sources and benchmarking rather than primary research. Where estimates differed, these were reconciled to provide what we would regard as a plausible consensus view of the likely future trajectory for Business As Usual. In many cases, published projections of demand only covered five or ten years, so projections for the remaining period were derived by consensus within the team, keeping the approach and assumptions consistent with those of the original forecasts. For each of the services we established a baseline ( Business As Usual ) forecast of demand, which was then flexed for each of the scenarios detailed in Section 1. The resultant forecasts comprise yearly estimates for: The initial population (including number of subscribers) using a particular service and application across geographical areas and bands, as well as a historic growth rate Initial and future take-up of each service and application (indicated by start date). Forecasting service demand requires projections of numbers of users and traffic per user for each of the services and applications between 2007 and The expected traffic is determined in appropriate units (e.g. MBytes, voice minutes, channels, terminals). Data at this stage provides conversion factors to combine different units of demand Develop Demand Forecasts for Remaining Scenarios For the remaining Scenarios, the principal differences in the projections of demand for each scenario were as follows: Different rates of growth in demand for each Application. As indicated in the Scenarios paper 7, the general approach has been to multiply growth rates by approximately: x 1.50 for Wire Free World x 2.00 for All You Could Want x 0.75 for Dystopia. The main exception to this is for Wifi and similar short range wireless technologies, where a slightly narrower range of growth factors have been used: x 1.25 for Wire Free World x 1.50 for All You Could Want x 0.75 for Dystopia. These differences between the growth rates in the six scenarios have the effect of varying the rate of progress and change in each scenario, rather than simply scaling the level of demand up or down by a constant factor, or adding an extra 10% or 20% to the annual rate of growth across all applications. 7 The Scenarios paper is included as an Appendix to this report. 3-2

25 3. Our Approach Different diffusion rates for the migration from the installed base of cellular technologies to new ones, which in turn affects the rate at which traffic can shift from older cellular technologies such as 2G to newer ones such as WiMAX and LTE. Different attractiveness for newer, more spectrally efficient, cellular technologies, which has a further effect on the rate at which traffic migrates to the newer technologies. Differences in the evolution of the mix of demand across Applications and Services in each Scenario e.g. Dystopia has a shift away from consumer-oriented application and services Convert Demand Forecasts into Common Units This conversion process took two forms; one for Cellular, and one for the other Services. Cellular For Cellular, most demand forecasts were expressed at the Application level, ie technology-independent. This reflects the end user viewpoint - there is a demand to be able to make a call, or send a message, with certain quality expectations (speed, sound quality etc.) and the user is indifferent to whether it is delivered via 2G, 3G, or 4G. This demand was then allocated to the individual cellular technologies, taking account of: Installed device base: traffic cannot be carried on 3G if the user s device is 2G-only The attractiveness and quality of each cellular Technology in delivering each Application to end users. Traffic was then converted into common units (MB / hour) from the different units used when formulating the projections of demand for each Application; voice-minutes, SMSmessages, in MB, and so forth. Note that it is necessary to allocate the traffic for applications to specific technologies, since some are more efficient at carrying applications than others (e.g. different coding rates, different frequency reuse). Non-cellular For the other (non-cellular) Services, raw estimates of demand were generally defined at the level of each individual technologies, so the only process required was to convert demand into the common units of MB / hour. 3-3

26 3. Our Approach 3.3 FORECASTING THE RELATIONSHIP BETWEEN DEMAND FOR SERVICES AND DEMAND FOR SPECTRUM Change in demand for Services does not necessarily translate into changes in the demand for spectrum. There are a number of steps needed to convert the Offered Traffic into demand for spectrum, and then into a net surplus or shortage of spectrum: Converting the offered traffic for each application and service into traffic volumes calculated on a common basis for all of the technologies (same units) Determining the required capacity that the networks need to be able to provide, to provide an acceptable service Spreading the demand across different neighbourhood types, and, for some technologies, down to sites Understanding the likely spectral efficiency, converting the required capacity into a demand for spectrum (MHz) Determining how these assumptions change over time in the period to Derive the Network Technical Parameters (e.g. Capacity, Geography) for Business As Usual To convert a demand estimate for services into a demand for spectrum requires a wide range of inputs to be specified for each type of network. These include: Differences between the peak time traffic, and the required capacity that would be specified for the Service / Technology / Application Spectral efficiency and frequency reuse Availability of useful sites, and the timing of major shifts in technology e.g. ramp-down of 3G and the emergence of other technologies such as 4G and femtocells. Differences between neighbourhood types, in particular differences between rural and urban neighbourhoods Changes to each of these factors over the period to These factors were derived using a mix of external research, expertise within the consortium, and calibration of results against currently observable network performance Derive the Network Technical parameters for remaining scenarios A similar set of parameters were derived relative to those used for Business As Usual, taking account of the different evolutionary path projected for each Scenario. These differences are a mix of scale (changes are larger or smaller), timing (faster or slower), and other types of divergence, according to what is appropriate to each of the network technical inputs. 3-4

27 3. Our Approach Derive the Demand for Spectrum The calculations can be split into four main steps: Convert overall demand (in MB/ hour) into required busy-hour capacity Allocate this requirement to neighbourhood types and then derive traffic per site (with simplifications for some non-cellular Services) Convert this from required capacity for traffic into required spectrum, taking account of spectral efficiency Aggregate these requirements for spectrum to get a national picture of the demand for spectrum, by Service and Band, over time. 3-5

28 3. Our Approach 3.4 PRESSURE ON SPECTRUM The focus of this report is on identifying potential spectrum shortages for specific applications / services. For example, in cellular, the pressure (potential net surplus or shortage) on spectrum is reported and analysed. A potential shortage of spectrum means that, for example, cellular may need more spectrum in future than it is scheduled to have allocated to it. In some areas, this net position may mask some imbalances between different technologies - for example, a new technology such as LTE or WiMAX may have a large surplus of spectrum, whilst a mature technology such as GSM may have a shortage. We cover the major examples of these shortages for the cellular market where there are important interactions of technologies and the gradual evolution of the installed base of devices and hence the routing of traffic between technologies and bands Define the Spectrum Available in Business As Usual Current spectrum allocations were taken as the starting point for 2007/08. The known planned and expected future changes were then factored into these allocations over the next 2-6 years, to reflect, for example: Forthcoming allocations at 2.6GHz Digital Switchover / Digital Dividend Review Refarming of 2G spectrum. Looking further ahead, it seems likely that there will be some releases of spectrum from UK public sector organisations such as the MOD, who are now facing financial pressures and incentives to release, or at least share more extensively, potentially valuable spectrum. There are also other potential changes within the timescale of the study, including the expiry of the existing 3G licences. These changes were also factored into the projections of available spectrum out as far as 2025, taking some account of the likely uses for any released tranches of spectrum Define the Spectrum Available in Remaining Scenarios All Scenarios start from the same spectrum available in , but thereafter there is some divergence between the Scenarios, reflecting, amongst other factors: Different levels of demand growth; if growth is faster then there will be greater pressure, over the medium term, to attempt to alleviate this by releasing more spectrum from other public sector / non-commercial users Some slight differences in the speed and efficiency with which spectrum is reused, shared, or released, for example in Dystopia there may be more pressure to retain spectrum for applications geared to protecting national security Determine the Net Spectrum Pressure Net shortages are calculated simply by comparing the demand for Spectrum against the final figures for what is available, then taking different views of this data eg by Service, spectrum band, year, and neighbourhood type. 3-6

29 4. ALTERNATIVE SCENARIOS THAT MAY UNFOLD IN THE NEXT YEARS In order to examine spectrum usage over the next years we have adopted a scenario-based approach. In developing the alternative scenarios we have drawn on discussions with a range of parties including the study s Stakeholder Panel and experts across the organisations within the consortium. The scenarios for analysis have been deliberately selected to cover a wide range of outcomes, in order to illustrate those conditions under which more or less shortage of spectrum may occur. 4.1 DRIVERS OF CHANGE Over the last years the drivers of demand for spectrum have been changing, and it seems likely that this will continue. This section outlines the principal drivers of change and how they are likely to filter through into changes in the demand for spectrum and the likelihood of any pinch points or shortages emerging. In addition to the applications, technologies, and services considered in this study (see Section 0 above), it is recognised that there will be a range of softer factors which may have a bearing on the demand for spectrum and which may, to varying degrees, drive demand growth in the latter part of the year period under consideration. Also, there are potential new applications, services and technologies that could significantly affect this picture, depending on their development. In order to understand what these other factors and emergent applications, services and technologies might be, we held a series of meetings with key stakeholders; the following sub sections are summaries of the key ideas postulated, split into: Emergent applications, that end users may want Enabling technologies, that support these advances in applications Possible Emergent Applications Mobile virtual world: Fully mobile high resolution video, e.g. in visor form, with 3D imaging, shifting the boundary between perception and reality. Short range devices: Personal wearable devices able to interact in real time with other devices when in proximity. Devices may steer the wearer towards specific services or products (given sufficient trust / regulation). Mobile broadcasting: Greater interactivity and personalisation of the media further blurring the distinction between broadcasting and communications. Digital media: The emergence of digital paper as a result of convergence between TV, radio, paper, magazines and interactive entertainment such as computer games. 4-1

30 4. Alternative Scenarios That May Unfold in the Next Years Digital jewellery: The concept of the mobile phone as a single platform device may be replaced with digital jewellery, containing much greater functionality than a current mobile phone but in interoperable individual component devices - made possible because of the increased miniaturisation of chip sets and batteries through nanotechnology and new materials subject to overcoming the challenges including the evolution of the user interface. Active skin: micro-electronics could be in use to monitor a person's medical state, using chip(s) the size of a grain or smaller, embedded under the skin. The devices would be able to communicate in real time allowing, for example, real time remote monitoring of medical conditions such as diabetes and the administration of required drugs in real time with no user intervention. Real-time transport applications: Increasing sophistication in sensing, communication, and processing, to provide greater assistance and coordination to road users Possible Enabling Drivers Improvements in semiconductor and nanotechnology making it possible to develop smaller and smaller chip sets with increased functionality e.g. processing power and storage capacity will increase, with user interfaces such as speech recognition becoming commonplace. Mobile and fixed communications networks converging to create a fully Wire Free World with extremely high data rates, which will be available to everyone at any location. Wireless networks providing more and faster short range links to fixed networks. Mobile technology migrating to a single network protocol, namely IP, and this will help promote cooperation and service development as well as constructive competition between operators. The Scenarios Paper, attached as Appendix E of this report, gives further details of the emergent applications and enabling drivers that were identified and considered. 4-2

31 4. Alternative Scenarios That May Unfold in the Next Years 4.2 DEFINING THE FUTURE SCENARIOS From analysis of the key future drivers for spectrum and their associated uses over the next years we have derived a set of six future scenarios: Business As Usual - The growth trend and technology developments continue the trends of the last years. There is some shift from wired to mobile wire-free, but this is a gradual trend rather than a sudden change. The detailed assumptions for Business As Usual are included in Section 3 of this report. Wire-Free World - This scenario is centred on faster growth in wireless technology and demand for services. Wide applicability and seamless convergence of fixed and mobile communication networks create a fully Wire-Free world with extremely high data rates, available at practically all locations. Users will have wireless access anytime anywhere, breaking down the some of the traditional barriers between being at home, at work, and elsewhere. All You Could Want - There is greatly accelerated growth and a significant shift from wired to wire-free. Services, demand, applications and technologies develop along the lines that they have been developing for the last years but with the pace accelerating and an upsurge in new applications using device to-device and deviceto/from-person communication. Dystopia - Society is more threatened and fearful, due to a bleaker social and economic outlook, driven by e.g. accelerating prices for raw materials, an upsurge in terrorist threats and activities, identity theft, and severe reining-in of credit and expenditure, throttling back economic growth. The mix of technology and demand shifts away from being diverse and consumer-centric, towards a focus on systems aimed at monitoring, recording and coordinating everything, aimed at reducing terrorism / crime / disorder / antisocial activity. Total demand continues to grow, but slower than in Business As Usual. Industry Fragmentation The growth in capacity does not keep up with the demand, as the degree of cooperation and standardisation across geographies and different parts of the supply chain deteriorates, reversing some of the gains in economies of scale achieved in the last years. Industry starts to fragment, from having major players and interoperability, to a profusion of smaller solutions and ventures fighting for spectrum, content, services, and revenue. National governments are more inclined to step in with special measures to protect the interests of local ventures. Other scenarios broadly cover the opposite case, where we assume that there are strong efficiencies of scale. Re-use There will be a significant shift in emphasis toward reuse and sharing possibly reflecting a similar shift in values in the wider society and economy. This will be reflected in (i) greater cooperation between networks, for example tolerance or encouragement for sharing of sites and/or spectrum; and (ii) greater releases of spectrum suitable for re-use in mobile and fixed networks possibly driven by introduction of AIP to public sector usage both in the UK and in around the world. 4-3

32 4. Alternative Scenarios That May Unfold in the Next Years 4.3 IMPLICATIONS OF SCENARIOS ON SUPPLY AND DEMAND OF CAPACITY In Figure 2 below we compare the six different future scenarios in term of their supply of capacity, and the latent demand for that capacity. Differences in the supply of capacity reflect both differences in spectrum availability and in techniques and architectures that increase efficiency (e.g. improved compression techniques, greater frequency reuse) Supply of Capacity Re-Use All You Could Want Wire-free Business as Usual Demand for Capacity Dystopia Industry Fragmentation Figure 2: Positioning of the different future scenarios 4-4

33 4. Alternative Scenarios That May Unfold in the Next Years 4.4 A COMPARISON OF THE SCENARIOS AND THEIR IMPLICATIONS The differences between the scenarios outlined above enable us to identify the likely main differences between them over the next years differences in demand growth, in networks and technologies, and differences in the release and reuse of spectrum. These are summarised below in Table 4 below. Business As Usual scenario Wire-Free World scenario All You Could Want scenario Dystopian scenario Industry Fragmentation scenario Re-Use scenario a. Demand (traffic) Total Demand growth per annum (amount in MB equivalent) Total Demand growth (mix) Demand growth (driven by growth in technology) b. Network architecture Moderate High Very High Slow Moderate Moderate Shift To Data Shift To Data Large Shift To Data Large Shift To Data Shift To Data Shift To Data Some Some High Some None None Cell proliferation Slow Moderate Very High Moderate High Moderate Fibre plus wireless Slow High Very High Slow Slow High Industry fragmentation Some Moderate Moderate Some Very High None Convergence and integration of services Cellular traffic carried by femtocells c. Availability of Spectrum Rate of release/re-use of spectrum Spectrum coordination across regions Slow Moderate Moderate None Negative Moderate 20% 40% 60% Minimal Minimal 20% Moderate High High Slow Moderate Very High Moderate High High Slow Negative High Table 4: A comparison of the set of future scenarios Each of the dimensions has a different out-turn in each scenarios, and their relationships are summarised in Table 4 in terms of Very High to Negative. Further details are provided in the following sections, each describing one Scenario in turn. It is also possible to assume combinations of scenarios, taking growth rates from different scenarios for different applications. In the results that follow the impact of these can be understood by taking the services independently for the scenarios of interest and summing the results. 4-5

34 5. ASSUMPTIONS FOR FORECASTING THE DEMAND FOR SERVICES In this section we present the forecasting methodology used for each service/application pair to determine the following data: Forecasts of number of subscriptions for the period Forecasts of total offered traffic for the period In order to derive the future traffic per service per application, the demand for each application is modelled using existing market surveys and reports. The key data used in the model are the number of subscriptions and total offered traffic (inbound and outbound) from for each service and application. There are two key issues to note in forecasts of demand for individual applications: First, the forecasting horizon in this study However in previous published studies on which we draw, the forecasts vary within the period of Where the forecasts fall short of 2025, we model a trajectory based on the last growth rate projected Second, new/young applications in order to project future traffic volume for either new or relatively new applications within the forecasting horizon, the number of subscriptions is required. An important part of the demand projections involves the penetration of each application. The diffusion models are used to forecast the take-up rate of an application. It is important to note that the full penetration may occur during the forecasting period for some applications and as a result a constant traffic in the later periods can be forecast. Sources used in the development of demand forecasts are listed in Appendix B. 5-1

35 5. Assumptions for Forecasting the Demand for Services 5.1 CELLULAR Voice and Messaging Subscriptions are assumed to equal the total of cellular users. This in turn grows with population, limited to ~1.5 subs per person. Voice and messaging traffic is modelled using log linear demand models. Reference data for 1999 to 2006 is taken from Oftel 8 and Ofcom 9 and income is taken from Office of National Statistics 10 for the same period. Data Subscriptions are assumed to follow overall 3G take-up, which in turn is assumed to reach around 80% of cellular subscribers by Data Traffic forecasts are based on three inputs: Number of subscribers: the assumption is made that these follow the diffusion of 3G. Number of sessions per sub: This is taken from the UMTS Forum 2003 and 2005 data estimates for 2012 and It is assumed that sessions per subscription vary with growth rates over the period. Uplink and downlink file size per session: An assigned uplink and downlink file size is assumed to be constant throughout the forecasting period. Application Sessions per month / user Data per session Uplink (KB) Data per session Downlink (KB) Year E Mail WWW Video Streaming Gaming Download P2P Telemetry & M2M Table 5: Input parameters for cellular services (Source: UMTS Forum) Note that the average uplink data for P2P goes down over time as the mix of different session types changes. Mobile IPTV is assumed to start in Take-up is estimated as proportional to mobile broadband (Forrester, 2008). We assume growth 11 from 58,000 users in 2010 to 965,000 by In parallel, an increasing proportion of users use MBMS, growing to 42.5% in Traffic is estimated based on forecast monthly sessions and average file size (WiMAX Forum, 2007) for The Market Information Mobile Update, published by Oftel, various years, The Communications Market: Mobile telecoms market data tables, Ofcom, Real households' disposable income per head, Economic Trends, Quarterly Values, published by ONS The percentage of IPTV users is based on study [40]. 5-2

36 5. Assumptions for Forecasting the Demand for Services Baseline Traffic and Subscriptions Demand Forecasts This section summarises the Business As Usual demand inputs used for cellular. Figure 3: Cellular - Total Subscriptions by Application The number of subscribers by application is shown in Figure 3 above. But what is of rather greater interest is the growth in Offered Traffic, and that is summarised in the next two charts. Figure 4: Cellular - Total Offered Traffic: Voice Calls (Million Minutes / Hour) The total voice traffic offered shown in Figure 5 above is assumed to rise by a factor of about two over the period, as we see continued increased use of mobile phones and increased fixed/mobile substitution. 5-3

37 5. Assumptions for Forecasting the Demand for Services Figure 5: Cellular - Total Offered Traffic: Traffic by Application (Terabytes / hour) The growth in Offered Traffic for data applications is shown in Figure 5, measured in Terabytes per hour (average hour, not busy hour). 12 The growth in data traffic is expected to be considerably greater than that for voice. 12 For reference, 1TB = 1,000GB. 100 TB/hour is approximately 28Gbps, corresponding to 10 million users running at 2.8kbps continuously. 5-4

38 5. Assumptions for Forecasting the Demand for Services 5.2 SHORT RANGE WIRELESS DECT and Analogue Cordless The primary use of DECT in the UK is assumed to be voice over fixed lines. It is assumed that 30% of fixed lines calls are made over DECT. The traffic trend is decreasing, according to Ofcom figures between 2005 and 2007 at a rate of about -15% per annum. Analogue cordless provides a similar voice capability, but is assumed to be small 13 and remaining below the 100MHz-10GHz spectrum of interest Wi-Fi Three user types are considered: home, work and mobile. Home and Work cover fixed broadband users using Wi-Fi technology; Mobile covers hotspot users. The total number of work and home subs is based on diffusion of fixed broadband and extrapolated to 2025; the proportion of fixed broadband users with Wi-Fi is 95% in For mobile users, the number of hotspot users is calculated using data from a survey by Ofcom (2003). We assume that this is the sum of both pay and non-pay public wireless or Wi-Fi network users. Subscriptions to individual applications are also considered. VoIP It is assumed that 18% of mobile SRW Wi-Fi users will use VoIP (Voice over Internet Protocol). Minutes of use (MoU) per session are estimated to be 15 minutes and the average number of sessions per month is 30. It is assumed that 88% of mobile SRW Wi-Fi subscribers use . Using the number of sessions per month and the maximum file size for the years 2010, 2015 and 2020, traffic per session is calculated. Data applications Annual traffic for other data applications is calculated as for . Mobile IPTV The number of IPTV subscribers was around 25,000 in which is 0.3% of the total fixed broadband users. Using this, it is assumed that only 20% of total fixed broadband users will have IPTV by Hence, the numbers of IPTV users are calculated using the diffusion of SRW Wi-Fi users. The number of sessions for 2010, 2015 and 2020, and the downlink file size is fixed throughout the period. Radio It is assumed that 13% of Wi-Fi subscribers use radio (live streaming audio). The number of sessions per month per user and the number of minutes per session are 30 and 164 respectively. Using these inputs, the annual traffic in minutes is calculated Bluetooth Bluetooth is available on a number of devices, with the majority of usage being driven by either mobile phone or laptop users. Communication between devices of hands-free headsets and mobile phones and PDAs and laptops is common. To assess traffic, the average use of phone and laptop users per day is considered for voice and data. 13 Analogue cordless indicated to have 25% in 2003, 9% in 2005 (Ofcom (2006) Complaint against BT s pricing of digital cordless phones) 14 Source: Ofcom 5-5

39 5. Assumptions for Forecasting the Demand for Services Ultra Wide Band (UWB) Forecasting demand for UWB requires device take-up and usage, at home and work, daytime and evening. Ofcom (2004) defines 22 uses and forecasts adoption rates in 2010 and Using these we determine the penetration of applications for and which of these are used at home or/and work. Using proportions given by Ofcom (2004), usage areas are determined. Further, to determine the traffic in minutes, each application is assigned percentages of total daytime and evening that they are actively in use Baseline Traffic and Subscriptions Demand Forecasts This section summarises the Business As Usual key demand inputs used for SRW. Figure 6: Short Range Wireless - Total Subscribers by Application 15 Note that current demand for UWB is much lower than even the low take-up case forecast in: At such an early stage of the market it is difficult to be certain whether the longer term forecasts should be downgraded, but it is clear that at present UWB is not being introduced into products in the way that was expected, although chipsets are becoming available. 5-6

40 5. Assumptions for Forecasting the Demand for Services The chart below shows the assumed growth in traffic for data applications. Unlike the previous chart, DECT is not shown (partly because it is measured in minutes rather than MB). DECT traffic is rapidly overtaken by the growth in data traffic. Figure 7: Short Range Wireless - Total Offered Data Traffic (Terabytes / Hour) Voice calls over Short Range Wireless are expected to remain broadly constant or decline, with a greater share of minutes per annum being taken by WiFi (VoIP). 5-7

41 5. Assumptions for Forecasting the Demand for Services 5.3 BROADCAST TV AND RADIO Whereas the demand for spectrum in other services (e.g. cellular and short-range wireless) is driven by the demand of users, the demand for spectrum for broadcast services is driven by the number of channels/stations/streams. The main input is therefore the number of Channels, together with the size of the content stream in each case. Quantifying this demand is somewhat problematic. In theory, broadcasters would always have some appetite for an additional terrestrial channel. In practice, at the margin this demand is unlikely to be strong enough to displace the demand from cellular services, and it seems likely that in a service-neutral auction, the broadcasters would not retain the spectrum if competing with the cellular operators. Furthermore, the broadcasters have the option - albeit at a price in terms of share of voice, cost of delivery and cost for consumers to receive - of shifting the content from terrestrial to satellite technologies and frequencies. In the light of this, we have assumed a degree of pragmatism and applied a degree of responsiveness of spectrum demand to spectrum cost; we have capped the projected demand for terrestrial channels at roughly the amount that the technology of the time will enable broadcasters to fit within the amount of terrestrial spectrum available Broadcast TV For TV, traffic forecasts are provided by technology. For this study, the important measure of demand is the number of content streams ie the number of TV and radio channels being transmitted. Regional, local, and sub-national channels are converted to the equivalent number of national channels. Projections were made for the number of subscribers and the consumption per head or household, allowing sense checks on the projected audience per channel, but these latter measures of demand do not directly impact on the requirement for spectrum. Analogue TV: the network was originally designed for 4 TV channels, each needing about five 8 MHz transmission channels to provide national coverage. Channel 5 has since used the white space of Band IV/V but does not have the national coverage of the other 4 channels. Between 2008 and 2012, digital switchover occurs, at the end of which all analogue channels will be decommissioned. The number of channels required for analogue is set to 4 until 2010, then reducing down to zero (as digital switchover progresses) by DVB-T: using historical data from 2000 to 2008, S-curve parameters are estimated and extrapolated to forecast total traffic for For offered traffic, there are three applications of relevance: television, radio and datacast. 16 Note that during the period of switchover (i.e ) a reuse factor is imposed on the number of channels in order to take account of the increasing demand for spectrum from digital. 5-8

42 5. Assumptions for Forecasting the Demand for Services DVB-H application is a mobile TV application and is not yet in operation in the UK. This is a new application and is assumed to start in Based on the introduction of mobile TV in Italy, the number of channels is assumed to reach twelve fairly quickly, and audience figures are projected to be sufficient to justify the continuation of this service but to not be sufficient to stimulate demand from broadcasters for significant additional channels using the technology. DVB-T2: there are plans to convert one of the 6 DVB-T multiplexes into a multiplex for HD channels. This is likely to have three channels at the outset, but may in time be able to carry more with improvements in coding schemes. We assume that these channels will operate under DVB-T2 and will increase by about one channel per year until DVB-S is the technology that operates satellite broadcasting, operating at frequencies just about our 10GHz boundary. It is able to broadcast a much greater capacity than terrestrial broadcasting. If the demand for more channels and higher definition content cannot be satisfied by terrestrial capacity, one of the options open to broadcasters would be to shift some content to this technology. We make this assumption in the results presented in this report Broadcast Radio Broadcast radio 17 uses four technologies: AM, FM, and DAB, with DRM on the horizon. AM radio (Medium wave and Long wave) operates in bands below 2 MHz and is outside the scope of this study. FM has a fixed frequency band between 88 and 108 MHz, with a fixed transmission spacing of 300 KHz. This demand for spectrum is assumed to continue throughout the forecast period. DAB currently operates in 7 multiplexes of 1.7 MHz bandwidth each with approximately 9 radio stations on each. There is the potential for 12 multiplexes according to current spectrum management. In the Business As Usual case, we assume that this figure of 63 stations modelled will remain broadly constant until A third digital technology such as DRM could be launched, possibly taking over from FM - and/or DAB - in the same way that DVB-T is doing from analogue TV. In this case, the same spectrum would be required as currently plus, perhaps, some additional spectrum during the transition period, so that the needs of consumers with devices operating each of the different technologies could be accommodated. We have not projected any additional demand for spectrum arising from the introduction of DRM or other additional radio technologies. 17 There has also been a growth in Internet radio station traffic, but this counted as is streaming audio, rather than broadcasting. 5-9

43 5. Assumptions for Forecasting the Demand for Services Baseline Traffic and Subscriptions Demand Forecasts This section summarises the baseline ( Business As Usual ) key demand inputs used for Broadcast-TV and Broadcast-Radio. Broadcast TV and Radio Figure 8: Broadcast TV - Total Offered Traffic (Number of Channels) We assume that broadcast radio continues to demand around 60 channels since this is sufficient to cover current requirements and internet radio is also showing growth. Note that in the above Figure, DVB-S falls outside the frequency bands of interest for this study. 5-10

44 5. Assumptions for Forecasting the Demand for Services 5.4 FIXED WIRELESS ACCESS (FWA) The offered traffic for FWA is estimated in the same way as for SRW-Wi-Fi. Therefore, this section only considers the calculation of subscriber numbers of FWA 18. The actual number of FWA subscribers in 2005 is used to define the initial ratio of FWA to fixed broadband subscriptions. This is forecast to grow over the period to a final figure of 1-2% of fixed broadband connections. Growth in subscribers is assumed to be linear Baseline Traffic and Subscriptions Demand Forecasts This section summarises the baseline ( Business As Usual ) demand inputs used for FWA, Figure 9: Fixed Wireless Access - Total Subscriptions by Data Application Figure 10: Fixed Wireless Access - Total Offered Traffic: Data Applications in Terabytes 18 For traffic assumptions and inputs refer to the section on SRW Wi-Fi. 5-11

45 6. THE BUSINESS AS USUAL BASELINE SCENARIO 6.1 DEMAND FOR SERVICES AND SUPPLY OF SPECTRUM In Section 5 we have considered the demand for services in the baseline Business As Usual scenario, and the traffic that those services would generate. In this section we consider the impact that this demand is likely to have on pressure for spectrum, taking into account changes in network technology that we anticipate. Business As Usual is the baseline as discussed in Section 1. It has moderate growth in traffic, mature and relatively slowly changing networks, and some moderate release of spectrum. The tabulated parameters below are extracted from Table Business As Usual Demand Growth in the demand for data in this scenario is up to 40% per annum, with a shift from voice to data. There will be a shift in broadcasting, post digital switchover, towards higher video resolutions, and some shift to satellite. Demand (traffic) Total Demand growth per annum (amount in MB equivalent) Total Demand growth (mix) Demand growth (driven by growth in technology) Moderate Shift To Data Some Business As Usual - Technology and Network Architecture There will be very little change if any in the current industry structure, with: Some fibre to the kerb/home/office replacing copper/xdsl in the last mile Introduction of Femtocells will go ahead at a modest pace Limited convergence / integration of broadcast and cellular wireless / fixed. Network architecture Cell proliferation Fibre plus wireless Industry fragmentation Convergence and integration of services Slow Slow Some Slow Cellular traffic carried by femtocells 20% Business As Usual - Availability of Spectrum There will be a moderate rate of release and /or re-use of spectrum, with limited change to promote and drive re-use/release of spectrum. Availability of Spectrum Rate of release/re-use of spectrum Spectrum coordination across regions Moderate Moderate 6-1

46 6. The Business As Usual Baseline Scenario 6.2 BUSINESS AS USUAL - RESULTS Business As Usual - Offered Traffic (by Service) Figure 11 and Figure 12 show offered total traffic by service in TeraBytes / hour (average hour not busy hour, uplink and downlink combined) for Business As Usual. Figure 11: Total Offered Traffic (Terabytes) in Business As Usual by Service Figure 12: Total Offered Traffic (Terabytes) in Business As Usual by Service (log scale) 6-2

47 6. The Business As Usual Baseline Scenario Business As Usual - Offered Traffic (by Application Group) Figure 13 and Figure 14 show assumed offered total traffic by Application group in TB (average not busy hour, uplink and downlink combined) for Business As Usual. Figure 13: Total Offered Traffic (Terabytes) by Application Group Figure 14: Total Offered Traffic (Terabytes) by Application Group (log scale) In the short term, offered busy hour traffic is dominated by cellular voice. Voice calls made by DECT also contribute strongly. In 2007, voice is indicated to form greater than 60% of the total. and web browsing are also strong components of traffic in the short term from both WiFi and cellular. In the medium to long term, short range wireless (WiFi and ultra-wideband) dominates traffic in the busy hour. Offered traffic from voice calls is minor compared to Web browsing, , video streaming and downloading. In addition, machine to person and telemetry / machine to machine are significant applications increasingly being used. 6-3

48 6. The Business As Usual Baseline Scenario Business As Usual - Demand for Spectrum Overall spectrum demand as shown in Figure 15 arises predominantly from continuing demand for both TV broadcasting and cellular, where both are operating very close to supply of spectrum in this band. Short range wireless begins to dominate the picture towards the end of the period. Figure 15: Spectrum Demand (MHz) in Business As Usual by service A key feature of this is the reduction in spectrum requirements for cellular and broadcast TV over the period. This is due to a number of technology improvements. In cellular technology, the move to 3G and 4G (LTE) technologies provides significantly improved spectral efficiency and reuse, up to a factor of ~40 compared with 2G (see Appendix A) 19. Since we are predicting an increase in traffic of ~15, this is more than sufficient to absorb all the growth in spectrum demand in this scenario (although not sufficient for the most aggressive growth scenario All You Could Want ). The introduction of femtocells, whilst assumed to carry 20% of the traffic, require dedicated 5MHz channels for operation, and in this scenario have a net negative effect on overall capacity / Hz (although they may still other benefits, such as providing indoor coverage in rural areas). This is a very conservative view, although the channel allocation issues are real. One solution could be for network operators to share a common femtocell channel. In Broadcast TV the improvement is due to improved coding efficiency which is assumed to be available from around For Short Range Wireless the demand is increasingly significantly, and demand will need to be provided for in bands other than the dominant 2.4GHz band used today. This is likely to be increased use of 5.8GHz and increased use of UWB technology. 19 Improvements in spectral efficiency arise from the move from 2G to 3G then 4G giving improved spectral efficiency, number of sites, and reuse (factor of ~40). 6-4

49 7. DEMAND GROWTH AND SPECTRUM IMPACT FOR INDIVIDUAL SERVICES 7.1 INTRODUCTION In this section we consider growth in demand for services and the consequent pressure on spectrum given changes in network technology. The following services are covered: Cellular (fully mobile device to/from fixed infrastructure) Short range wireless (mobile to mobile or to access point) Broadcast TV (fixed device to multiple devices, one way, video plus audio) Broadcast Radio (fixed device to multiple devices, one way, audio only) Fixed Wireless Access (nomadic or fixed device to/from a fixed device). Results follow a standard format with discussion of the applications supported service and the overall pattern of growth, followed by spectrum demand forecasts for each scenario. The level of detail provided for each service depends on the complexity and scale of spectrum demand. Additional data is included in Annex E for completeness Demand for Services The demand for services figures are taken from the scenarios and their assumed growth patterns given in Section 3. They are represented in common units of TeraBytes / hour Demand for Spectrum The demand for spectrum for each service is shown by scenario. These charts do not necessarily follow the increase in demand for services, since there may be differences in the way services are delivered, particularly in spectral efficiency of wireless networks. 7-1

50 7. Demand Growth and Spectrum Impact for Individual Services Pressure on Spectrum Pressure on spectrum is shown in two ways depending on the context. It is either shown as a line chart with supply, demand, and difference plotted, or as a graph of supply plotted against demand with time, as shown in Figure 16 below. Very high risk of serious spectrum shortage Nominal Supply = Demand line (1 : 1) Spectrum Demand (MHz) High risk of spectrum shortage Some risk of spectrum shortage Low risk of spectrum shortage Spectrum Supply (MHz) Figure 16: Spectrum Demand (MHz) in Business As Usual by service In this graph the mid-way line running approximately from corner to corner of the chart is the nominal Supply = Demand line, i.e. a ratio of 1 : 1 between the two quantities. Where demand falls below this line there is less risk of spectrum shortage; where demand falls above this line there is a significant risk of serious shortage. The Supply and Demand axes are calibrated in MHz. Although it may be thought that it is desirable to make supply exactly equal demand, in practice this is not possible since demand is not an absolute quantity but depends heavily on cost of spectrum, cost of networks, price levels of applications etc. So whilst we discuss potential shortage of spectrum in this report, this needs to be interpreted as a shortage according to a particular set of assumptions. In practice, if cellular spectrum is in relatively short supply, networks operators will tend to introduce more cell sites, and will introduce fewer advanced high traffic services (or will tariff them very highly). Similarly if spectrum is readily available, networks will tend to be less dense and operators will be more generous with their offerings. There is a similar grey area between surplus and shortage for other services and the size of this grey area can vary from one service or technology to another, depending on: The peakiness of the demand for the service (busy hour to average, and bursts versus steady traffic) The spatial distribution of demand for the service The effect of congestion on the networks (graceful degradation, or long queues, or unacceptable drops in service quality) The alternatives available, if demand exceeds supply. 7-2

51 7. Demand Growth and Spectrum Impact for Individual Services The reader should therefore not interpret potential shortage in this report as meaning that consumers will necessarily experience poor Quality of Service, nor attempt to draw conclusions about appropriate price levels for spectrum from this report. With this in mind we have used the interpretation that should Demand be more than approximately 60% of Supply then non-broadcast networks will begin to be affected by pressure on spectrum, potentially leading to a limitation on the services they introduce, and/or leading to escalating network costs. For broadcasting, the boundary is more tightly defined, since it is less common to price off demand or to allow a gentle degrade service quality at peak times. Demand could reach 80% or 90% of supply without there being any detrimental effects, and there is an alternative delivery channel, in the form of broadcasting via satellite. 7-3

52 7. Demand Growth and Spectrum Impact for Individual Services 7.2 UNDERSTANDING THE MAIN SOURCES OF THE GAINS IN EFFICIENCY IN THE NETWORKS The casual reader, inspecting the charts showing large growth in traffic for most services and applications, in particular for cellular data and for short range wireless, would expect this to translate into a surge in the demand for spectrum and a severe net shortage, starting in a few years time and getting steadily worse over the remainder of the study period. However, the final outcome is somewhat different there is some pressure on spectrum over the next five years, but after that the position eases. The main reason for this is the improvements in spectral efficiency expected from the networks, in particular cellular networks, as they migrate infrastructure and users to new technologies 20. This section sets out the assumptions regarding these efficiency gains. The gains come from two sources: Spectral efficiency bits per second per Hertz per site or sector Improvements in compression and coding algorithms Spectral Efficiency - Bits / Sec / Hz / site (/Sector) The Table below gives the assumptions regarding the spectral efficiency of the networks and technologies. They represent the combined effect of both Frequency Reuse and Single-site Spectral Efficiency. The contributions from the Frequency Reuse factors are in the column shaded yellow. The combined effect of the two factors is given in the columns shaded blue. 20 There are further effects from a range of factors in the model, such as increases in the number of useful sites and sectors, and from the deployment of femto cells. The effect of these is outlined in Appendix A: they may make a difference - particularly if femto cell technology takes off - but, in most of our six scenarios, their impact is less than that from the gains from spectral efficiency and from coding improvements. 7-4

53 7. Demand Growth and Spectrum Impact for Individual Services Service, Technology and spectrum allocation Cellular Freq. Reuse Factor Assumptions: Combined effect of Frequency Reuse and single site spectral efficiency GSM/ GPRS / EDGE 1/ MHz 1/ G UMTS (+ entry-level HSDPA) 2100MHz G (HSPA+) G LTE Mobile WiMax MBMS G/3.5G 700/900 MHz G 700/900 MHz Short Range Wireless Analogue Cordless** 1/ DECT** 1/ Bluetooth 1/ UWB 1/ WiFi 1/ Broadcast Radio Analogue (FM) 1/ DAB DRM 1/ Broadcast TV Analogue TV 1/ DVB-T 1/ DVB-H 1/ DVB-T2 1/ DVB-S Fixed Wireless Access Fixed WiMax 1/ TD/CDMA 1/ Proprietary 1/ Table 6 : Changes in Spectral Efficiency, by Technology and Spectrum Band The charts below show these results graphically first for cellular technologies, then for broadcast. 7-5

54 7. Demand Growth and Spectrum Impact for Individual Services Figure 17 : Changes in Spectral Efficiency Cellular Technologies At present the bulk of the cellular traffic is carried over 2G networks with the rest going over 3G. By 2025 the projection is for 2G to be decommissioned and for the most of the traffic to be carried over 4G networks and technologies. Cellular spectral efficiency improves from the current figure of about 0.04 to about 1.3 a factor of approximately 30. Figure 18 : Changes in Spectral Efficiency Broadcast Technologies At present the bulk of the broadcast traffic is carried over analogue networks and DVB-T. Analogue TV is scheduled for switch-off over the next five years, and by 2025 the projection is for most broadcast traffic to be carried over DVB-T and DVB-T2. In the broadcast sector (particularly for TV) is likely that there will be some balancing feedback between gains in spectral efficiency and the demand for terrestrial channels, with any surplus channels being delivered via satellite. 7-6

55 7. Demand Growth and Spectrum Impact for Individual Services How The Spectral Efficiency Assumptions Vary Between Scenarios The same set of assumptions for spectral efficiency are used for all six scenarios, with the following exceptions: In Scenario 2, Wire Free World, the efficiency of 4G cellular technologies is assumed to rise from 1.3 to 1.7 over the period In Scenario 3, All You Can Want, the efficiency of 4G cellular technologies is assumed to rise from 1.3 to 2.1 over the period In Scenario 3 there is also a further 60% improvement in the spectral efficiencies of DVB-T, DVB-H, and DVB-T2, taken up by growth in the number of channels and higher video resolution per channel. The mian effect of this is to allow for greater growth in the number of terrestrial high-definition TV channels, rather than to reduce the demand for spectrum Changes in Coding Efficiency, Compression, and channel content The Table below gives the assumptions made regarding changes in coding efficiency over time. Figures are informed by past trends and by the view that of the different types of content, video offers the greatest opportunities for enhanced compression, and text and data offers little or no opportunity. Still images, audio / voice traffic lie between these extremes. Application Annual Multiplier (in BAU) Long term effect (in BAU) Comments Voice x 1.03 x 1.65 Audio: 3% pa - Messaging (SMS, MMS, IM) x 1. x 1.00 Condensed data: 0% - E Mail x 1.02 x 1.40 Mix of Data / text, attachments, images, video scope for efficiency coding is quite low WWW x 1.03 x 1.65 Mix: data, text, images, video scope for efficiency coding is moderate Video Streaming on Demand x 1.06 x 2.69 Video: 6% pa codec improvement, no channel bloat (raw figures for Traffic are already in MB). Changes to assumptions in other Scenarios (WFW, AYCW, DYST, FRAG, and REUSE)

56 7. Demand Growth and Spectrum Impact for Individual Services Application Annual Multiplier (in BAU) Long term effect (in BAU) Comments TV (live video) x 1.06 x 2.69 Video: 6% pa codec improvements, no channel bloat Changes to assumptions in other Scenarios (WFW, AYCW, DYST, FRAG, and REUSE) TV resolution: add 5% pa: WFW, 10% pa in AYCW. Radio (live audio) x 1.03 x 1.65 Audio: 3% pa codec improvements, no channel bloat Codec improvements: 5% pa: AYCW and REUSE. Net effect: annual factor becomes 1.01 in WFW and AYCW, x1.11 in REUSE. Datacast x 1.02 x 1.40 Mix of data and images Content bloat in datacast - slightly faster than in TV - add 10% pa: WFW and REUSE, 20%: AYCW. Codec improvements as TV (5% pa: AYCW and REUSE). Net effect: annual factor is x0.92 in WFW, x0.88 in AYCW, x0.97 in REUSE PMSE x 1.02 x 1.40 Video, but may be less amenable to applying new coding schema Gaming x 1.06 x 2.69 Mix: video, some shift to caching, local intelligence on device Download x 1.04 x 1.95 Mix: Audio, Video, some data - P2P x 1.04 x 1.95 Mix: Audio, Video, some data - Telemetry & M2M x 1.02 x 1.40 Mix: Audio, Video, some data - M2P and P2M x 1.02 x 1.40 Mix: Audio, Video, some data - Table 7: Changes in Coding Efficiencies, by Application - Corresponding changes in PMSE - as in TV. Net effect: the annual factor is x0.97 in WFW and in AYCW, x1.07 in REUSE. These efficiency gains are applied to most technologies within each service, but not to the following: 2G /2.5G cellular, analogue broadcasting, and DECT

57 7. Demand Growth and Spectrum Impact for Individual Services 7.3 CELLULAR GROWTH The penetration of cellular services in the UK is now close to saturation. However, with many people now having a second mobile line, e.g. separate ones for work and personal use, or separate voice/data devices, the number of subscriptions is continuing to rise. There appears to be a continuing shift of calls to from fixed to cellular, as evidenced in traffic volumes and reflecting the statistics in Ofcom s The Communications Market 2008, reporting that 70% of mobile users use their mobile phone to make calls within the home. Currently, voice and messaging are the most widely used and established applications, accounting for three-quarters of the traffic in the busy hour, with the rest coming from data applications such as , www, and video streaming. Traffic is carried over 2G/2.5G and 3G networks, and over three different spectrum bands, with the typical user oblivious or indifferent to the routing choices made by the technology. Cellular includes the following applications to mobile devices with wide-area coverage: Voice Messaging (SMS, MMS) Internet access Direct video streaming Broadcast TV, Radio, Data (e.g. teletext) and PMSE services Gaming Music and video download, e-commerce, m-commerce File sharing (YouTube etc.) Telemetry and Machine - Machine communications Navigation and tracking Machine-to-person communications and vice versa. 7-9

58 7. Demand Growth and Spectrum Impact for Individual Services Demand for Cellular Services (TeraBytes) Figure 19 and Figure 20 below show the predicted growth in cellular traffic for scenario (linear and log scales). Figure 19: Cellular Traffic Demand Growth (TeraBytes) for six scenarios (linear scale) Figure 20: Cellular Traffic Demand Growth (TeraBytes) for six scenarios (log scale) 7-10

59 7. Demand Growth and Spectrum Impact for Individual Services Demand for Spectrum for Cellular Services (MHz) Figure 21 below shows the resultant growth in spectrum demand for cellular, across all bands. The key differences between this and traffic growth is the forecast future improvements in spectral efficiency of cellular technologies, as discussed in the previous section. This significantly reduces the amount of spectrum required to meet the high growth in service demand under some scenarios. Note that we assume that there is no cellular spectrum demand above 4GHz in the timescale of this study. Figure 21: Demand for Cellular Spectrum (All bands, 100MHz 4,000MHz) The initial point of interest is the growth in demand over the period to 2012, caused by increased data traffic being offered over existing networks. From approximately 2012, spectrum demand tends to decrease as the move to more spectrally efficient 3G and 3.5G continues (with 3G/3.5G terminals and network infrastructure becoming more commonplace) 21. From 2016 to 2019, 3G900 is assumed to take off as a result of 900MHz spectrum refarming. Services at 700MHz also become significant, rebalancing traffic towards the band below 1GHz. Also, from around 2016 it appears that improvements in network technology and the benefits of femtocells will have matured, and we show a tendency for spectrum demand to start increasing again in the higher growth scenarios. However, it is also possible that further technology and network improvements may be introduced by that time. 21 Also note the discussion in Section1.3: Effects of pressure on spectrum. 7-11

60 7. Demand Growth and Spectrum Impact for Individual Services Supply of Spectrum for Cellular Services Figure 22 below shows the total of spectrum available for cellular over all bands, with some assumptions about future release. Figure 22: Supply of Cellular Spectrum (All bands, 100MHz 4,000MHz) The assumed step increase in available spectrum shown above at around 2013 is from the Digital Dividend (release of spectrum from broadcast below 1GHz). The increase in spectrum availability beyond 2010 is the 2.6GHz band becoming available. The shifts around 2015 / 2017 are caused by refarming of spectrum, and the delay in it becoming available for the new technology. There is an additional block of spectrum assumed to become available from public sector release (MOD) providing capacity from 2015 / 2016 in the 2.7-4GHz region. 7-12

61 7. Demand Growth and Spectrum Impact for Individual Services Pressure on Spectrum Here we show the net surplus / shortage of spectrum for cellular. We see here that the pressure on spectrum for cellular services generally becomes less acute from around 2012 because of the shift from 2G to 3G, and subsequently (in 2014) because of the Digital Dividend release. Again, beyond 2014 the demand for cellular services is largely matched by the availability of spectrum, except in our most aggressive growth scenario. This arises because of a combination of two factors: the growth in demand for services is paralleled by growth in spectral efficiency of advanced technologies and use of smaller cells, and there is additional spectrum made available through this period. The aggregated picture across bands is shown in Figure 23 below, with an alternative representation in Figure 24. Figure 23: Surplus / Shortage of Cellular Spectrum (All bands, 100MHz 4000MHz) Figure 24: Surplus / Shortage of Cellular Spectrum (All bands, 100MHz 4000MHz) Referring to the explanation of this chart given in Section 7.1.3, we can see that there is a high risk of spectrum shortage for the top three scenarios though until Referring to Section 1.3 however suggests that this will tend to limit growth of new high bandwidth applications rather than causing specific user problems with current applications. 7-13

62 7. Demand Growth and Spectrum Impact for Individual Services Differences in pressure on spectrum between bands Because of the historical split between 900MHz and 1800MHz GSM operators, there is still a significant difference in spectrum supply between the two bands. In Figure 25 and Figure 26 below we show the pressure on spectrum for bands below 1GHz and above 1GHz, showing the extent of the potential shortage over the next 5 years. Figure 25: Surplus / Shortage of Cellular Spectrum (100MHz 1000MHz) Figure 26: Surplus / Shortage of Cellular Spectrum (All bands, 1000MHz 4000MHz) 7-14

63 7. Demand Growth and Spectrum Impact for Individual Services Discussion on each Scenario for cellular Business As Usual. In Dense Urban environments, we predict that the traffic demand growth of a factor of 10 over the next 15 years will broadly be met by spectrum supply, through a combination of new technology and additional spectrum allocation. There is however a risk of shortage in , particularly in the 900MHz band. After 2012, the view of cellular in all bands combined indicates that for the most part, there is enough spectrum across the period of analysis for Business As Usual. Wire Free World. In this scenario, with traffic demand growth around 8x that of Business As Usual, we see considerable risk of excess spectrum demand below 1GHz up to After this period (post Digital switchover) there is probably sufficient spectrum below 1GHz. Above 1GHz demand for spectrum tends to climb after 2015, but this is partially balanced by increasing supply (notably at 2.6GHz). Overall, as expected, there is more risk of shortage in Wire Free World than in Business As Usual. All You Could Want. With this very high level of demand growth there is a very serious risk of excess spectrum demand below 1GHz in the period up to 2012 (we show a net shortage in the latter part of this period). There is also a longer term issue from 2020 onwards, although it is likely that new technologies or strategies that are not visible to us today may be introduced by this time. It is worth noting that pressure on spectrum from rapid growth in demand is heavily compensated by improvements in network technology and additional site rollout requiring major investments by network operators. Dystopian. The reduced growth in demand in this scenario leads to ample supply of spectrum in both bands, despite rather less spectrum becoming available to the commercial market than in other scenarios. Industry Fragmentation. The spectrum demand / supply in this scenario is broadly similar to Business As Usual, and give no major cause for concern. Re-Use. Again, there appears to be little cause for concern in this scenario, particularly after 2012, when demand for spectrum falls rapidly. 7-15

64 7. Demand Growth and Spectrum Impact for Individual Services 7.4 GROWTH IN SHORT RANGE WIRELESS Short range wireless consists of a number of technologies that have different user focus. Analogue cordless 22 and DECT are specifically used for voice, providing the air interface between a handset and a fixed line connection. DECT phones make up a significant proportion of the fixed line telephone network. Voice traffic on landlines has been decreasing in recent years and this is factored into the projections for DECT traffic. WiFi is mostly used for broadband type activities such as , web browsing, video streaming etc. Three environments can be considered; home, work and public access hotspots. Traffic mostly consists of a mix of data applications. Bluetooth and Ultra-wideband (UWB) are mostly used to connect the request(s) of a person to a machine. For example, telephone headsets, mobile phone or PDA to PC synchronisation and PC to printer. This activity has grown significantly in recent years and is envisaged to continue. UWB is not yet operational but is scheduled to come onstream in the next year or so. Applications: Voice Messaging (SMS, MMS) Internet access Broadcast TV, Radio, Data (e.g. teletext) and PMSE services Gaming Music and video download, e-commerce, m-commerce File sharing (YouTube etc.) Telemetry and M-M communications Navigation and tracking Machine-to-person communications and vice versa. This is the same range of applications as cellular, but as a wireless tail, with the omission of Direct Video Streaming. 22 It is assumed that the majority of such devices currently in use are of DECT rather than analogue cordless. The latter operates at frequencies of around 30MHz and is therefore outside the spectrum of interest for this study (100MHz to 10GHz). 7-16

65 7. Demand Growth and Spectrum Impact for Individual Services Demand for short range services Figure 27: Short Range Wireless Traffic Demand Growth (TeraBytes / Hour) for six scenarios Figure 28: Short Range Wireless Traffic Demand Growth (TeraBytes / Hour) for six scenarios (log scale) For Short Range Wireless we are showing a very large growth in traffic, mainly due to growth in fixed-line data and increasing use of wireless communications in the home, workplace and other locations, including leisure venues. Growth in Short Range Wireless traffic is likely to outstrip that of cellular. 7-17

66 7. Demand Growth and Spectrum Impact for Individual Services Demand for Short Range Wireless spectrum (MHz) Figure 29 shows predicted growth in spectrum demand for short range wireless. Figure 29: Demand for Short Range Wireless Spectrum Spectrum demand for Short Range Wireless is calculated as an average for Dense Urban, and does not take into account demand quantisation 23. However, it is growth that is more significant, showing that after demand for spectrum will increase significantly Supply of spectrum for Short Range Wireless services Figure 30: Supply of Short Range Wireless Spectrum Figure 30 shows predicted growth in spectrum supply for Short Range Wireless services. The step change around 2012 comes from assumed public sector release below 4GHz. It should be noted that the majority of this spectrum allocation is Ultra Wide Band (shared spectrum), which is not yet used significantly as a technology, but which we expect to be needed in the medium term to satisfy demand. 23 For Wifi and other SRW technologies, a minimum channel bandwidth is needed if users are to achieve the high instantaneous data rates of 10Mbps+ that the technology can support. This quantity of standby bandwidth would not vary with traffic, and is not included in the figures shown here or used in the model. There are similar quantisation effects in the demand for other technologies e.g. in 2G cellular, demand for spectrum is quantised in multiples of 200kHz, paired and possibly scaled up to allow for frequency reuse. These quantisations of demand are in general not represented in the model. Some exceptions are made on a case by case by case basis, eg in considering the provision of spectrum for femto cells. 7-18

67 7. Demand Growth and Spectrum Impact for Individual Services Pressure on Spectrum Figure 31 and Figure 32 show pressure on spectrum for Short Range Wireless services. Figure 31: Surplus / Shortage of Short Range Radio Spectrum Today s apparent spectrum surplus seems to be contrary to experience of WiFi in dense urban areas where congestion occurs. We believe that a significant amount of congestion is due to inefficient use of spectrum, the number of different networks and limits on user numbers rather than spectrum shortage per se 24. Increasing introduction of 5.8GHz systems will alleviate this, as will moves to licensed bands for business critical services. A complicating factor for Short Range Wireless will be the extent to which UWB succeeds. There is significant unused UWB spectrum available today (it is shared and cannot be used at high power). If widely adopted it could take some of the very high bandwidth short range traffic from WiFi, although propagation through walls will be a limiting factor. There is a risk of serious spectrum shortage arising in the All You Could Want scenario, after 2020, even allowing for UWB usage. However, at present we are using some conservative assumptions about traffic handling which apply for WiFi it is likely that these will be significantly improved over the next 10 years for UWB. Figure 32: Surplus / Shortage of Short Range Radio Spectrum 24 See report (to be published by Ofcom): Estimated Utilisation of Licence Exempt Spectrum

68 7. Demand Growth and Spectrum Impact for Individual Services Discussion on each Scenario for Short Range Wireless Business As Usual. We do not see any problems meeting the overall demand for spectrum under this scenario, although the UWB spectrum allocation is masking a specific problem around WiFi. Wire Free World. Again, we do not see any problems meeting the demand for spectrum under this scenario, although there is some risk of shortage below 4GHz from Again, there is a specific issue around WiFi spectrum usage, particularly at 2.4GHz. All You Could Want. Within this scenario we show problems arising after 2020 below 4GHz although if we assume that UWB is successful it is likely that demand can be satisfied. Given the nature of ad-hoc WiFi networks we would expect to see problems in the near future in the 2.4GHz band. Dystopian. We do not see any problems meeting the demand for spectrum under this scenario, despite the very limited amount of spectrum being available. Industry Fragmentation. We do not see any problems meeting the demand for spectrum under this scenario. Re-Use. We do not see any problems meeting the demand for spectrum under this scenario. 7-20

69 7. Demand Growth and Spectrum Impact for Individual Services 7.5 GROWTH IN BROADCASTING - TV Terrestrial television broadcasting is in the process of moving to digital. A service is in operation (i.e. based on DVB-T) that currently covers approximately 85% of the population. As analogue is switched off, digital transmitters will be established to ensure national coverage by Currently there are five analogue TV channels and 37 digital terrestrial standard definition TV channels. Handheld TV (DVB-H) has not yet started in the UK and there are still choices to be made about how it might be delivered. When launched, it is likely that the most popular TV channels, given audience figures, will be available (Indepen and Aegis, 2005). It is estimated that 12 channels will provide the core of this service when started. In order to deal with the demand for High Definition Television (HDTV), there are already plans to convert one of the six DVB-T multiplexes to allow for 3 high definition channels on the terrestrial platform. Broadcasting behaves somewhat differently from cellular; a change in the number of viewers or listeners does not alter the demand for spectrum for broadcasting. Demand for broadcasting is therefore expressed in terms of the broadcasters demand for transmission capacity, rather than the consumers demand for content and reception. This distinction applies throughout this analysis and report. Applications: Broadcast TV, Radio, Data (e.g. teletext) and PMSE services. 7-21

70 7. Demand Growth and Spectrum Impact for Individual Services Demand for Broadcasting - TV (Mbytes) Figure 33 shows the predicted growth in Broadcasting - TV traffic for each of the scenarios. Figure 33: Broadcasting - TV Demand Growth (MByte) for Six Scenarios In this chart, demand figures are expressed as the demand for national transmission of content streams, rather than as the addition of demand for those streams. The overall picture to 2013 is dominated by digital switchover. After that the growth of High Definition TV is balanced by the progress of compression technology, allowing additional channels within the same traffic level, except in All You Could Want, where demand grows dramatically. Since broadcast TV can be provided efficiently though satellite transmission, we assume that any excess demand for spectrum is provided by satellite. 7-22

71 7. Demand Growth and Spectrum Impact for Individual Services Demand for Broadcasting - TV services (MHz) Figure 34 below shows the predicted growth in spectrum demand for Broadcasting - TV traffic services. Figure 34: Demand for Broadcasting - TV Spectrum (all bands, constrained) Demand has been modelled as being broadly constrained to be within or close to what is available. The rest is assumed to be provided over DVB-S satellite links above 10GHz Supply of spectrum for Broadcasting - TV services Figure 35 below shows the predicted growth in supply of spectrum for Broadcasting - TV services, again broadly limited by what is currently available. Figure 35: Supply of Broadcast TV Spectrum 7-23

72 7. Demand Growth and Spectrum Impact for Individual Services Pressure on Spectrum Figure 36 below shows the predicted pressure on spectrum for Broadcasting - TV services. Figure 36: Broadcast TV - Spectrum Demand (MHz) Under Six Scenarios (log scale) Figure 37: Broadcast TV - Spectrum Demand (MHz) Under Six Scenarios From these charts we can see that broadcast spectrum is, as would be expected, highly utilised. However, since satellite is regarded in this model as the overflow option, the fact that demand equals supply is not necessarily a problem. 7-24

73 7. Demand Growth and Spectrum Impact for Individual Services 7.6 GROWTH IN BROADCASTING - RADIO Current radio technologies considered in this study include FM radio and DAB. FM radio occupies all 20MHz of spectrum between 88 and 108MHz. DAB has been in operation for several years and typically offers the same channels as FM plus others. There are now 7 DAB multiplexes, each carrying 8 to 12 radio stations. There is a third radio technology available, DRM, but at present this is not used in the UK Demand for Broadcasting - Radio services (Mbytes) Figure 38 shows the predicted growth in traffic for Broadcasting - Radio. Figure 38: Broadcasting - Radio Demand Growth (MByte) All scenarios offer the same demand Demand for Broadcasting - Radio services (MHz) Figure 39 shows the predicted change in spectrum demand for Broadcasting - Radio. Figure 39: Change in Spectrum Demand for Broadcast Radio. The downward trend in spectrum demand is caused by improvements in signal coding rather than a decrease in number of channels. 7-25

74 7. Demand Growth and Spectrum Impact for Individual Services Supply of spectrum for Broadcasting - Radio services Figure 40 below shows the predicted growth in supply of spectrum for Broadcasting - Radio services. Figure 40: Change in Spectrum Supply for Broadcast Radio The downward shift in supply in Dystopia around 2015 is assumed to be withdrawal of spectrum either in the DAB or analogue bands Spectrum Oversupply / Shortage Figure 41: Change in Spectrum Surplus / Shortage for Broadcast Radio 7-26

75 7. Demand Growth and Spectrum Impact for Individual Services 7.7 GROWTH IN FIXED WIRELESS ACCESS In the context of this study, FWA is a means of delivering broadband access, competing with cable. It currently serves under 0.1% of households in the UK. In a recent study undertaken for Ofcom (Plextek, 2006 [30]), the potential of FWA in the UK was assessed as being limited, and this is factored into the forecasts of demand for the service. Unlike other services, which are typically most heavily used in urban areas, Fixed Wireless Access is primarily assumed to be rolled out in rural areas, and it is therefore this Geotype that is modelled here. Traffic mix is taken to be similar to that for broadband. Applications: Voice Messaging (SMS, MMS) Internet access Direct video streaming Broadcast TV, Radio, Data (e.g. teletext) and PMSE services Gaming Music and video download, e-commerce, m-commerce File sharing (YouTube etc.) Telemetry and M-M communications Navigation and tracking Machine-to-person communications and vice versa. 7-27

76 7. Demand Growth and Spectrum Impact for Individual Services Demand for Fixed Wireless Access services (TeraBytes) Figure 42 and Figure 43 below show the predicted growth in Fixed Wireless Access. Figure 42: Fixed Wireless Access Demand Growth (TeraBytes) Figure 43: Fixed Wireless Access Demand Growth (TeraBytes) (log scale) For FWA we assume that the bulk of the demand is in rural neighbourhoods (constituting 10% of the population and 80-90% of the land area). In high growth scenarios, which might result if a major telecoms provider were to adopt FWA as a high speed local loop technology for rural areas, we estimate 1-2% of UK households using FWA for all their traffic (approximately 10% of rural households). Otherwise the history of Fixed Wireless Access in the UK suggests that demand will be modest. Services and traffic offered will be similar to Short Range Wireless (WiFi). 7-28

77 7. Demand Growth and Spectrum Impact for Individual Services Demand for Fixed Wireless Access services (MHz) Figure 44 below shows the predicted growth in spectrum demand for Fixed Wireless Access services. Figure 44: Fixed Wireless Access Growth (MHz) Under Six Scenarios For modelling purposes we assume that spectrum needed for FWA cannot be reused for cellular in urban areas for interference reasons, since many subscribers would be within interference range of urban areas should penetration reach 1-2% of households. This assumption is perhaps conservative and there could potentially be opportunities to reuse spectrum if required Supply of spectrum for Fixed Wireless Access services Figure 45 shows predicted growth in spectrum supply for Fixed Wireless Access. Figure 45: Fixed Wireless Access Growth (MHz) Under Six Scenarios The additional spectrum is from allocations in the 2 5 GHz region, resulting in sufficient spectrum, as shown in Figure 46 below. 7-29

78 7. Demand Growth and Spectrum Impact for Individual Services Pressure on Spectrum Figure 46 and Figure 47 show pressure on spectrum for Fixed Wireless Access. It is clear that under most scenarios there is little risk of spectrum shortage for the foreseeable future. Only beyond 2019 do we see a significant issue, and only then under the All You Could Want scenario. Should the demand for communications take off to the extent modelled in All You Could Want, it is likely that we would see a much wider roll-out of fibre to the home, which in turn would reduce the pressure on FWA as a technology. Figure 46: Fixed Wireless Access Growth (MHz) Under Six Scenarios Figure 47: Fixed Wireless Access Growth (MHz) Under Six Scenarios 7-30

79 8. DEMAND FOR SPECTRUM ALL SERVICES AND ALL SCENARIOS In this section we consider the total supply and demand for spectrum across all services, and under the various scenarios that we have considered. These results should however be treated with caution since they may mask some of the potential service-specific shortages highlighted earlier in this report. In Figure 48 and Figure 49 below we show the pressure on spectrum for all services in the highest three growth scenarios (All You Could Want, Wire Free World and Business As Usual). The two figures present the same data, but with an expanded scale in Figure 49. It can be seen from Figure 48 and Figure 49 that there is a risk of spectrum shortage under the two most aggressive demand scenarios. Figure 48: Overall pressure on spectrum from 100MHz to 10GHz Figure 49: Overall pressure on spectrum from 100MHz to 10GHz (expanded scale) 8-1

80 8. Demand for Spectrum All Services and All Scenarios The implication of this overall picture is that for the higher growth scenarios, spectrum could be a limiting factor if not managed carefully. As noted in Section 1.3, a shortage of spectrum may not show itself as poor quality of service or as customer complaints, but may instead act as a brake on rollout of new applications, particularly where these are data-intensive. There appears to be sufficient spectrum to meet the need, but it will not necessarily be in the ideal band, and active management and refarming will be needed to make the most of it. Looking more closely at this picture we can see the same features that were clear from the cellular service results shown in Section 7.3.4, namely increasing pressure on spectrum over the next few years as usage of data applications grows ahead of new network rollout. The falloff as 3.5G and 4G is widely introduced is apparent, leading to a general reduction in pressure on spectrum for the following few years. Short Range Wireless spectrum is a significant concern, particularly in the 2.4GHz band, with WiFi being heavily used in some areas today, and UWB not yet nearing widespread introduction. Finally, we see the rapid growth in demand for Short Range Wireless having a major impact on spectrum requirements for the two most aggressive scenarios of All You Could Want and Wire Free World. 8-2

81 9. ULTRA-DENSE URBAN AND OTHER NEIGHBOURHOOD TYPES The majority of this report has considered Dense Urban as the geographical type ( geotype ) of interest, with Dense Rural being used for Fixed Wireless Access. These have been considered as the main areas of interest since they cover the majority of the population and the majority of urban areas. However, Ultra-dense Urban is the limiting case, and is exemplified in certain areas of the City of London, transport termini, sporting venues etc. 9.1 SPECTRUM DEMAND IN ULTRA-DENSE URBAN AREAS The two services where demand in Ultra-Dense urban is a particular issue are Cellular and Short Range Wireless. Broadcast spectrum demand is not affected by spectrum density, and Fixed Wireless Access is primarily of interest in rural areas. Finally, Backhaul, which was initially considered in some aspects of this work, mostly operates at above 10GHz and does not have limitations in urban areas since high capacity fibre is readily available. Cellular in Ultra-Dense Urban For cellular coverage we have assumed a similar approach to network planning as used today, where macro-cells and micro-cells are used in low and medium-density population areas up to Dense Urban. We have also assumed increased density of cell sites in future, particularly in Dense Urban areas, to meet future demand in the higher growth scenarios these would typically be street level microcells and picocells. Ultra-Dense Urban areas are today served by significantly increasing the number of cell sites, either through use of microcells, or through remote radio heads (equivalent to picocells). These techniques enable network operators to cover very high population densities in constrained geographic areas at a manageable cost, without excessive burden of planning issues to erect transmission towers, and without requiring unreasonable amounts of spectrum. Cell spacing can be as little as a few hundred metres, and buildings will have indoor coverage specifically designed to provide very high capacity through use of low power cells. This pattern of feedback between the pattern of urban demand and the distribution of base stations is borne out by evidence from both Ofcom s sitefinder web pages 25, and from discussions with operators. We assume that similar techniques will be used in future, with the introduction of more micro cells and pico cells, coupled with femtocells when these become widely available to provide the very high capacity, very short range, coverage that data-centric applications may require. Network architectures are assumed to develop in ways that make reasonably good use of sites and spectrum e.g. using one frequency band or carrier for wide area coverage, complemented by another one providing higher capacity in smaller areas of high demand density

82 9. Ultra-Dense Urban and Other Neighbourhood Types For this reason we have not predicted spectrum requirements based on covering ultradense urban areas in the same way as is assumed for lower population density regions; rather we have made the assumption that, as today, spectrum will be dimensioned to take account of the majority of users and locations, and that special techniques will be applied elsewhere. Short Range Wireless in Ultra-Dense Urban The widespread use of WiFi that we see today will continue and grow in urban areas. This poses specific difficulties, since there is no management of quality between the very many independent service providers (which include various cafes, venues, corporate offices as well as the public service providers such as Openzone and T-Mobile). We expect continued pressure on traffic within the 2.4GHz band, with congestion and Quality of Service issues growing over the next few years. As the use of very high traffic applications such as audio and video downloads and video streaming grows, spectrum shortage is likely to become more serious particularly in town and city centres. In addition there may be increasing competition for short-range wireless spectrum, with other technologies and services seeing to make greater use of this unlicensed band we have treated the band as if it were 100% available for Wifi and bluetooth. In later years we make the assumption that UWB will become increasingly popular, offering very high traffic rates over short ranges, operating at low power and sharing spectrum with other users. If UWB is not successful in the marketplace then other measures may need to be taken if demand is to be met. 9-2

83 10. CONCLUSIONS AND RECOMMENDATIONS Throughout this report we have presented the results of our forecasts and modelling, and noted a number of conclusions. In this section we bring together those conclusions to give our view on areas of spectrum shortage between now and CONCLUSIONS Demand for Services In our consideration of services that require significant amounts of spectrum, and that are likely to be of most interest in understanding potential shortages in the future we have considered the following services: Cellular, Short Range Wireless, Broadcast TV, Broadcast Radio, Fixed Wireless Access and Backhaul. Of these services it is apparent that Cellular and Short Range Wireless are of most interest since it is for these services that demand is likely to change significantly over the next years. Demand for Broadcast TV channels could also increase significantly, although we recognise that this can be overflowed to satellite transmission, and hence is likely to be of lower priority than those services that fundamentally require mobility and which have no alternative means of delivery. Increasing take-up of free-to-air satellite TV could mean that in 10 or 20 years time most subscribers can get everything that Digital Terrestrial TV offers them via other means,with relatively few subscribers dependent on the terrestrial distribution channel.. Broadcast Radio allocations are forecast to be relatively steady. Rapid growth seems unlikely and there is even some potential for rationalisation in some scenarios, but it seems likely that two of the three technologies (FM, DAB, and DRM) will continue to require a similar amount of spectrum to the current allocation for most of the study period. For Fixed Wireless Access, the traffic growth forecasts depend on both the takeup of the FWA services, and the growth in fixed-line traffic per household. In the high growth scenario it is likely that the demand for spectrum will exceed supply from about In the other scenarios, there is little pressure on FWA spectrum. Backhaul has not been considered in detail in this report, since most of the frequencies in use in communications networks are above the range of interest, and those below 10GHz could potentially be displaced relatively easily if required Applications Our primary forecasts for demand growth have come from previously published results, extrapolated as necessary to Sources such as Ofcom, UMTS Forum and WiMAX Forum have been consulted and incorporated into the model. We have considered a wide range of applications, including a number of breakout options such as 3D television, growth of mobile virtual worlds etc. We have then amalgamated these applications into a number of scenarios and applied growth rates to them according to the scenario of interest. This gives a range of demand figures for traffic (Bytes/second) varying by a factor of 100 between Business As Usual and All You Could Want. 10-1

84 10. Conclusions and Recommendations The following conclusions do not arise from this work, but are key inputs to it from previous studies, and are included here for completeness: The main applications that will drive growth in spectrum demand for cellular are video streaming and downloads, with increasing demand for machine-machine and person to/from machine communication. Voice will cease to dominate as these new applications grow in popularity. and web browsing will bolster the data traffic over the next 6-8 years but will then be overtaken by the other applications. The main applications that will drive spectrum demand for Short Range Wireless are web browsing, , gaming, and video streaming, recognising that much of this traffic is in the home Pressure on Spectrum Pressure on Spectrum - Cellular For the three high growth scenarios of Business As Usual (baseline), Wire Free World and All You Could Want, we believe that there is likely to be significant pressure on spectrum over the next 3-4 years as take-up of data services increases ahead of new network rollout. Operators may be hard pressed to cater for this additional traffic by adding sites and upgrading technology and infrastructure. This will tend to limit growth of new applications. The pressure on spectrum is likely to become less acute from around 2012 because of the shift from 2G to 3G+, and subsequently (in 2014) because of the Digital Dividend release. By 2015, the position should have eased, with most traffic migrated away from 2G. and with increases in available and useful spectrum, e.g. at 2.6GHz. Beyond 2015 the demand for cellular services is largely matched by the availability of spectrum, except in our most aggressive growth scenario. This arises because of a combination of two factors: the growth in demand for services is paralleled by growth in spectral efficiency of advanced technologies and use of smaller cells, and there is additional spectrum made available through this period. It is necessary to consider frequencies below 1GHz separately from higher frequencies because of differences in spectrum available in the two bands. We believe that there will be a disproportionate shortage of spectrum below 1GHz over the next few years. Pressure on Spectrum Short Range Wireless For Short Range Wireless, demand is currently considerably less than supply. However, there are grounds for concern. Overall we are forecasting very high growth in demand for spectrum for Short Range Wireless services in the All You Could Want scenario, resulting in a clear shortage of spectrum, possibly in 5 or 10 years time. This arises as a result of very high growth in fixed line data traffic, and Short Range Wireless being used as the link to the PC or other terminal in the home and other locations including work and leisure venues. Additionally, there is a greater risk of the supply of spectrum being in the wrong bands or the wrong technologies. Current demand, and perhaps much of the future growth, is for Wifi, in partiicular at 2.4GHz. which only has some 15% of the SRW spectrum supply. 10-2

85 10. Conclusions and Recommendations Unlike cellular, we cannot predict with accuracy the potential improvements in frequency usage that may develop for SRW. This is because, as an unlicensed technology, users are free to install and use equipment that may have very poor spectral efficiency. The short range nature of WiFi and UWB, and the very high burst rates that these technologies aim to deliver, means that this may result in a reasonably graceful degradation of network performance as congestion starts to have an impact, reducing data rates and perhaps having a knock-on impact on the volume of offered traffic. The problems may be localised, confined to a small percentage of the country, but, because demand tends to cluster in time and space, it will affect a significant proportion of users of the technology in the very high growth scenarios. Pressure on spectrum Broadcast TV and Radio As noted above, the high growth scenarios that we have modelled do allow for a dramatic increase in the number of broadcast TV channels. However, we assume that this can be provided over satellite and therefore do not consider that it will create significant pressure on spectrum competing with applications that are fundamentally mobile. Pressure on spectrum Fixed Wireless Access For Fixed Wireless Access the majority of demand is in rural areas and we are currently forecasting significant spectrum surplus for the foreseeable future. Only beyond 2019 do we see a significant issue, and only then under the All You Could Want scenario. Should the demand for communications take off to the extent modelled in All You Could Want, it is likely that we would see a much wider roll-out of fibre to the home, which in turn would reduce the pressure on FWA as a technology Technology and network improvements It is clear from the results that improvements in technology are critical to meeting the demand for spectrum over the coming decade in the high-volume services, ie cellular and short range wireless. Previously published results for growth in traffic demand show factors of ten or more compared with today; in the higher growth scenarios this can be much higher as use of wireless data become ubiquitous. Although it is possible to increase supply of spectrum by a significant amount, it is difficult to achieve much more than a factor of 2-3, partly because of the short ranges exhibited in the higher frequency bands. Similarly it is difficult to envisage all this additional capacity in the wide area being provided by additional macrocells the cost, logistics and acceptability of these are already partially limiting factors. There is some scope for additional cellular capacity to be provided through the use of micro and pico cells, but the economics of this will place a limit on the capacity gains that this can provide. We envisage that they will deliver approximately a further doubling of capacity due to rollout of additional sites and sectors. This leaves a very large gap to be made up by the introduction of new technologies that improve spectral efficiency, such as newer modulation formats, improved reuse etc. We have assumed in this work that these improvements come into being over the next 5-10 years, and are adopted by the cellular industry. However, it is clear that if this does not happen then lack of capacity could form a barrier to growth. 10-3

86 10. Conclusions and Recommendations In the Broadcast services there is some scope for improvement in technology via better coding algorithms and signal compression but these are not likely to reduce the demand for terrestrial spectrum, which is broadly in equilibrium: any excess content streams may migrate to satellite, and if additional terrestrial channels become available, the capacity will be taken up by additional channels or higher resolution video RECOMMENDATIONS In order to minimise any negative effect on consumers that may arise from a shortage of spectrum over the period from now until 2025 we make the following recommendations: Spectrum allocation and re-farming Ofcom should examine closely the likely shortage of spectrum in cellular that we predict may occur over the next 3-5 years. The shortage is if anything likely to be more pronounced below 1GHz than in the range 1GHz to 4GHz although this depends on the mix of customers and traffic across the five operators networks. Depending on the rate of growth of cellular applications, this pressure on spectrum may require more rapid rollout of 3G+ than has previously been envisaged. Ofcom should consider the extent to which additional spectrum may be released for use in Dense Urban, Less Dense Urban, and Ultra-Dense Urban areas. This might be from Defence use 26 or from existing civil users in the bands above 1GHz. We note that there is little pressure on spectrum in rural areas and that regional release covering major towns and cities would suffice to significantly improve capacity of networks for the vast majority of the population. Ofcom should examine the options for Short Range Wireless. WiFi in particular has, grown to be an important service, and as its importance grows further this will sit uncomfortably with its position sharing spectrum in an unlicenced band. Ofcom should recognise that the balance of supply and demand is naturally prone to fluctuations over the short to medium term. Assuming that the pressure on spectrum does increase over the next few years - albeit mitigated by the commercial strategies of operators there may be calls for the regulator to do even more to alleviate the shortage. Such calls should only be heeded if at the time it seems likely that the growth in demand will outstrip the significant gains in spectral efficiency which are likely to come into effect over the next ten to fifteen years. We recommend that, as far as possible, Ofcom should minimise transition periods when bands are re-farmed since this introduces significant short-term inefficiencies in the use of spectrum. Ofcom should continue its general stance of issue licences on a service- and technology-neutral basis. 26 PA Consulting Group recently completed a report for UK Ministry of Defence, reviewing the current and future use of Defence spectrum: munications/publicconsultations/ukdefencespectrummanagement htm 10-4

87 10. Conclusions and Recommendations Monitoring growth Ofcom should continue to update demand forecasts for mobile services, particularly for Cellular and Short Range Wireless. These should be monitored alongside the results presented in this report to follow the likely track of pressure on spectrum. Ofcom should continue to track improvements in technology of wireless standards such as LTE and MIMO, to maintain an understanding of the extent to which these will live up expectation in terms of additional traffic capacity. We also recommend that Ofcom repeat this work in , to track progress against this report. By doing this Ofcom will be in a position to understand which of the scenarios we have presented is most likely to come to fruition, and the extent of pressure on spectrum over the period. Efficient use of spectrum Within the results presented here, femtocells are not projected to deliver a significant capacity gain for 3G networks, primarily because a large fraction of cellular traffic is unsuited to femtocells, and they require a dedicated (5MHz) channel which detracts from the operator s overall network capacity. We recommend that Ofcom consider this issue to see whether more should be done to stimulate use of 3G femtocells through alternative spectrum allocations (e.g. shared channels for low power 3G applications). Other Ultra Wide Band technology will be important to meet growth forecasts in the top three growth scenarios. We recommend that Ofcom continue to monitor this to ensure that there are no obstacles to UWB introduction should it be successful as a standard and be necessary in the UK. We note that machine-to-machine applications are forecast to grow significantly over the next 10 years; success of these will demand changes to both equipment cost and tariff models to allow the proliferation of low cost devices. 10-5

88 APPENDIX A: ASSUMPTIONS FOR NETWORK ARCHITECTURE AND SPECTRUM USE FOR FORECASTING THE DEMAND FOR SPECTRUM

89 A: Assumptions for Network Architecture and Spectrum Use for Forecasting the Demand for Spectrum In the Spectrum Shortage Model, there are a number of input assumptions that are used in the process of converting the Offered Traffic into a demand for spectrum, and then into a net surplus or shortage of spectrum. These cover the following areas: 1. Determining how traffic is likely to be routed to different technologies, and up/down link asymmetry effects 2. Converting the offered traffic for each application, into a set of traffic figures calculated on a common basis for all of the technologies (same units) 3. Determining the required capacity that the networks need to be able to provide in the busy hour 4. Breaking down the demand by neighbourhood type and then down to sites 5. Converting the demand from traffic volumes into spectrum (MHz) 6. Determining how these assumptions change over time in the period to The assumptions in each of these six areas are described in the six subsections below. A.1 ROUTING TRAFFIC AND UPLINK / DOWNLINK EFFECTS A.1.1 Routing of Offered Traffic between Different Network Technologies For Cellular, the offered traffic is routed to the available technologies as follows: Only route traffic to compatible technologies Only route traffic to technologies whose networks are still live Traffic is more likely to be routed to technologies that more users have access to, i.e. which are present on more devices. All things being equal, traffic is more likely to be routed via the fastest / most suitable technology and to the spectrum band that has the most capacity. The typical routeing factors for data applications are of the order of: 0.01 to for 2G/2.5G, 0.1 to 0.25 for 3G, 1.0 for 3.5G, 2.0 for WiMAX, and 4.0 for 4G (1.0 for 4G if deployed below 1GHz. A proportion of this traffic is routed via femtocells, rather than via the macro layer network. This proportion is assumed to be zero for 2G, 2.5G, and for any refarmed spectrum below 1GHz. For the remaining cellular technologies, the assumptions on the proportion of traffic routing via femto are set out in Section A.5 below. For non-cellular traffic, the demand data is provided by technology, so there is no need to consider routeing calculations to get from the whole service down to individual technologies. A.1.2 Mapping Offered Traffic onto Uplink / Downlink Offered Traffic maps onto spectrum differently for each application: Voice: all inbound and outbound traffic uses both the uplink and the downlink. Messaging and inbound traffic uses the downlink, outbound uses the uplink. Inbound traffic is about four times outbound traffic. A-1

90 A: Assumptions for Network Architecture and Spectrum Use for Forecasting the Demand for Spectrum For P2M and M2M, inbound traffic uses the downlink, outbound uses the uplink. For broadcasting: downlink only. Some types of channel require more spectrum that others, and this is separate from HDTV / DVB-T2 i.e. there are differences in spectrum requirements within the Standard definition DVB-T. For PMSE: uplink only. For all other Applications (web browsing, video streaming, downloads), inbound traffic is typically four to ten times the outbound traffic, and inbound traffic uses the downlink, outbound traffic uses the uplink. A.1.3 Effect of the routing parameters The routing of traffic to the uplink or downlink has little effect on the net surplus or shortage of spectrum, since the net position is calculated on an aggregate basis. But it does provide one of the grounds for concern in situations where the available spectrum is greater than the demand, but only marginally so, perhaps by a factor of 2 or thereabouts. The other cases where a surplus of this order of magnitude could turn out to be quite restrictive are: If the service is offered over several different frequency bands and technologies and the installed base of handsets has a markedly different mix of technologies from the installed base stations. Moderate differences are handled by the model, routing traffic pro rata to the mix of devices and the attractiveness of each technology, but in extreme cases there may be a shortage in the spectrum used by the handsets and a surplus in the spectrum used by the base stations. If the spectrum is licensed to individual operators whose installed base of sites does not maintain a reasonable equilibrium with their share of traffic. So one operator may have a surplus of spectrum (perhaps having completed a new build, or made a major effort to recruit customers and traffic) whilst another operator has a shortage. If the demand for the service is highly localised (although the introduction of neighbourhood types in the calculation goes a considerable way to overcoming this, in particular for cellular) A.2 CONVERTING DEMAND INTO COMMON UNITS FOR EACH TECHNOLOGY All traffic volumes for all Applications are converted into one common set of units, namely MB / hour. The conversion factors vary, for example: If the raw demand is in MB / year there is a conversion factor of x 1/365 x 1/24 (about x ). If the raw demand figure is in video channels, the conversion factor will be of the order of: x2 (Mbps, channel content, net of compression) x 1/8 (Mb to MB) x 3600 (seconds to hours), i.e. about x1,000. Conversion factors can vary over time as coding efficiencies improve. This varies between applications and technologies but typically yields an improvement of a factor of 1.5 to 2 over the study period. A-2

91 A: Assumptions for Network Architecture and Spectrum Use for Forecasting the Demand for Spectrum A.3 DERIVING THE REQUIRED CAPACITY FOR TRAFFIC A.3.1 Factors for the Busy Hour For Cellular traffic, typically the load in the Busy Hour is about 4x the average over the whole week (24 hours, 7 days). This is reduced to x2 for telemetry and M2M, because some of their usage will be more geared to 24/7 operation, and increased to x6 for any PMSE traffic over Cellular. Similar factors apply to Short Range Wireless (also x4),and Fixed Wireless Access (x3 for voice and messaging, x4 for most data) with the same adjustments for Telemetry, M2M, and PMSE. For Broadcasting, and for broadcasting applications over Backhaul, there is no real difference between busy hour and the rest of the week, so the Busy Hour factor is just x1. For Backhaul of everything other than broadcast content, the Busy Hour factor will be similar to that for the same application when delivered to the user: x1 for broadcasting applications, x2.5 to x3.5 for others, reflect the scope for some modest aggregation of variations in traffic on the backhaul link. A.3.2 Factor for Required Headroom For Cellular there is a requirement to maintain capacity at a higher level than the average busy hour traffic, in order to allow an acceptably high call completion rate for voice in the Busy Hour, of the order of say 98%. This translates into a capacity requirement, for voice, of some 1.5 times the busy hour traffic. The requirement is somewhat lower for other data, because it is delivered as packets and is more tolerant of delay, so a factor of 1.25 is assumed for this traffic, except for telemetry and M2M, some of which may require continuous uninterrupted capacity, so a required headroom factor of x2 is assumed for these applications. For Short Range Wireless, and Fixed Wireless Access, the required headroom is greater, because spectrum availability is not pooled in the same way as for cellular. The required headroom is assumed to be the order of 10x to 20x the busy hour traffic. For Broadcast, there is no additional headroom requirement, so the factor is x1. For Backhaul, a small headroom margin is required, of the order of x1.5 for any voice traffic and x1.25 for all other traffic. A-3

92 A: Assumptions for Network Architecture and Spectrum Use for Forecasting the Demand for Spectrum A.4 BREAKING DOWN DEMAND BY NEIGHBOURHOOD TYPE AND THEN DOWN TO SITES A.4.1 Breakdown of Demand by Neighbourhood Types The land area of the UK is divided into seven neighbourhood types according to population density, ranging from hyper dense urban 27, through dense urban, down to unpopulated areas. The effect of these on the demand for spectrum varies slightly by Service: For Cellular and Short Range Wireless, demand is distributed more or less pro rata to the number of people in each neighbourhood type. In those areas where the daytime and night-time populations are very different, the higher of the two figures is used, i.e. capacity planning is assumed to serve each type of neighbourhoods when that type is fairly busy, rather than simply aiming to provide for the population of the UK at a set time of the week, e.g. overnight, or during a weekday morning. For Fixed Wireless Access, this approach is modified slightly to reflect the likelihood that the service is likely to be attractive for users in rural areas, but less attractive in urban areas. We assume that 80% of the users and traffic are in rural neighbourhoods, i.e. the 90% of the land area that contains only some 10% of the population. Within the rural neighbourhoods, it is expected that each household or person is more likely to subscribe to FWA if they live in a less dense rural area, but the greatest density of demand (measured in terms of lines or MB per square kilometre or MHz) is likely to be highest in the more dense rural areas. Accordingly, much of the analysis of the demand for spectrum for FWA is focused on these dense rural neighbourhoods, rather than the urban areas. For broadcast, and broadcasting backhaul, demand is more or less the same in all neighbourhood types. A.4.2 Availability of Useful Sites The availability of sites to a network will vary by service and by neighbourhood type, and it may also vary over time as new technologies roll out or deepen their coverage. For cellular technologies, the number of sectors per cell is also important, because this scales up the capacity pro rata. The availability of sites is a particularly important factor for cellular; other services are less affected, and the impact on the demand for spectrum for broadcast services is limited. The assumption is made that the number of useful sites available to each cellular network is different from the current headline figure of 10-20,000 or so sites. The assumptions informed in part by taking examples from Ofcom s sitefinder facility - are as follows: For 2G /2.5G spectrum and services; some 12,000 sites are available, and - at least for those sites where they are needed for capacity - they currently have equipment installed. 60% of these available sites are in urban areas, and these generally have three sectors each, giving a figure of about 22,000 for the total number of useful urban sectors. No further growth in available sites or sectors is expected. 27 Hyperdense urban covers areas such as major venues for sporting events, high density business districts such as the City of London, and some major airports. A-4

93 A: Assumptions for Network Architecture and Spectrum Use for Forecasting the Demand for Spectrum For 3G: site density is about ¾ of that for 2G in the hyper dense neighbourhoods, half the density of 2G in other urban areas, falling to 1/10 if the density of 2G sites in the most rural areas. The rural numbers will grow to match the current site numbers for 2G in , and then stabilise. For 3.5G and MBMS: 20% of the 3G sites are available and equipped. The numbers will grow to match those for 3G in 2015, and then for MBMS only - double again for For new cellular technologies, the availability of sites is assumed to be constrained initially, because it is not practicable to roll out to the whole country instantaneously, and there may be conflict within a site e.g. positioning on masts, or regarding obtaining planning consents or negotiating with the owner of the site. These constraints are assumed to be overcome, with the number of useful sites available assumed to take about five years to catch up with the existing technologies: For 4G and WiMAX: no sites are currently available / equipped. The numbers available will grow to be similar to those for 3G in 2020, and then double again (only for 4G) by For 3G 900 and 4G 700, for few sites are available, but then in the period the availability of sites will catch up with that for the same technology in the higher bands. The effects of these changes on the number of sites available in urban areas i.e. the 10% or so of the country where 90% of the population and traffic is to be found 28 - are summarised in the Table below. Number of useful urban sectors available Current technologies: 2G/2.5G GSM/ GPRS / EDGE 23 k 23 k G/2.5G 1800 GSM/ GPRS / EDGE 23 k 23 k G UMTS (2100 etc) 12 k 17 k 23 k 23 k 3.5G (HSPA) (2100 etc) 2 k 13 k 23 k 34 k Emerging technologies: 4G LTE (2600 etc) k 26 k 68 k Mobile WiMAX (2600 etc) k 9 k 9 k MBMS 2 k 9 k 17 k 17 k Refarming technologies: 3G/3.5G 700/900 1 k 7 k 23 k 23 k 4G 700/ k 26 k 26 k The additional sites in 4G technologies are assumed to be largely microcell and picocell implementations, which allow far higher densities than at present. Figure 50: Number of Useful UK Urban Sites Available to Networks Using each Technology 28 The figures cover the first three neighbourhood types: hyperdense urban, dense urban, and less dense urban (/suburban), all areas with a population density of greater than 400 people per km2. A-5

94 A: Assumptions for Network Architecture and Spectrum Use for Forecasting the Demand for Spectrum For Short Range technologies, in general each traffic-bearing wireless hub or cordless handset is treated as one site and the assumption is that some 20% of these sites are in use in the busy hour, the others are unable to carry any traffic because they are not paired with any terminal devices. A.5 CONVERTING THE DEMAND FROM TRAFFIC VOLUMES INTO DEMAND FOR SPECTRUM A.5.1 Frequency Reuse This allows for the fact that in many networks, several carrier frequencies are required in order to provide every site with the capacity that one carrier frequency can carry for a single site: For 2G / 2.5G cellular, this is assumed to be 13. Whilst 7 or 9 is achievable in theory, in practice 12 to 15 is regarded as a sensible planning assumption. For other cellular technologies a reuse of 1 is assumed. For short range wireless, a frequency reuse is assumed to be 3 for DECT and analogue cordless (generally indoors), 10 for the other technologies, due to their inherent lack of network planning. For Broadcasting, reuse is assumed to be 6 for analogue, 5 for digital terrestrial, and 1 for DVB-S (satellite). For Fixed Wireless Access, reuse is set at 3. For Microwave Backhaul, reuse is set at 2. A.5.2 Spectral Efficiency This governs how much data can be transmitted and received reliably for each unit of spectrum available. All figures are expressed in bits per second per Hertz. For cellular technologies, spectral efficiency is 0.5 for 2G / 2.5G 29, 0.17 for 3G, 1.0 for 3.5G, and 1.3 for 4G. For WiMAX it starts at 0.67, but over the period it rises to 1.3 from For MBMS it is For SRW it is set at 0.1 for analogue cordless, 0.33 for DECT, 0.5 for Bluetooth, and 1.0 for UWB. For Wifi,.it starts at 1.0 and rises to 1.9 over the period , but all of the efficiencies are halved to allow for the overhead of the Ethernet architecture. For broadcast it is set to approximately 0.4 for analogue technologies, 0.9 for digital radio, 2.5 to 3.0 for most digital TV technologies, 4.5 for DVB-T2, and 1.5 for DVB-S. For FWA and Backhaul it is set to The figure may be lower for pure 2G infrastructure, but in neighbourhoods where there is a capacity constraint, we assume that the operators generally install 2.5G infrastructure. A-6

95 A: Assumptions for Network Architecture and Spectrum Use for Forecasting the Demand for Spectrum A.5.3 Allowing for Spectrum Required for Femtocells In Cellular, some spectrum will be required to be reserved for femtocells, in order to carry their share of the traffic. This is assumed to be as follows: 3G/3.5G: first significant use in 2012, rising steadily to 60Mz - 2x5MHz for each of six networks - in Thereafter remaining at 60MHz. 4G: requires additional spectrum from 2019, ultimately adds a further 50% on top of the 3G spectrum from 2021, i.e. we assume that operators have to support an installed base of two different types of incompatible equipment, but there is some cooperation or variation in launch dates or switch off dates when allocating spectrum for the less popular generation of technology. WiMAX: First significant use in 2014, rising to 21MHz (12 x 1.75MHz carriers) in No further changes thereafter. A.5.4 Proportions of Traffic Staying on the Macro Layer vs. Diverted to Femtocells Femtocells will take a share of the cellular traffic off the main macro / micro layer cells, as follows: 3G: 0% until 2012, rising to 1% in 2015, 15% in 2020, and with a long-term level of 20% 4G LTE: two years behind the trajectory for 3G, with the same long-term level of 20% WiMAX: two years behind the trajectory for 3G, but with a long-term ceiling of 10% rather than 20%. Taken together with the previous subsection, arguably these figures take quite a conservative view of the assistance that femtocells provide in offloading traffic from the cellular networks over the time period up to The projected gains in spectral efficiency summarised below - are not reliant on a massive programme of diverting traffic using femtocell technology. A.5.5 Overall Spectral Efficiency Improvements The combination of the factors discussed above gives rise to a potential improvement as follows: For Cellular: Content compression: improvement of a factor of about 1.5. Spectral efficiency: improvement of a factor of 2, from 0.5 to about 1.0. Frequency reuse: from mostly 1/13 (some at 1/1), to 1/1. Number of useful sites and sectors (for each combination of technology and band): increase of a factor of 1 to 2. Combined effect: 1.5 x 2 x 10 x 1.5, an overall factor of approximately 40. For Short Range Wireless: Content compression: a factor of 1.5 to 2 (a figure of 1.7 is used to illustrate). A-7

96 A: Assumptions for Network Architecture and Spectrum Use for Forecasting the Demand for Spectrum The shift from DECT to Wifi, Bluetooth and UWB gives a gain in spectral efficiency from 0.33 to , a factor of about 2 but this is offset by a worsening of frequency reuse (1/3 to 1/10). Among the data technologies, there is some improvement in spectral efficiency from 0.5 to about 1.0. There is no material change to frequency reuse. The combined effect is therefore an overall factor of approximately 1.7 x 2 x 3/10 x 2, i.e. about 2. There are differences between adding more wireless base stations and adding more traffic to the same number of base stations. Broadly speaking, it is the traffic per base station that drives the demand for spectrum, rather than the number of base stations. For Broadcasting : Over time, there are improvements in signal compression algorithms a fact or of x 2, or slightly more - that allow each channel to be squeezed into fewer Mbps. This is largely taken up by additional channels and by increases in the level of video resolution that each channel offers. The switchover from analogue to digital improves the spectral efficiency for TV from 0.3 to about 3, i.e. a factor of 10. This switchover may also reduces frequency reuse slightly, from 1/6 to 1/5, giving a combined factor of approximately 12. The switchover from FM radio to DAB provides a similar gain in efficiency. There is some slight inefficiency during transition periods, as the installed base of equipment takes time to catch up with the technology. Each content stream may need to be broadcast over two or three different technologies and bands, for example the terrestrial TV stations are broadcast over digital as well as analogue, and many terrestrial radio stations are currently broadcast over FM, DAB, and DVB-T. There is no gain from increasing the number of transmitter sites. The overall picture for Broadcasting is an improvement in efficiency of a factor of about 25 : X12 from the switchover to from analogue to digital X2 from the gains in coding algorithms (generally taken up by offering additional content and/or higher resolution). For FWA : there are gains from the rollout of additional sites, but no improvements in spectral efficiency or frequency reuse. A.6 CHANGES IN SPECTRUM AVAILABILITY A.6.1 Developments in the Existing Cellular Bands Cellular - 2G/2.5G 900MHz. This spectrum is refarmed over the period , it is reused for 3G fairly quickly, being redeployed after lying fallow for about one year. Cellular - 2G/2.5G 1800MHz. This is refarmed in 2017, reused for 4G LTE, 1-2 yr later. A-8

97 A: Assumptions for Network Architecture and Spectrum Use for Forecasting the Demand for Spectrum Cellular - 3G/3.5G 2100MHz. No changes or refarming between now and The 20MHz of 3G TDD spectrum gets used for MBMS. A.6.2 Other Developments for Cellular and for Broadcast TV 2.6GHz spectrum is released and usable (network deployment plus some device deployment) from , as follows: 20MHz to 3G/3.5G, 50MHz to WiMAX, 80MHz to 4G LTE, 20MHz to MBMS, and 20MHz to FWA. TV Broadcasting spectrum 470MHz - 860MHz is refarmed in , with 290MHz remaining in use for TV (including 10MHz for DVB-H), 72MHz for 4G LTE. 100MHz of new spectrum at about 3GHz is released in 2016, taken up by 4G LTE. A.6.3 Developments for Other Services Short Range Wireless, Broadcast Radio, FWA, and Backhaul Short Range Wireless: 575MHz allocated for UWB, and 150MHz allocated circa 2014 for expansion of Wifi/Bluetooth. No changes to allocations for DECT, analogue cordless, or the existing WiFi band. Broadcast Radio: no changes to existing allocations, i.e. no expansion in DAB and no introduction of DRM. Fixed Wireless Access: When the 2.6GHz band is released, FWA takes 20MHz of it. There may be reassignments between the different FWA technologies, but this is assumed to be done efficiently and to not have a material impact on the supply of spectrum. Backhaul: (i) for wireless comms networks; all usage is above 10GHz. (ii) for broadcasting networks: there is a ramping down of spectrum allocations for backhaul in 2012, as the demand ramps down due to the switch off of analogue TV broadcasting, but the technology continues to be used. A.7 HOW THESE ASSUMPTIONS VARY OVER THE PERIOD TO 2025 Many of these assumptions remain unchanged over the whole period to The main ones that change over time other than the release and refarming of spectrum described immediately above - are as follows: The availability of the different mobile technologies, e.g. the switch off of 2G and the launch of WiMAX and LTE For broadcasting, the factors for the conversion of demand figures (expressed in channels) into MB/hour The factors for required headroom and for spectral efficiency, in a few cases (e.g. to reflect likely improvements in 3G / 3.5G and in WiFi) The availability of useful sites The share of traffic that is carried by the femtocell layer, and the amount of spectrum that those cells need. A-9

98 APPENDIX B: DEMAND FORECASTS SOURCES

99 B: Demand Forecasts Sources The following sources were used either for specific data points and forecasts, or to inform the thinking and methodology behind the demand forecasts. Alptekin, A., Levine, P. and Rickman, N. (2007), Estimating spectrum demand for cellular industry in the U.K. Technical report, Department of Economics, University of Surrey, the U.K. Analysys (2001), The Long Run Incremental Cost of UK Mobile Network Cost, Report for Oftel. Analysys (2002), The Long Run Incremental Cost of UK Mobile Network Cost, Report for Oftel. Analysys Mason (2005), Spectrum for non-government services , Final Report for the Independent Audit of Spectrum Holdings. Analysys, DotEcon and Mason (2004), Assessment of options for allocating available spectrum within VHF Band III ( MHz) and L-Band ( MHz): Final Report to Ofcom Cave, M. (2002) Review of Radio Spectrum Management An independent review for Department of Trades and Industry and HM Treasury. Cave, M. (2005) Independent Audit of Spectrum Holdings an Independent Audit for Her Majesty s Treasury. Cave, M. and W. Webb (2003) Spectrum licensing and spectrum commons Where to draw, Warwick Business School, Papers in Spectrum Trading, No.2 DotEcon (2001) Estimation of Fixed to Mobile Price Elasticities, Report of BT. DotEcon (2001) Fixed to Mobile Substitution, A Report Prepared for BT. DotEcon and Analysys Mason (2005) Allocation options for selected bands Final reports for Ofcom. Forrester (2008), European Mobile Forecast: , March Kensuke Fufuda, Kenjiro Cho, and Hiroshi Esaki (2005). Impact of Residential Broadband on Japanese ISP Backbones, ACM SIGCOM Computer Communications Review, volume 35, Number 1: January Mason and DotEcon (2004) Value of UWB Personal Area Networking Services to the United Kingdom, Final Report to Ofcom. Mobile TV Joint UMTS Forum/GSMA Work Group (2008) "Sustainable Economics of Mobile TV Services". White Paper OECD (2006), IPTV: Market Developments and Regulatory Treatment, JT Ofcom (2005), A guide to the spectrum Framework Review. Ofcom (2005), The Communications Market Ofcom (2005), Ultra Wideband. Ofcom (2006), The Communications Market Ofcom (2007), The Communications Market B-1

100 B: Demand Forecasts Sources Ofcom (2007), The Communications Market: Broadband Digital Progress Report, Ofcom (2007), The Future of Digital Terrestrial Television: Enabling new Services for viewers, Ofcom (2008), The Communications Market Ofcom (2008), The Nations & Regions Communications Market 2008 (May). Oftel (2001), Review of Price Control on Calls to Mobiles, A Consultative Document issued by the Director General of Telecommunications. Open IPTV Forum White Paper (2007). Ovum (2005), UK Broadband Status Report, June 2005 Plextek (2006), Wireless Last Mile Final: Report to Ofcom, QinetiQ (2006), Cognitive Radio Technology: A Study for Ofcom QINETIQ/06/ Quotient Associates (2006), Supply and demand of spectrum for Programme Making and Special Event in the UK: Final report to Ofcom, RedM (2006), Improving the Sharing of the Radio Spectrum: Final Report to Ofcom, Sagentia (2008), PMSE: Future Spectrum Access, report to Ofcom, Spectrum Framework Review a Consultation on Ofcom s views as to how radio spectrum should be managed, Ofcom UMTS Forum (2003), 3G Offered Traffic Characteristics, Report No. 33, UMTS Forum (2005), Magic Mobile Future , Report No. 37, Wimax-Forum (2007), Spectrum Requirement for Mobile Wimax Equipment to Support Wireless Personal Broadband Services, B-2

101 APPENDIX C: SPECTRUM DEMAND IN WIRE FREE WORLD

102 C: Spectrum Demand in Wire Free World C.1 DEMAND AND SUPPLY OF SPECTRUM The Wire Free World scenario is based on the premise that the current growth rate in technological development and new applications will continue developing in the same manner as they are today but at an accelerated rate. There will also be a significant shift from wired/fixed location demand to mobile wire-free demand. Services, demand, applications and technologies continue to develop along the lines that they have been developing for the last years, with the pace accelerating. C.1.1 Demand Demand for data applications is likely to see a significant increase in this scenario, potentially 60% per annum, with a shift in the traffic mix towards data. The shape of the demand growth would be exponential, ostensibly putting a higher risk of spectrum shortage occurring within the planning period (i.e. 2025). As with Business As Usual, demand growth will be driven by customer demand and not through any increase in the feedback from network changes. There will be a faster shift in broadcasting service towards HDTV and some shift to satellite. a. Demand (traffic) Total Demand growth per annum (amount in MB equivalent) Total Demand growth (mix) Demand growth (driven by growth in technology) High Shift To Data Some C.1.2 Network In cellular, macro-cell network architecture will be complemented by a femto layer taking a considerable share of the traffic. b. Network architecture Cell proliferation Fibre plus wireless Industry fragmentation Convergence and integration of services Moderate High Moderate Moderate Cellular traffic carried by femtocells 40% C.1.3 Availability of Spectrum In this scenario, the rate of release and re-use of spectrum is somewhat higher than in Business As Usual. c. Availability of Spectrum Rate of release/re-use of spectrum Spectrum coordination across regions High High C-1

103 C: Spectrum Demand in Wire Free World C.2 WIRE FREE WORLD OFFERED TRAFFIC Figure 51 shows offered traffic in TB in the busy hour (uplink and downlink combined) for Wire Free World. By 2025 Traffic is about four times that in Business As Usual, growing on average some 10% more per year. Figure 51: Total Offered Traffic (Terabytes) in Wire Free World by Service (log scale) Figure 52 shows offered total traffic by Application group in the busy hour (uplink and downlink combined) for Wire Free World. Telemetry and M2M account for a significantly greater proportion of the busy hour traffic. Figure 52: Total Offered Traffic (TeraBytes) in Wire Free World by Application Group (log scale) C-2

104 C: Spectrum Demand in Wire Free World C.2.1 Wire Free World - Demand for Spectrum The increased demand for all services means that the spectrum required is significantly more than in Scenario 1. Spectrum demand reaches its maximum in 2025, but it also peaks in 2011/12 at around 800MHz. The main changes, moving from Scenario 1 to this scenario, are as follows: In Cellular, the demand for spectrum post 2013 is some 100MHz higher. In Broadcast, there is more demand, but more of it diverts to satellite, so the net effect on demand below 10GHz is minimal. In SRW, up to 2013 the demand for spectrum is comparable, but thereafter it continues to grow much more rapidly, ending at 600MHz in this scenario, contrasting against 200MHz in Scenario 1. Figure 53: Spectrum Demand (MHz) in Wire Free World by Service C-3

105 APPENDIX D: THE ALL YOU COULD WANT SCENARIO

106 D: The All you Could Want Scenario D.1 DEMAND AND SUPPLY OF SPECTRUM The All You Could Want scenario is driven by consumer demands for applications and improved devices. This is indicative of greatly accelerated growth in individualised interactive entertainment and broadcast services delivered to wireless users as the norm. There is a significant shift from wired to wire-free with services providing entertainment on the move. Services, demand, applications and technologies develop faster than in the previous decade(s), as consumers, vendors, and operator s get into something of a positive feedback loop. D.1.1 Demand This scenario will result in the largest increase in demand for data traffic, up to 80% per annum, for personalised data with much greater specialisation and distribution of sources and destinations. Demand growth for data is more or less exponential, as in Wire Free World. Demand growth for Broadcasting, and in its levels of video resolution, is slightly faster than in Wire Free World. a. Demand (traffic) Total Demand growth per annum (amount in MB equivalent) Total Demand growth (mix) Demand growth (driven by growth in technology) Very High Large Shift To Data High D.1.2 Network In cellular, a femtocell layer will complement the existing macro/microcell layer. For fixed network the last mile will be increasingly based on fibre which will reduce upstream bottlenecks. Convergence and integration of services and applications will be faster than Wire Free World. b. Network architecture Cell proliferation Fibre plus wireless Industry fragmentation Convergence and integration of services Very High Very High Moderate Moderate Cellular traffic carried by femtocells 60% D.1.3 Availability of spectrum In this scenario, the rate of release and re-use of spectrum is higher than in Business As Usual, partially offsetting the greater rates of growth in traffic, with the increases in total spectrum available 100MHz-10GHz of the order of a 50% total increase over the year period. c. Availability of Spectrum Rate of release/re-use of spectrum Spectrum coordination across regions High High D-1

107 D: The All you Could Want Scenario D.2 ALL YOU COULD WANT - RESULTS D.2.1 All You Could Want - Offered traffic (by Service) Figure 54 shows offered traffic in MB in the busy hour (uplink and downlink combined) for All You Could Want Traffic is about times that in Business As Usual, growing an additional 15-20% per annum. Figure 54: Total Offered Traffic (Terabytes) in All You Could Want by Service (log scale) D.2.2 All You Could Want - Offered traffic (by Application Group) Figure 55 shows offered total traffic by Application group in the busy hour (uplink and downlink combined) for All You Could Want. Figure 55: Total Offered Traffic (TeraBytes) in All You Could Want by Application Group (log scale) D-2

108 D: The All you Could Want Scenario D.2.3 All You Could Want - Demand for Spectrum Figure 56: Spectrum Demand (MHz) in All You Could Want' by Service It is notable that demand for spectrum does not increase by the same factor as traffic, especially for cellular, compared with Business As Usual. The reason for this is the assumed higher density of new cell sites, including extensive use of femtocells. The main changes, moving from Scenario 1 to this scenario, are as follows. In Cellular, the demand for spectrum post 2013 is some 100MHz higher, and grows by a further 200MHz post In Broadcast, there is more demand, but more of it diverts to satellite, so the net effect on demand below 10GHz is minimal. In SRW, up to 2012 the demand for spectrum is comparable, but thereafter, it continues to grow much more rapidly, ending at 1800MHz in this scenario, contrasting against 200MHz in Scenario 1. D-3

109 APPENDIX E: SCENARIOS PAPER

110 E: Scenarios Paper TABLE OF CONTENTS E1. Introduction E1-1 E2. The World As We See It Developing Between Now and 2025 E2-1 E2.1 Other applications by 2025 E2-3 E2.2 Other services by 2025 E2-4 E3. Defining the Future Scenarios E3-1 E3.1 Identifying spectrum usage drivers until 2025 E3-1 E3.2 Defining future scenarios until 2025 E3-4 E3.3 The dimensions of the scenarios E3-5 E4. The Future Scenarios E4-1 E4.1 Classification of scenarios E4-1 E4.2 Scenario 1: The Business As Usual scenario E4-2 E4.3 Scenario 2: The Wire-Free scenario E4-3 E4.4 Scenario 3: The All You Could Want scenario E4-5 E4.5 Scenario 4: The Dystopian scenario E4-8 E4.6 Scenario 5: The Industry Fragmentation scenario E4-9 E4.7 Scenario 6: The Re-Use scenario E4-10 E5. Summary E5-1 EZM R_A Ofcom 7 April 2009 E-i

111 E: Scenarios Paper E1. INTRODUCTION This document defines a set of possible future scenarios for spectrum usage over the next years which will be modelled in order to recognise potential areas of spectrum shortage. All the future scenarios presented in this document have been developed in conjunction with the Stakeholder Panel. The following sections of the document are structured as follows: Section E2 A description of the what the World in 2025 may look like in terms of applications and technologies Section E3 The process employed to identify and define the set of future scenarios Section E4 A description of the different future scenarios that have been identified and their expected impact on spectrum usage. EZM R_A Ofcom 7 April 2009 E1-1

112 E: Scenarios Paper E2. THE WORLD AS WE SEE IT DEVELOPING BETWEEN NOW AND 2025 The purpose of the future scenarios is to define a set of different likely future worlds. In the model, these are represented in terms of specific applications and technology developments associated with mobile, broadcast and fixed wireless services as shown in Table 1 (see below). EZM R_A Ofcom 7 April 2009 E2-1

113 E: Scenarios Paper Application Service Technology Voice Messaging (SMS, MMS, IM) E Mail WWW Video Streaming TV Radio PMSE Gaming Download P2P M2M and Telemetry Frequency Simple Voice Rich Voice Voice over IP Voice Messages Video Conference SMS MMS Instant Messaging Video Messages Internet Access (e.g. VPN) Web Browsing Direct Video Streaming Broadcast TV Broadcast channels Radio stations Broadcast PMSE Online Gaming Mobile Gaming Gambling Lotteries Music/Video File Transfer E-Commerce M-Commerce File Sharing U-Tube Blogs Chat Rooms Navigation Tracking Push/Pull Advertising Cellular 2G GSM N/A x x x x Possible but slow 2.5G GPRS/ EDGE N/A x x N/A N/A 3G UMTS N/A N/A 3.5G (HSPA) N/A 4G LTE N/A 900MHz 1800MHz 900MHz 1800MHz 900MHz 1800MHz 2100MHz 900MHz 1800MHz 2100MHz 700MHz 900MHz 1800MHz 2600MHz Paired Paired Paired Paired Paired & Unpaired Mobile WiMax N/A 700MHz 900MHz 1800MHz 2600MHz Unpaired MBMS x x Paired x x x x x x x 2100MHz (Unpaired) Wireless Analogue Cordless** x x x x x x x x x x x 47MHz Paired (Short Range) DECT** x N/A N/A x N/A N/A N/A N/A N/A N/A N/A 1880MHz Paired Bluetooth N/A x N/A N/A N/A 2.4GHz Unpaired UWB N/A N/A GHz Unpaired WiFi N/A 2.4GHz 5.8GHz Unpaired Analogue N/A N/A 0.2MHz MHz 2-30MHz 88- Unpaired Radio 108MHz DAB x x MHz Unpaired DRM x x 2-30MHz Unpaired Analogue Not in isolation x N/A Not in isolation MHz Unpaired DVB-T x MHz Unpaired TV DVB-H x N/A MHz Unpaired DVB-T2 x N/A MHz Unpaired DVB-S x N/A ~10GHz Unpaired Fixed WiMax GHz Unpaired Fixed Wireless 3.4GHz TD/CDMA Acc. 1920MHz Unpaired Proprietary 3.4GHz Unpaired Back-haul Microwave >10GHz Paired = application available via technology; N/A = technology potentially capable, but not used to provide application. x = technology not capable of providing application; ** not currently confirmed as in scope, will be kept under review Table E1: Application and service family matrix EZM R_A Ofcom 7 April 2009 E2-2

114 E: Scenarios Paper To model the demand for spectrum over the next years it is important to understand what the future World could look like, in terms of demand for applications, services and spectrum. These will be modelled to investigate whether the current allocation of spectrum is likely to be able to support them or if there is a discontinuity. Therefore, in addition to the applications and services shown in Table 1, it is recognised that there will be "others, which have not yet been envisaged but that will be in existence and generating significant traffic by In order to understand what these other applications and services might be, we held a series of meetings with the stakeholders. The details of these sessions can be found in Appendix A. The following is a summary of the key ideas postulated. E2.1 OTHER APPLICATIONS BY 2025 Mobile Entertainment: By 2025 it is expected that 3D high resolution video glasses will be available. These will enable the user to have the ability to view, hear, or interact with entertainment media wherever or whenever desired. In addition, the user will have the ability to personalise the media. Mobile Broadcasting: By 2025 it is proposed that personal wearable devices will be available. These devices will broadcast personal information and will be able to interact in real time with other devices when in close proximity. The radiating devices could be used to advertise products and service or even make recommendations. Therefore the level of trust that is needed in the information being broadcast is high and would require regulation of the business layer platform to stop abuse and provide fair competition. Applications of this type also raise the issue of digital trespassing and how this can this be prevented is something that needs to be resolved. Digital Media: In 2025 we may see the emergence of digital paper this will be as a result of the convergence between TV, radio, paper, magazines and computer games. However this will be dependent on the corresponding convergence of regulations of these sources. Digital Jewellery: By 2025 mobile phones could be replaced with digital jewellery, which will contain the same functionality as mobile phones (e.g. camera, radio, telephone, etc) but just dismantled into the individual components, and sold separately as individual low cost jewellery items. This will be possible because of the increased miniaturisation of chip sets and batteries through nanotechnology and new materials. The complication will be around the user interface which will also rely on new technology being developed. Active Skin: By 2025 active skin electronics could be in use to monitor a person's medical state. This involves a chip, the size of a grain or smaller, being embedded under the skin. The chip is able to communicate in real time with a central system allowing for real time remote monitoring of medical conditions such as diabetes and then administering the necessary drugs in real tine with no user intervention. EZM R_A Ofcom 7 April 2009 E2-3

115 E: Scenarios Paper E2.2 OTHER SERVICES BY 2025 Improvements in semiconductor and nanotechnology over the next years will make it possible to develop smaller and smaller chip sets with increased functionality e.g. processing power and storage capacity will increase, and user interfaces such as speech recognition will become common place. In the next years it is predicted that the mobile and fixed communications networks combine to create a fully cordless world with extremely high data rates which will be available to everyone at any location. Additionally, there will be the emergence of new wireless networks, which will provide short range hop to a fixed networks and will result in service providers having to offer new bundled services to the consumer such as free services combined with paid for services. Mobile technology is expected to migrate to a single network protocol, namely, IP and this will help promote competition between operators. Optical wireless (pt-pt) will be available in 2020 offering both directional and non directional short range backhaul. EZM R_A Ofcom 7 April 2009 E2-4

116 E: Scenarios Paper E3. DEFINING THE FUTURE SCENARIOS To provide reliable predictions of spectrum usage over the next years a set of realistic future scenarios must be defined. This has been achieved, in this study, using the following 3 step approach (Figure E1): Step 1 - Identify key drivers that are predicted to affect spectrum usage over the next years Step 2 - Group the drivers in the most appropriate manner to define a set of possible future scenarios Step 3 - Define each of the future scenarios to be modelled in term of their demand for applications, services and spectrum. Identify possible driver of spectrum usage in the next years Group drivers and associated uses to define future scenarios Evaluate each future scenario in terms of demand for applications, services and spectrum Drivers Scenario 1 Drivers Scenario 2 Drivers Scenario 3 Drivers Drivers Scenario 4 Drivers Application Demand Service Demand Spectrum Demand Application Demand Service Demand Application Demand Service Demand Spectrum Demand Spectrum Demand Application Demand Service Demand Spectrum Demand Figure E1: Development of the future scenarios Full details of this approach can be found in Appendix A. The following sections however provide an overview and the conclusions. E3.1 IDENTIFYING SPECTRUM USAGE DRIVERS UNTIL 2025 The drivers that will affect the future use of spectrum were identified by considering the potential growth and developments of applications and services over he next years combined with an understanding of what demands these requirements these will place on the current spectrum allocation. The process identified six key drivers as playing an important role in the usage of spectrum between 2008 and Continuous Steady State Development 2. Technology Development a. Supply of Bandwidth b. Demand of Bandwidth EZM R_A Ofcom 7 April 2009 E3-1

117 E: Scenarios Paper 3. Social and Cultural Trends 4. Macro-Economic Changes 5. Provider Business Model(s) 6. Alternatives to Wireless Spectrum 1. The key uses associated with each of the drivers (expressed in terms of applications and technology developments) expected to affect the demand for spectrum over the next years are presented in Table E2. 1 Out of scope for this study. EZM R_A Ofcom 7 April 2009 E3-2

118 E: Scenarios Paper Continuous Steady State Development Technology Development (Supply of Bandwidth) Technology Development (Demand of Bandwidth) Social & Cultural Trends Macro- Economic Changes New Business Model (s) Alternatives to wireless spectrum Interactive gaming Real time, efficient network optimisation 3D Printers Data security De-regulation Spectrum trading/ sharing Very high performance copper Personalised advertising Cognitive wireless networks High Definition TV Demise of fixed office Digital switch over High quality voice and video Low cost satellite Software bloat Ubiquitous seamless networks Device miniaturisation Environmental costs and planning Military spectrum requirements Price and price elasticity Optical backhaul On demand Short range networks Electronic paper Radiation concerns Big Brother applications All you can eat tariff Use of fibre Junk/spam/ spit calls, messages, videos, etc Ad-hoc or MESH networks Compression technology Generation change Monopoly provision Device push vs. network push High definition wireless CCTV High altitude provision 3D TV Change in population distribution Inefficient spectrum allocation Monolithic vs. aggregator Data verification Intelligent vehicles and highways Wire-free access Competition/o pen access Market opens up to new providers Device capability Device proliferation Wireless office Wider economic growth Industry consolidation Personal Video Recording M2M applications Social networking Advertising payload Network agnostic applications Wireless home Virtual worlds Diagnostic and monitoring systems Immersive TV Demand for > 50 channels Personalisation Ultra mobile PC's User generated content Table E2: Key drivers for spectrum until 2025 EZM R_A Ofcom 7 April 2009 E3-3

119 E: Scenarios Paper E3.2 DEFINING FUTURE SCENARIOS UNTIL 2025 From analysis of the key future drivers for spectrum and their associated uses over the next years we have derived a set of six future scenarios: Business As Usual - The growth trend and technology developments continue the trends of the last years. There is some shift from wired to mobile wirefree, but this is a gradual trend rather than a sudden change. The detailed assumptions for Business As Usual are included in Section E4.2 and Section 5 of the main report. Wire-Free World - This scenario is centred on faster growth in wireless technology and demand for services. Wide applicability and seamless convergence of fixed and mobile communication networks create a fully Wire- Free world with extremely high data rates, available at practically all locations. Users will have wireless access anytime anywhere, breaking down the some of the traditional barriers between being at home, at work, and elsewhere. All You Could Want - There is greatly accelerated growth and a significant shift from wired to wire-free. Services, demand, applications and technologies develop along the lines that they have been developing for the last years but with the pace accelerating and an upsurge in new applications using device to-device and device-to/from-person communication. Dystopia - Society is more threatened and fearful, due to a bleaker social and economic outlook, driven by e.g. accelerating prices for raw materials, an upsurge in terrorist threats and activities, identity theft, and severe reining-in of credit and expenditure, throttling back economic growth. The mix of technology and demand shifts away from being diverse and consumer-centric, towards a focus on systems aimed at monitoring, recording and coordinating everything, aimed at reducing terrorism / crime / disorder / antisocial activity. Total demand continues to grow, but slower than in Business As Usual. Industry Fragmentation The growth in capacity does not keep up with the demand, as the degree of cooperation and standardisation across geographies and different parts of the supply chain deteriorates, reversing some of the gains in economies of scale achieved in the last years. Industry starts to fragment, from having major players and interoperability, to a profusion of smaller solutions and ventures fighting for spectrum, content, services, and revenue. National governments are more inclined to step in with special measures to protect the interests of local ventures. Other scenarios broadly cover the opposite case, where we assume that there are strong efficiencies of scale. Re-use There will be a significant shift in emphasis toward reuse and sharing possibly reflecting a similar shift in values in the wider society and economy. This will be reflected in (i) greater cooperation between networks, for example tolerance or encouragement for sharing of sites and/or spectrum; and (ii) greater releases of spectrum suitable for re-use in mobile and fixed networks possibly driven by introduction of AIP to public sector usage both in the UK and in around the world. Figure E2 below compares the six different future scenarios in term of their supply and demand for capacity. EZM R_A Ofcom 7 April 2009 E3-4

120 E: Scenarios Paper Supply of Capacity Re-Use All You Could Want Wire-free Business as Usual Demand for Capacity Dystopia Industry Fragmentation Figure E2: A comparison of the different future scenarios Note that in the diagram above, Demand for capacity includes latent (i.e. unmet) demand. Supply of capacity will be affected not only by additional spectrum availability, but also by techniques that increase efficiency (e.g. improved video coding technology). E3.3 THE DIMENSIONS OF THE SCENARIOS The six future scenarios are used to evaluate the spectrum required for various futures over the next years. Three dimensions have been identified to describe scenarios for future demand in terms of applications, services and spectrum; these are: Demand (Traffic) Network architecture (Demand for spectrum) Availability of spectrum in steady state condition. The details of the individual dimensions with regard to the scenarios are defined in the following sub-sections. EZM R_A Ofcom 7 April 2009 E3-5

121 E: Scenarios Paper a. DEMAND (TRAFFIC) The Business As Usual scenario contains the baseline forecasts for traffic based on the growth of demand increasing as currently predicted. The growth in demand relative to this base case for each future scenario will be affected in four dimensions: The total amount of growth in demand by application in MB The mix of growth in demand by application (including future unknown applications) The timing of growth in demand by application across the planning period The impact of growth (and improvements) in technology on the growth in demand. b. NETWORK ARCHITECTURE/DEMAND FOR SPECTRUM In addition to changing demand growth, scenarios are also defined by potential changes in network structure and architecture. Key dimensions identified in this respect are: A move from the current situation of a predominantly macro cell structure to one that is based predominantly around smaller, more numerous cells (e.g. Femto) A move from the current situation of a predominantly copper (e.g. xdsl) last mile delivery to one that is based predominantly on fibre plus wireless A move towards a fragmented and un-coordinated industry (e.g. lots of small players, weak or absent standards) Convergence and integration of services. c. AVAILABILITY OF SPECTRUM Further, the changing availability of spectrum both nationally and internationally will vary according to scenario. The key dimensions identified include: The rate of release/re-use of spectrum The impact of international standardisation that promotes and drives re-use and/or release of spectrum. EZM R_A Ofcom 7 April 2009 E3-6

122 E: Scenarios Paper E4. THE FUTURE SCENARIOS In section 3 we identified the six future scenarios to be modelled in this study as: 1. Business As Usual World 2. Wire-Free World 3. All You Could Want World 4. Dystopian World 5. Industry Fragmentation World 6. Re-Use World. Details of each of these future scenarios are discussed in the following sections. E4.1 CLASSIFICATION OF SCENARIOS The three dimensions discussed in section E3.3 enable each scenario to be classified in a way that establishes the likely risk of spectrum shortage over the next years. The comparison of the dimensions of each of the scenarios is shown in Table 3. EZM R_A Ofcom 7 April 2009 E4-1

123 E: Scenarios Paper Business As Usual scenario Wire-Free World scenario All You Could Want scenario Dystopian scenario Industry Fragmentation scenario Re-Use scenario a. Demand (traffic) Total Demand growth per annum (amount in MB equivalent) Total Demand growth (mix) Demand growth (driven by growth in technology) b. Network architecture Moderate High Very High Slow Moderate Moderate Shift To Data Shift To Data Large Shift To Data Large Shift To Data Shift To Data Shift To Data Some Some High Some None None Cell proliferation Slow Moderate Very High Moderate High Moderate Fibre plus wireless Slow High Very High Slow Slow High Industry fragmentation Some Moderate Moderate Some Very High None Convergence and integration of services Cellular traffic carried by femtocells c. Availability of Spectrum Rate of release/re-use of spectrum Spectrum coordination across regions Slow Moderate Moderate None Negative Moderate 20% 40% 60% Minimal Minimal 20% Moderate High High Slow Moderate Very High Moderate High High Slow Negative High Table E3: A comparison of the set of future scenarios Each of the dimensions has a different outturn in each of the six scenarios, and their relationships are summarised in Table E3 in terms of the categorisations Very High to Negative. Further details are provided in the following sub-sections. E4.2 SCENARIO 1: THE BUSINESS AS USUAL SCENARIO The Business As Usual scenario can be best described as the as is world. It assumes moderate growth (projecting the growth trend from the last 5-10 years, no great acceleration) and is based on present business strategy. Services, demand, applications and technologies develop along the same lines as in the last years and at the about same pace. There is a bit of a shift from wired/fixed location demand, to mobile wire-free, but this is a gradual trend rather than a sudden change. EZM R_A Ofcom 7 April 2009 E4-2

124 E: Scenarios Paper E4.2.1 Dimensions a. DEMAND The growth in demand in this scenario is assumed to be approximately 40% per annum in total MB with a shift from voice to data. This growth in traffic is assumed to be exponential. It is expected that there will be no coupling in growth to technology/bandwidth/speed. There will be a shift in broadcasting service towards HDTV and some shift to satellite. b. NETWORK ARCHITECTURE There will be very little change if any in the current industry arrangements and key players although we can expect to see some fibre to the kerb/home/office replacing copper/xdsl as the last mile dominant technology although this will be slow. This will be coupled with the slow introduction of Femto/Micro cells. There will be a very little convergence and integration of broadcast communication and cellular wireless-fixed in this time frame. c. AVAILABILITY OF SPECTRUM There will be a moderated rate of release and /or re-use of spectrum standards will remain as they do today with very little change promote and drive re-use/release of lots of spectrum, standardised across international regions E4.3 SCENARIO 2: THE WIRE-FREE SCENARIO The Wire-Free world scenario is based on the premise that the current growth rate in technological development and new applications will continue developing in the same manner as they are today but at an accelerated rate. There will also be a significant shift from wired/fixed location demand to mobile wire-free demand. Services, demand, applications and technologies continue to develop along the lines that they have been developing for the last years, with the pace accelerating. The drivers that have the greatest impact on this scenario are advances in Technology that will change the usage of spectrum through the demand of bandwidth and Social and Cultural Trends. As a result of these drivers, the key applications and services that will have the most significant growth and therefore result in the greatest growth in demand for this scenario are presented in Table E4. EZM R_A Ofcom 7 April 2009 E4-3

125 E: Scenarios Paper Driver Effect Service Application Technology (i) Ad-hoc or MESH networks Mobile All (i) (ii) Short range networks (Femto & mega-wifi) Mobile All Device proliferation (everything has IP) Mobile All (ii) M2M applications Mobile M2M (ii) Wireless home applications Mobile All Social and Cultural Trends E4.3.1 Dimensions a. DEMAND Change in population distribution All All Working from home (demise of fixed office) Mobile Social networking Mobile P2P Wire free access Mobile All Wireless office applications Mobile Table E4: Key drivers in a Wire-Free world Voice, , WWW, Downloads Voice, , WWW, Downloads, Video streams Demand for data applications is likely to see a significant increase in this scenario, potentially 60% per annum in total amount of MB, with a shift towards data. The shape of the demand curve in this case would be exponential, putting a higher risk of spectrum shortage occurring within the planning period (i.e. 2025). In this case, demand growth will be driven by customer demand and not through feedback from network changes. There will be a shift in broadcasting service towards HDTV and some shift to satellite b. NETWORK As currently, macro-cell network architecture will dominate in this scenario, placing a moderate demand on spectrum. The last mile however will be based on an xdsl strategy which does not impose an increased demand on spectrum. It is likely that the use of spectrum by providers becomes more efficient via developments in convergence and integration of services and applications over the planning period that would indicate a low risk of impact on spectrum shortage. c. AVAILABILITY OF SPECTRUM In this scenario, the rate of release and re-use of spectrum is expected to be moderate, placing a greater demand on spectrum available than otherwise expected given the significant increase in customer demand for applications. The international management of spectrum would probably follow similar trends initially but will show acceleration from 2012/2015. EZM R_A Ofcom 7 April 2009 E4-4

126 E: Scenarios Paper E4.4 SCENARIO 3: THE ALL YOU COULD WANT SCENARIO The All You Can Want scenario is driven by consumer demands for applications and improved smaller devices. This is indicative of greatly accelerated growth in individualised interactive entertainment and broadcast services delivered to wireless users as the norm. There is a significant shift from wired to wire-free with services being delivered to people, not places, providing entertainment on the move. Services, demand, applications and technologies develop faster than in the previous decade(s), as consumers, vendors, and operator s get into something of a positive feedback loop. This scenario emanates from the Provider Business Model(s) with other relevant drivers from Continuous Steady State Development and Technology. As a result, the key applications and services that have the most significant consequence and therefore result in the greatest growth in demand for this scenario are presented in Table 5 below. EZM R_A Ofcom 7 April 2009 E4-5

127 E: Scenarios Paper Driver Effect Service Application Continuous Development Technology (ii) (ii) Demand for > 57 terrestrial channels Broadcast TV Demand for immersive TV Broadcast TV Participation in interactive gaming Mobile Gaming Unrequested junk/spam/spit calls and videos Massive personal video recording (PVR) Broadcast TV All Messaging, , WWW, Video streaming On demand broadcasts All TV, Video streaming Personalisation All M2M, P2P Personalised advertising All P2P Software bloat across applications Mobile Messaging, WWW, Ultra mobile PC's Mobile All User generated content All P2P Demand for virtual worlds Mobile P2P Improvements in compression technology Broadcast Video Streaming, TV, Radio General use of diagnostic and monitoring medical systems Mobile M2M (ii) Demand for High Definition TV Broadcast TV (ii) Provider Business Models E4.4.1 Dimensions a. DEMAND Use of intelligent vehicles and highways Mobile M2M All you can eat tariff All All High quality voice and video Mobile Table E5: Key drivers in an All You Could Want world Voice, Messaging, Video Streaming This scenario will result in a vast increase in spectrum due to the large demand, potentially up to 80% per annum in total amount of MB, for personalised data with much greater specialisation and distribution of sources and destinations. This will be assisted by the increase in device. The shape of the demand curve in this case would be exponential, putting a high risk of spectrum shortage occurring within the planning period In this case, traffic growth will be driven by customer demand but this will be coupled with feedback from network changes. Additionally, we will expect there to be a move in broadcasting to IP platforms EZM R_A Ofcom 7 April 2009 E4-6

128 E: Scenarios Paper b. NETWORK A Femto/Micro cell network architecture will begin to dominate in non- broadcast services, providing greater re-use of spectrum. For fixed network the last mile will be increasingly based on fibre which will reduce upstream bottlenecks. Convergence and integration of services and applications will be faster than Business as Usual. c. AVAILABILITY OF SPECTRUM In this scenario, the rate of release and re-use of spectrum is expected to be moderate, placing a greater demand on spectrum available than otherwise expected given the significant increase in customer demand for applications. International coordination of the management of spectrum would probably follow similar trends initially but will increase from 2012/2015. EZM R_A Ofcom 7 April 2009 E4-7

129 E: Scenarios Paper E4.5 SCENARIO 4: THE DYSTOPIAN SCENARIO The Dystopian scenario is where society becomes more threatened and more fearful, due to a bleaker social and economic outlook, driven by (for example) accelerating prices for raw materials, an upsurge in international terrorist threats and activities, increased identity theft, and severe reining-in of credit and expenditure, throttling back the rate of economic growth. The mix of technology and demand shifts away from being diverse and consumer-centric, towards a narrower focus on systems aimed at monitoring and recording and coordinating everything, aimed at reducing terrorism/ crime/ disorder/ antisocial activity. Individuals no longer trust, and therefore want to use, technology. The Dystopian World scenario arises from a combination of the Macro Economic and Social and Cultural Trends drivers with the former being the more dominant driver. The key applications and services that result in the greatest growth in demand for this scenario are presented in Table E6. Driver Effect Service Application Macro- Economic Social and Cultural Trends E4.5.1 Dimensions a. DEMAND Roll out of big brother applications Mobile M2M, P2P Digital switch over Broadcast TV, Radio Release of military spectrum requirements for other uses Mobile PSME Demand/provision of data security Mobile All Table E6: Key drivers in the Dystopian world This will see large centralised it systems, with huge automations and huge numbers of small transactions. Emergent behaviour will see users take up services because it makes things easier or cheaper (e.g. Lower health insurance if you have your health monitored). Therefore, growth in demand in this scenario is assumed to be moderate, with a shift in the mix from voice to data such as M2M and telematics. This growth in data demand is assumed to follow a more or less exponential trajectory, possibly slowing down by Additionally, in the broadcasting sector there will be a cull of smaller terrestrial channels to allow room for a few (4-10) major high definition TV channels. b. NETWORK ARCHITECTURE As currently, macro-cell network architecture will dominate, placing a moderate demand on spectrum. It is likely that spectrum use becomes more efficient but there will be very little developments in convergence and integration of services and applications over the planning period that would indicate a low risk of impact on spectrum shortage. c. AVAILABILITY OF SPECTRUM There will be little release of spectrum. EZM R_A Ofcom 7 April 2009 E4-8

130 E: Scenarios Paper E4.6 SCENARIO 5: THE INDUSTRY FRAGMENTATION SCENARIO The Industry Fragmentation world scenario is where after a few years of business as usual, the industry starts to fragment; from the four or five major players in each service sector, to a profusion of smaller ventures fighting for spectrum /content /services. The current co-operation between different parts of the supply chain, and between regions and countries, is reduced and government act to protect vested interests. This scenario is driven by the Provider Business Model. Therefore, the key applications and services that result from this driver and its associated effects with the most significant growth in demand are highlighted below in Table E7. Driver Effect Service Application Provider Business Models E4.6.1 Dimensions a. DEMAND Advertising payload All Messaging, WWW, Market opens up to new providers. Mobile All High quality voice and video Mobile Industry consolidation Mobile All Spectrum trading/sharing All All Table E7: Key drivers in an Industry Fragmentation world Voice, messaging, video streaming The growth in demand in this scenario is assumed to be approximately 40% per annum in total amount of MB with a shift in the mix from voice to data. This growth in demand is assumed to be exponential, but with little or no coupling of demand growth to technology/bandwidth/speed. b. NETWORK ARCHITECTURE As currently, macro-cell network architecture will dominate but with the increase introduction of Femto/Micro cells, placing a moderate to high demand on spectrum. The last mile however will be based on an xdsl strategy which does not impose an increased demand on spectrum. It is likely that the use of spectrum by providers deteriorates and there will be no developments in convergence and integration of services and applications over the planning period indicating a low risk of impact on spectrum shortage. c. AVAILABILITY OF SPECTRUM There will be a moderated rate of release and /or re-use of spectrum standards will deteriorates even within the EU compared with today as there will be no promotion or drive to re-use/release of lots of spectrum. EZM R_A Ofcom 7 April 2009 E4-9

131 E: Scenarios Paper E4.7 SCENARIO 6: THE RE-USE SCENARIO The Re-Use world scenario consists of a significant release of suitable spectrum for reuse in mobile and fixed networks, both in the UK and in the other major markets around the world. Consequently the supply of capacity develops at a faster pace than the Business as Usual scenario. In locales where spectrum may be scarce, operators are able to mitigate the impact by cooperating over spectrum sharing and re-use. E4.7.1 Dimensions a. DEMAND Services, demand, applications and technologies develop along the lines that they have been developing for the last years approximately 30% per annum in total amount of MB, There is a significant shift from wired to wire-free, with a shift in the mix of demand from voice to data. Demand growth follows an exponential trajectory, not coupled to advances in technology or growth in capacity. b. NETWORK There is a proliferation of short range wireless capacity and an associated growth in fibre deployments. Industry coordination remains strong, with some use of spectrum sharing to mitigate what would otherwise be pockets of spectrum shortage. c. AVAILABILITY OF SPECTRUM The rate of release and re-use of spectrum is expected to be very high, due to release or sharing of spectrum currently held within the public sector (in both the UK and elsewhere). Assignments of spectrum are coordinated across different countries and regions, to maximise the returns to scale. EZM R_A Ofcom 7 April 2009 E4-10

132 E: Scenarios Paper E5. SUMMARY In order to investigate how developments in mobile, broadcast, and fixed wireless communication services, combined with growth in demand for, new and existing, applications will affect the usage of spectrum in the UK over the next 15 to 20 years six future scenarios have been identified. These six future scenarios have been identified through a series of workshops and meetings with PA colleagues and members of the stakeholder panel. This document has identified six possible future scenarios for spectrum usage over the next years. In addition it also highlights how they will be modelled to recognise potential areas of spectrum shortage. The six future scenarios are: 1. Business As Usual World 2. Wire-Free World 3. All You Could Want World 4. Dystopian World 5. Industry Fragmentation World 6. Re-use World These six scenarios are represented in the model in term of three dimensions - Demand (Traffic), Network architecture (Demand for spectrum) and Availability of spectrum in steady state condition. Theses three dimensions enable each scenario to be classified in a way that establishes the likely risk of spectrum shortage over the next years. The table in section 4.1 provides a comparative overview of how each of the six scenarios will be represented in the model in terms these three dimensions. The model will evaluate each of these six future scenarios against the current demand in spectrum across the planning period, providing Ofcom with an understanding of which future scenario(s) could result in a shortage of spectrum and in what timeframe it is likely to occur. The results of this study will provide Ofcom with insight into the issues and uncertainties affecting spectrum usage over the next years and enable them to preempt decisions using policy levers which are wholly or partly within their control in order to manage the future. EZM R_A Ofcom 7 April 2009 E5-1

133 APPENDIX F: STAKEHOLDER MEETING NOTES EZM R_A Ofcom 7 April 2009

134 F: Stakeholder Meeting Notes TABLE OF CONTENTS F1. Notes from meeting with Professor Simon Saunders F1-1 F2. Notes from meeting with Graham Plumb - Head of Distribution, BBC F2-1 F2.1 Current areas of spectrum shortage for broadcasting F2-1 F2.2 Spectral efficiency F2-1 F2.3 What the future of broadcasting will look like F2-1 F3. Notes from Meeting with Ian Pearson - Futurologist, Futurizon GmbH F3-1 F3.1 The most likely scenario is a major shift to a wire-free world F3-1 F3.2 Network protocols will converge on IP - and charges for voice calls may become a thing of the past F3-1 F3.3 Devices will miniaturise and may proliferate ( Digital Jewellery ) F3-1 F3.4 Technology will continue to get easier to use, and content will proliferate F3-2 F3.5 Technology, in particular wirefree, will have a dramatic impact on the transport sector F3-3 F3.6 Micro-electronic devices may be deployed on - and inside - the body F3-3 F3.7 These changes may occur even if the socio-economic outturn is poor F3-3 EZM R_A Ofcom 7 April 2009 Fi

135 F: Stakeholder Meeting Notes F1. NOTES FROM MEETING WITH PROFESSOR SIMON SAUNDERS Notes from meeting with Professor Simon Saunders - Independent Consultant, Chair of the Femto Forum, Visiting Professor at the University of Surrey, and a Member of Ofcom Spectrum Advisory Board(OSAB). F.1.1 Modelling the trend for growth in Demand The demand for communications is currently limited by: Availability Price Speed. It does not seem to have any extrinsic limits ("consumers want n MB /mo"). As traffic grows - backhaul may become more constrained. F.1.2 Services and technology feedback to affect growth in demand Services, technology, spectrum, and demand are a coupled system. Growth in services/speeds/capacity does stimulate growth in demand. Ideally the model and report will reflect this Better devices/services/prices/data rates will lead to more usage. Continual tension between "build it and they will come" vs. getting funding and investment/business models that are a priori viable. F.1.3 Distribution of demand needs to be considered explicitly Most usage is sedentary, in-building followed by localised peaks (train, airport and highway). There is little true outdoor fully mobile usage in low density areas. Airports operators need/desire to capture roaming customers - networks therefore have an incentive provide excess capacity/speed in these areas. Airports [and other sites] - landlords/freeholders are in strong position to extract rents from network operators/wifi providers the site is at a premium (more so than the service or spectrum). For spectrum shortage we are as interested in peaks, exceptional events (e.g. 7/, political/military crises) and localised high demand as we are in the average, or in 90-99% of usage. Averages alone are dangerous/misleading, and urban/rural geotypes alone aren t really enough. F.1.4 Factors affecting future scenarios Some of the main factors on which the alternative futures vary: Lots of feedback from technology/capacity growth, to services, and to end-user demand and traffic or little or no capacity-inspired growth. SS believes it s the former EZM R_A Ofcom 7 April 2009 F1-1

136 F: Stakeholder Meeting Notes Vertically integrated operators (as now, or even more of a walled garden) vs. rich supply chain - some or many operators focus on infrastructure and let other players focus on content and/or services Cell proliferation or continued dominance of the macro cellular architecture Degree of inefficiency in solutions/technologies/services e.g. spectrum in the wrong hands Anarchy (like WiFi) or orderly and distinct Spectrum usage rights Backhaul: plenty of fibre to the street/kerb/home or wireless backhaul from base station to core network. EZM R_A Ofcom 7 April 2009 F1-2

137 F: Stakeholder Meeting Notes F2. NOTES FROM MEETING WITH GRAHAM PLUMB - HEAD OF DISTRIBUTION, BBC F2.1 CURRENT AREAS OF SPECTRUM SHORTAGE FOR BROADCASTING Currently the BBC has a shortage of spectrum across all broadcast: Digital TV (HD but not freeview) DVB-2 will soon need to be implemented Each operator currently has only 1 HD channel Satellite - not enough transponders to broadcast to UK BBC has all the content for 3G. Programme Making and Special Events (PMSE) requires a lot of spectrum and currently the BBC share this spectrum with the MoD. This could become a problem if this spectrum is removed from MoD. F2.2 SPECTRAL EFFICIENCY Spectral efficiency is the focus for the BBC going forward this is achieved by improving the technologies. The introduction of Administered Incentive Pricing (AIP), which reflects the amount of spectrum they use in 2014, is forcing the BBC to make efficient use of their frequencies. F2.3 WHAT THE FUTURE OF BROADCASTING WILL LOOK LIKE In all the survey carried out by the BBC - Satellite mobile TV technologies is the lowest demanded service. Without any cost constraints the demand for spectrum would increase and in 2025 demand for spectrum will be high: Radio and broadcasting analogue BBC had 2 channels and needed 8 channels + for interactive this will move to around channels in the future They predict a growth in low price niche channels There will be a time shift from PVR to real-time There will be convergence and everything will be on demand IP then can switch off broadcast terminals. Therefore there will be a need a device that can do all of these things e.g. broadcast + IP (niche events) this will be the hybrid world There could be a shift to user generated content and there role could change to just being the content aggregator in the future The licence fee will probably not exist. EZM R_A Ofcom 7 April 2009 F2-1

138 F: Stakeholder Meeting Notes Technologies in 2025 will be: HD Super HD PVR s 3D TV. If the BBC were to convert all its existing channels to HD, using MPEG 4 technology DVB- 2 then a 50% saving would be made. This would then release some spectrum. In a broadcasting network it is simply about the number of channels unless an IP network is employed and then the number of users is important. EZM R_A Ofcom 7 April 2009 F2-2

139 F: Stakeholder Meeting Notes F3. NOTES FROM MEETING WITH IAN PEARSON - FUTUROLOGIST, FUTURIZON GMBH F3.1 THE MOST LIKELY SCENARIO IS A MAJOR SHIFT TO A WIRE-FREE WORLD It seems likely that cellular and other wireless communications networks will combine to create a ubiquitous wire-free world with extremely high data rates,.fixed networks will survive, and the core networks will likely remain as fixed infrastructure, but access will be overwhelmingly available over a wireless interface. Demand for everything wireless will increase; it is less clear whether capacity will keep pace with that demand. There will be the emergence of new networks which will provide short range hop to a fixed network similar to Skype phones today. This will result in service provider having to offer new bundled services to the consumer such as free services combined with paid for services. F3.2 NETWORK PROTOCOLS WILL CONVERGE ON IP - AND CHARGES FOR VOICE CALLS MAY BECOME A THING OF THE PAST Mobile technology is very likely to join fixed line comms in the migration to a single network protocol, namely IP, and this standardisation will promote service competition between mobile operators. By 2025 the charge for making a call could be, in practical terms, nothing. The reason for the level of charge today is because the network operators need to recover (and achieve a return on) the network build costs, and the difficulty of using fixed tariffs in a market where operators wish to encourage consumer take-up. In the future, this may not be the case; customers will be able to route their traffic via lower cost alternative channels (e.g. over IP), driving down volume-related charges so that in the long run they reflect long run marginal costs of capacity. This coupled with steep declines in the cost of incremental capacity may bring call charges down to, or close to, zero. F3.3 DEVICES WILL MINIATURISE AND MAY PROLIFERATE ( DIGITAL JEWELLERY ) By 2025 we may observe the disappearance of the mobile phone as we know it today. The functionality currently contained in a telephone (e.g. camera, mobile phone, music/ data store, etc) would be split into its individual components, which will be sold separately as individual interoperable miniaturised items, likely costing under 10 apiece. This will be possible because of the increased miniaturisation of chip sets and batteries by The complication with this digital jewellery view of future devices will be around the user interface, and this will rely on new technology being developed. The first generation of these interfaces is currently being developed, and will be a implemented in the form of either glasses or more likely in a full video visor. The cost would initially be around 300. The video visor may initially be VGA quality but will be 3D high resolution by EZM R_A Ofcom 7 April 2009 F3-1

140 F: Stakeholder Meeting Notes The interface would enable the user to straddle the boundary between the physical and virtual worlds, for example to display virtual imaging personalised to the individual, time, location, and context; for example, to either draw attention to, or to reduce their awareness of, a nearby fast food outlet and its current promotions, according to the preferences of the user. It would also support infotainment applications such as streaming real time video. This physical/virtual hybridisation would give rise to new areas of concern, e.g. digital trespassing: who controls the user s experience, how is it safeguarded against the latter day equivalents of junk mail, viruses, and other unsolicited or unwanted material? The other constraint on device miniaturisation is the requirement for an antenna, however even this constraint might eventually be overcome by using the body of the wearer as the physical antenna, coupled with suitable signal processing technology. F3.4 TECHNOLOGY WILL CONTINUE TO GET EASIER TO USE, AND CONTENT WILL PROLIFERATE The internet and social networking sites will become common place. They will be made easier to use so ordinary/everyday people will be able to use it as a source to communicate and promote themselves personally or their business. Today there are some 100,000 ISPs in the UK but this will become 65million ISPs, each individual will using the technology to promote and position either themselves and/or their business. Ease of use will continue to accelerate so that every individual can develop and operate their own web presence(s) - website, blog, trading shell, other frontage. The capability to generate good content will most likely be spread more widely than it is a present. Individuals will be using high quality images and considerably more video than today. Delivering this content will require a large amount of bandwidth. There may be diversification of business models with this proliferation of content pay to receive, pay to publish, etc. and some further erosion of the power of the broadcast networks. Business models are always going to be a consideration e.g. the continual tension between musical artist and record label, or the imposition of the obligatory one minute or so of copyright warning at the start of every DVD, whether legitimate or counterfeit!.. A related development is the likely overlap and convergence between broadcast media, publishing, other printed media, gaming, and communications, for example technologies approaching digital paper would blur several of these boundaries. This convergence will also need a parallel convergence in regulatory supervision. Regulators may also need to alter their ground for intervention, e.g. to address confusion-based marketing strategies. F.1.1 Social devices / Ego Badges may provide intermediation for users Technology will provide services to intermediate for people, for example social / ego badges could share personal information according to proximity and other factors, similar to what is currently contains in social networking sites for example. These devices could tell you that you have a mutual interest or opportunity (e.g. in the context of a singles dating application) shared with another person that you have just passed in the street who is also wearing an interoperative device. These devices may also support ad hoc peer to peer data sharing / publishing / communication. EZM R_A Ofcom 7 April 2009 F3-2

141 F: Stakeholder Meeting Notes As with other technologies, these devices could be used purely for the benefit of the consumer, or they may operate for the benefit of other parties, e.g. in the same example, the nearby supermarket may use them to selectively promote products and make recommendations; the development path will be shaped by issues of trust as well as the underlying business model (wearer pays, advertiser pays, etc). F3.5 TECHNOLOGY, IN PARTICULAR WIREFREE, WILL HAVE A DRAMATIC IMPACT ON THE TRANSPORT SECTOR Intelligent highways, intelligent in-(and inter-)vehicle systems have huge potential to increase the efficiency of the transport sector. For example to take on the more routine elements of driver activity, to provide additional safeguards leading towards vehicles which interact with each other, and the infrastructure, in an intelligent manner. This may lead to considerable advances in other aspects of the transport; for example, if the vehicle and infrastructure cooperate intelligently then acceleration and braking can be reduced, and if the risk of a collision is eliminated then the weight of the vehicle can be reduced. The consumer perception of transport might shift away from the personal vehicle as ego/status symbol, towards a more utilitarian and networked view of what vehicles are and how they operate. F3.6 MICRO-ELECTRONIC DEVICES MAY BE DEPLOYED ON - AND INSIDE - THE BODY 10nm electronics could allow a number of health and wellbeing technologies to emerge, for example: Telemedicine, for example insulin monitoring: chips the size of a grain or even smaller by 2025 can be embedded under the skin sending information back to a central centre in real time monitoring insulin levels and administering the necessary drugs in real tine as is required. Active skin, to allow the user to change the cosmetic aspects of their appearance (makeup, tattoos, etc) This will present new challenges for regulatory authorities, as technology opens up new possibilities that straddle existing boundaries, e.g. Ofcom for comms, GMC for medical issues. As with other areas of development, there is the issue of for whose benefit it is deployed and used, linked to the issues of business models and who pays; one of the more frivolous examples of this might be the potential for digital makeup to be hacked F3.7 THESE CHANGES MAY OCCUR EVEN IF THE SOCIO-ECONOMIC OUTTURN IS POOR It is possible that GDP per head will stagnate or even decrease in real terms. Whilst the UK / Europe will benefit from the enlarged global markets as the emerging BRIC+ economies start to develop, in the case of some natural resources there will be greater and greater competition. It will be crucial to retain proper incentives in society, to encourage compliance and discourage behaviours with high external costs, e.g. crime and pollution. The UK (and other) governments may continue on the trajectory towards a heavily supervised society driven by the fear of threats both external and internal. EZM R_A Ofcom 7 April 2009 F3-3

142 F: Stakeholder Meeting Notes The recent pattern of migration may shift away from UK / Europe / North America, towards Asia, reflecting the wider rebalancing of economic power. Per capita incomes in the developed word are likely to remain ahead of those elsewhere, but the gap between the developed and developing economies will narrow. As life expectancy continues to increase perhaps following the same linear trend as the past 200 years, i.e. adding two years every decade - the UK and the rest of Europe will have to escalate the state retirement age to reflect this trend. EZM R_A Ofcom 7 April 2009 F3-4

143 APPENDIX G: MODEL SPECIFICATION EZM R_A Ofcom 7 April 2009

144 G: Model Specification TABLE OF CONTENTS G1. Introduction G1-1 G1.1 Purpose of the model G1-1 G1.2 Model constraints G1-1 G1.3 Structure of this document G1-2 G2. The Dimensions of the Spectrum Shortage Model G2-1 G2.1 Scope of the model and the defiintion of the key terms G2-1 G2.2 Building blocks G2-3 G2.3 Services and applications G2-3 G2.4 Technologies G2-4 G2.5 Geography G2-4 G2.6 Scenarios G2-5 G2.7 Spectrum G2-5 G2.8 Network dimensioning and costs G2-6 G2.9 Out of scope G2-6 G3. Methodology for Predicting Areas of Spectrum Shortages G3-1 G3.1 Deliverables G3-2 G4. Service Demand G4-1 G4.1 Inputs G4-1 G4.2 Process G4-2 G4.3 Final preprocessing G4-3 G4.4 Outputs G4-3 G5. Spectrum Demand G5-1 G5.1 Inputs: Network data G5-1 G5.2 Processing: Network analysis and modelling G5-1 G5.3 Outputs G5-5 G6. Quantification of Spectrum Shortages G6-1 G6.1 Inputs G6-1 G6.2 Process G6-1 G6.3 Outputs G6-1 G7. Model Interfaces and Environment G7-1 G7.1 Fitting the model within the IT environment G7-2 G8. Glossary G8-1 EZM R_A Ofcom 7 April 2009 Gi

145 G: Model Specification G1. INTRODUCTION This appendix describes the specification for the model that PA Consulting Group is developing for Ofcom to Predict Areas of Spectrum Shortage over the next years. It sets the expectations of what the model will do and how it will do it, through: Setting out the main dimensions that shape the choice of our approach Defining the scope of the analysis Confirming how the model will be used in practice and how it will handle scenarios Outlining what input data is required and what the outputs will be. G1.1 PURPOSE OF THE MODEL The purpose of the model is to allow areas of spectrum shortage to be predicted, by considering future demand and growth of capacity in networks and services, and the likely changes in spectrum allocations and availability. The model will analyse a range of agreed potential scenarios, outlined in our Future Scenarios document. These scenarios were derived from discussions with stakeholders and they describe likely future changes that may affect the supply and demand of, spectrum in the next years. The model will then determine under which scenarios and timescales there is likely to be a shortage in spectrum, enabling Ofcom to proactively manage their spectrum planning in the immediate to medium future. The timeframe within which the model results will be presented on a yearly basis is from the present 1 until G1.2 MODEL CONSTRAINTS Data and forecasts from publicly available reports The same modelling framework will be employed throughout The model is to offer the user a series of predefined inputs which they may choose to alter for any or all scenarios Availability of historical data for all the services and applications identified at an appropriate level of detail Access to relevant data, for example usage, network architectures, and network dimensioning parameters Availability of relevant network technical architecture information to be used in the model Timely advice from experts and the Stakeholder Panel regarding key inputs and relationships used in the model as well as the development of scenarios. 1 Generally for fitting to current/most recent demand data. EZM R_A Ofcom 7 April 2009 G1-1

146 G: Model Specification G1.3 STRUCTURE OF THIS DOCUMENT The main body of the Model Specification document comprises: Section 2 An appraisal of the issues that define our approach to modelling spectrum shortages Section 3 A methodological overview of the model design that introduces each element of the problem Section 4, 5 and 6 A detailed description of each of the three individual components of the model, which includes: The input data required for the model An outline description of the calculations to be carried out in the model: the main calculation steps, data interfaces, and data structures The outputs from the model Section 7 How the model is to be used, its environment and its hardware and software requirements. EZM R_A Ofcom 7 April 2009 G1-2

147 G: Model Specification G2. THE DIMENSIONS OF THE SPECTRUM SHORTAGE MODEL G2.1 SCOPE OF THE MODEL AND THE DEFIINTION OF THE KEY TERMS The model is based on a series of Building Blocks, (i)-(iv) are set out in Figure G1, and (v) enables presentation of results at more detailed geographic levels: (i) Services six services, comprising: cellular, short range wireless, broadcast radio and TV, fixed wireless access, and backhaul (ii) Technologies within each service e.g. for cellular examples include GSM / LTE, for short-range wireless WiFi / Bluetooth, and for broadcast Mobile TV / HDTV etc (iii) Applications 2 voice, messaging, , download, TV, radio, gaming, etc (iv) Spectrum Bands (Frequencies) both the current allocations (e.g. 900, 1800 and 2100 MHz for cellular) and those in the pipeline (v) Geographies the neighbourhood types such as urban and rural, that primarily reflect the density of population / demand, and also influence signal propagation and hence coverage and number of sites. Using the above Building Blocks, different sets of network planning parameters and input data scenarios are developed to describe a likely future. A matrix of services, technologies, applications and frequencies in scope is provided in Figure G1. This details the four key Building Blocks in the model, with an indication of which applications are available by technology. Throughout the remainder of this paper, the words service, technology, and application are to be read and interpreted with the above descriptions and definitions in mind. 2 Traffic for each application will be divided into uplink and downlink components, given sufficient data. EZM R_A Ofcom 7 April 2009 G2-1

148 G: Model Specification Application Service Family Service Technology Voice Messaging (SMS, MMS, IM) E Mail WWW Video Streaming TV Radio Datacast PMSE Gaming Download P2P Telemetry and M2M Frequency 2G GSM N/A x x x x x 900MHz 1800MHz Paired Possible but slow 2.5G GPRS/ EDGE N/A x x N/A x N/A 900MHz 1800MHz Paired 3G UMTS N/A N/A N/A 900MHz 1800MHz 2100MHz Paired Cellular 3.5G (HSPA) N/A 4G LTE N/A 900MHz 1800MHz 2100MHz 700MHz 900MHz 1800MHz 2600MHz Paired Paired & Unpaired Mobile Mobile WiMax N/A 700MHz 900MHz 1800MHz 2600MHz Unpaired MBMS x x x x x x x x x 2100MHz Paired (Unpaired) Wireless Analogue Cordless** x x x x x x x x x x x x 47MHz Paired (Short Range) DECT** x N/A N/A x N/A N/A N/A N/A N/A N/A N/A N/A 1880MHz Paired Bluetooth N/A x N/A N/A N/A N/A 2.4GHz Unpaired UWB N/A N/A GHz Unpaired WiFi N/A 2.4GHz 5.8GHz Unpaired Broadcast Fixed Key: Radio Analogue N/A N/A N/A 0.2MHz MHz 2-30MHz MHz DAB x x x MHz Unpaired DRM x x x 2-30MHz Unpaired Analogue Not in isolation x N/A Not in isolation MHz Unpaired DVB-T x MHz Unpaired TV DVB-H x N/A MHz Unpaired DVB-T2 x N/A MHz Unpaired DVB-S x N/A ~10GHz Unpaired Fixed WiMax GHz Unpaired Fixed Wireless 3.4GHz TD/CDMA Acc. 1920MHz Unpaired Proprietary 3.4GHz Unpaired Back-haul Microwave >10GHz Paired = application available via technology; N/A = technology potentially capable, but not used to provide application. x = technology not capable of providing application; ** not currently confirmed as in scope, will be kept under review Unpaired Figure G1: Services, technologies, applications, and spectrum bands EZM R_A Ofcom 7 April 2008 G2-2

149 G: Model Specification G2.2 BUILDING BLOCKS The data and forecasts used in the model will be obtained from published sources and benchmarking, rather than primary research or original market analysis. The model will comprise of a set of data and parameters that will enable users to develop a range of scenarios by configuring the data and parameters to their needs. current data will provide a baseline (i.e. as-is scenario) for the relative comparison framework. Forecasts for service demand and spectrum demand will be calculated. From this, the results of a scenario can be compared to the as-is scenario to compare the scale of shortfalls in spectrum. Data Demand data (subscriptions, usage) Network Architecture and Technologies Available spectrum / allocations Processing and Results Demand for each Service (call minutes, MB, channels) Demand for Spectrum for each Service (MHz) Net surpluses or shortages of Spectrum (MHz) Figure G2: Model process The model will evaluate service demand, spectrum demand, and net shortage of spectrum year by year over the year time period. Where it is impractical to map the details, e.g. for technologies in the pipeline, the Stakeholder Panel will be used to generate or verify the assumptions used in the model. G2.3 SERVICES AND APPLICATIONS The demand for spectrum arising from each application and service will be shaped by three key elements: Underlying traffic / demand for services Required quality of service (including blocking probabilities, capacity to cope with peaky and bursty demand, speed in Mbps, and latency) Network constraints and capacity (including coverage of different neighbourhood types, site availability, and spectral efficiency). These three pieces of information will define the core description of the service within the model. EZM R_A Ofcom 7 April 2009 G2-3

150 G: Model Specification Of the services, cellular mobile is the area of most complexity and will be given the greatest emphasis, drilling down into demand in greater detail to the three main application families within cellular, namely voice, messaging, and data, and then into greater detail where necessary e.g. within data, as a minimum: , web browsing and transactions, downloads and peer to peer file sharing, gaming and other web applications, IPTV / content streaming. G2.4 TECHNOLOGIES For each service, there are currently a range of technologies that exist, and there are several more planned or expected over the time frame of interest. Figure G1 provides an overview of technologies and their frequency ranges by service that are currently identified to be in scope of this project. The distribution of demand across the differing technologies (e.g. 2G/3G / LTE) will be made using the current known demand combined with an evaluation of future development of demand, which will be agreed with the Stakeholder Panel. Forecasts for the demand in services where associated technologies are not yet available will also be estimated and agreed with the Stakeholder Panel and/or subject matter experts within the PA and the Stakeholder Panel. G2.5 GEOGRAPHY The majority of the data used in the model is only published at the national level. However, for some services, provision can vary according to the neighbourhood type, or geotype, e.g. Urban, Suburban, and Rural. The primary difference between these neighbourhood types is the density of population and demand, but they also have an impact on the availability of sites, and on RF propagation characteristics and ranges, and can affect the economics of rolling out sites and transmitters. The model will allow for and the density of sites to vary by these neighbourhood types and also by technology. High-level published statistics, such as that 90% of the UK population live in urban areas 3, will be used to derive the required level of disaggregation by neighbourhood type. Some assessments will be made for a special subset of hyper-dense urban areas, such as the City of London, major sporting or entertainment venues, or major transport hubs, where the daytime population can greatly exceed the night-time population, and where the population density can reach extremely high levels (of the order of 100,000 per km 2 and above). However this will not distract from the main thrust of the analysis, which will be on the regular urban areas where most of the population live and work for most of the time. 3 Economist publications, EZM R_A Ofcom 7 April 2009 G2-4

151 G: Model Specification G2.6 SCENARIOS Scenario s are one representation of how service demand, spectrum demand and spectrum shortages will unfold over a year period. A description of the scenarios that the model will support is provided in more detail in the Future Scenarios document, however at the highest level the model will support the analysis of the following types of scenarios: Changes to the demand for services Changes to the spectrum assignments for different services Changes to technologies (new technologies added, others dropped) Changes to the network architectures Changes to spectrum availability. An as is scenario will be established to indicate the future position of the networks, given the present configurations and strategies and continuations of current demand trends. This view will be the model s baseline 4. Other future to be scenarios will be defined and evaluated using the model by comparing the results against the baseline. Similarly, the model sensitivities can be tested to the planning parameters by setting up sensitivity scenarios. G2.7 SPECTRUM Spectrum demand will be modelled in the first instance at the level of service and then converted into a total demand for spectrum from each of the six services cellular, short range wireless, and the rest. However, there is also a desire to obtain results that break down the demand of spectrum into spectrum bands. For example, for cellular, at present there are three UK bands: 900 MHz, 1800 MHz and 2100 MHz, with others likely to be added. Broadly there are three options for handling this: 1. Attempt to predict demand for each individual band of spectrum for which there is a separate entry in the UK frequency allocation table. So e.g MHz would be one band, separate from 1800 MHz and 2600 MHz (but not subdivided by operator). 2. Apply some grouping,, aggregating bands with similar propagation characteristics. So e.g. 2100MHz might be aggregated with 1800Mz and/or 2600MHz, but not with 900MHz or 700MHz, because the propagation characteristics are too different. 3. Apply very broad groupings to bands e.g. 100MHz 470MHz, 470MHz-4GHz, 4GHz- 10GHz, so that as far as possible each service only falls into one band. 4 We have considered the case for including a true do-nothing scenario, i.e. where the demand for services is static or growing slowly (under 20% pa year on year), and minimal changes in network configurations. Our view is that this scenario is unlikely to occur and is not of great interest, and accordingly it has not been taken forward, EZM R_A Ofcom 7 April 2009 G2-5

152 G: Model Specification The modelling will take the second of these options. The model will group spectrum bands into five clusters, in the light of the available data on usage and propagation, these will be as follows: 100 to 300 MHz 300 to 1,000 MHz 1,000 to 3,000 MHz 3,000 to 5,000 MHz 5,000 to 10,000 MHz. The analysis will cover the spectrum 100MHz to 10GHz, with particular focus on the areas of greatest concern: 300MHz to 1GHz, followed by 1-3GHz, and then 3-5GHz. G2.8 NETWORK DIMENSIONING AND COSTS Estimates of sensible trade-offs will need to be made in the model, to reflect the decisions that network operators have to make e.g. whether or not to over-dimension the network to improve speed, resilience, and service levels. The assumptions made for these trade-offs in the key network components will be visible and editable to the user. They will be made at an aggregated level; the model will not go into great detail, since it is not attempting to determine the cost impact of spectrum shortages or the valuations for different spectrum bands. The focus of the project is in assessing true spectrum shortages, so for example in assessing the density of sites in a cellular or broadcasting network, to determine frequency reuse across the UK, what really matters is the number of useful sites that are realistically available 5, rather than the cost per site. The model will allow for a calibrated degree of efficiency / inefficiency in spectrum use by operators over the forecasting period. That is, there may be some redundancy in the use of spectrum to deliver applications to users, due to networks needing to over-dimension to allow for busy hour traffic, bursts in demand, and infrastructure maintenance. G2.9 OUT OF SCOPE It is worth clarifying some of the areas which are out of scope for this study: Services Satellite data and communications Defence and Military systems Technologies PMR/PAMR (Analogue, Tetra, Pages) GSM-R ISM/RFID Amateur radio 5 In the short to medium term, there are further constraints, namely the current installed base of sites, and the rate at which new sites can be added. EZM R_A Ofcom 7 April 2009 G2-6

153 G: Model Specification Radio astronomy Marine and Aviation. Spectrum Supply and demand outside the range of greatest interest and commercial value, 100MHz 10GHz. EZM R_A Ofcom 7 April 2009 G2-7

154 G: Model Specification G3. METHODOLOGY FOR PREDICTING AREAS OF SPECTRUM SHORTAGES The model will be made up of three main components as illustrated in the Figure G3. These are: Service demand (Shaded yellow) Spectrum demand (Shaded green) Quantification of spectrum shortages (Shaded blue). Figure G3: Modelling the demand for specific services and the net surpluses and shortages of spectrum The Service Demand component (Section G4) determines the projections of numbers of users and traffic per user for each of the services and applications. As a result the expected yearly traffic will be determined for each application, each in units appropriate to the application (e.g. MB, voice minutes, channels, terminals). The model will allow for several different demand levels for each application within each scenario. EZM R_A Ofcom 7 April 2009 G3-1

155 G: Model Specification The Spectrum Demand component (Section G5) determines the technology required for each service and application and its supporting network. The spectrum (MHz) required is calculated, taking account of network architectures, geography, service quality, and availability of useful sites. The Spectrum Shortage component (Section G6) determines the net shortage or surplus of spectrum by comparing the demand against the current and future supply of spectrum. G3.1 DELIVERABLES The deliverables will include both a model which delivers a single set of spectrum demand levels and a report covering the work done and the assumptions made. The output(s) of the model will include for each agreed scenario: Projections of future spectrum demand per service A sensitivity analysis of the results 6 The report will include: Details of the agreed scenarios A set of outputs for the model Projections of future excess supply / demand for spectrum, by band / by service An indication of bands where capacity might be available as a substitute for the band being studied Outputs from the research and validation stages of the study which support the reasoning behind the scenarios A user guide and technical documentation to support the use of the model by Ofcom. 6 The sensitivity analysis will evaluate the effects of minor changes in value of key input parameters on the results of the model (c.f. the identification of spectrum shortage). This will enable levels of uncertainty to be investigated and confirm the level of stability that the results provide. EZM R_A Ofcom 7 April 2009 G3-2

156 G: Model Specification G4. SERVICE DEMAND This section describes how the forecasts of demand will be determined for both network traffic and subscribers for each of the services and applications. The process employed to determine the Service Demand is highlighted in the Figure G4. Figure G4: The evaluation of Service Demand As shown originally in Figure G3, the Service Demand module can be thought of as consisting of two parts: The input data and the associated economic model, which generate the demand forecasts The final stage of pre-processing to get the data into the form required for the next module (the Spectrum Demand module). G4.1 INPUTS As previously noted, the data and forecasts used in the model will be obtained from published sources and benchmarking rather than requiring primary research or original market analysis. The input data in this component will largely comprise yearly estimates for: The initial population (including numbers of subscribers) using a service and application across geographical areas and bands, as well as a historic growth rate Units of demand for each application, including initial average use parameters. For example, for cellular voice average use by minutes of use per month, for SMS average use by number of messages per month Initial and future take-up of each service and application (indicated by start date). In addition, the data collected at this stage will also provide any conversion factors needed to combine different units of demand. Where raw published data exists and is sufficient, then it will be used to provide additional evidence for the forecasts of demand by service and application used in the model. Such data will consist of numbers of subscriptions (devices, users, or households) and volume (e.g. total usage) for a service and application, and will be gathered from existing published data either from Ofcom of other sources. EZM R_A Ofcom 7 April 2009 G4-1

157 G: Model Specification G4.2 PROCESS Where raw data is sufficient the forecasts of future network traffic will use short-run relationships between key economic variables (i.e. price, quantity, GDP etc). With such short run specification, forecasting equations for expected traffic for each service and application are provided from which forecasts of each service and application will be derived. This has the potential to provide not only the forecasts for each service and application over the next years but also the dynamic relationship between, for example, changing prices and demand. A brief overview of an econometric approach to undertake this is provided in the following frame: Frame 1. An econometric approach to measuring service demand. The econometric approach employed in modelling the Service Demand is an Engle-Granger model. This is based on the economic foundations that there exists a simple long run relationship between network use, E, (e.g. total revenue), economic activity, Y, (e.g. consumer income) and network price, P: E = f ( Y, P, µ ) where µ represents the underlying demand trend for a service. Using this model, long-run price elasticities are estimated and co-integration tested. The importance of the test of co-integration indicates that the model confirms a valid long-run relationship. Based on this long run relationship, the short-run forecasting equation can then be determined that includes the relevant lags for price and income. Concluding that the model developed is calculating the response of consumers to price changes accurately. The output will be a set of price elasticities that measure this relationship. In cases where raw data unavailable, forecasts reported in the literature will be taken and applied. In this case, diffusion curves per service and application will be applied to growth in traffic per subscriber to gain a yearly forecast. An example curve is provided in Figure G5. (Source: Alptekin et al., University of Surrey) Figure G5: Example diffusion curve for a cellular service EZM R_A Ofcom 7 April 2009 G4-2

158 G: Model Specification Where reliability of forecasts cannot be verified through analysis, the Stakeholder Panel will be asked to check validity of assumptions that have been made in the model. G4.3 FINAL PREPROCESSING The forecasts of demand will be organised in a layout that makes it reasonably manageable for a user to browse them and apply sense checks, e.g. by comparing the demand figures for related items by eye. This data will then be pre-processed in order to get it into the structures required for the downstream modelling and analysis of each Scenario in turn. G4.4 OUTPUTS Three main measures of demand for network services will need to be estimated: Average demand Busy hour demand the level of demand reached on a typical day Required capacity what the network needs to be able to handle. Of these three, the first is calculated in the Service Demand module, the remaining adjustments are carried out in the Spectrum Demand module. Supplementary outputs such as numbers of subscribers / subscriptions, will also be passed to the Spectrum Demand module where required. Demand will be expressed in different units for each application and the definitions of the units in each case will also be output. Projections of total traffic in common units (MB/hour) will also be output 7. In addition to the prediction of demand by service and application, this component of the model will also set up forecasts regarding the migration of subscribers between technologies. For example, for cellular, users will switch in coming years from 2G to 3G and from 3G to 4G as well as other switching patterns. The user will be able to set parameters at this stage to define the percentage of customers who are estimated to transfer between technologies in the base-year and the likely increase in this rate of change in future years.. The results determined for each service and application will then be fed into the Spectrum Demand module. 7 Recognising that this projection of traffic in common units can be done across technologies within the same service, and across similar services, but that aggregation of broadcast and nonbroadcast services should be treated with some caution. EZM R_A Ofcom 7 April 2009 G4-3

159 G: Model Specification G5. SPECTRUM DEMAND This section describes how the architecture of a spectrum is represented in the model and how it is used to estimate the capacity implications on the proposed network structure given a projected traffic forecast. As shown in Figure G3, the module can be thought of in two parts: the network input data, and the network analysis and modelling. G5.1 INPUTS: NETWORK DATA The Building Blocks of the architecture are developed from the model dimensions with key inputs to the network architecture that includes: Network plans enables the user to control the main network planning parameters for a service (e.g. for cellular this includes uplink and downlink capacities and bands by geotypes). These parameters form the basis on which scenarios will be defined for each service. Service traffic and demand (the output from the Service Demand - see Section 4) this contains the projected subscriber traffic until 2025 for a service and application and is used to estimate the capacity demands on the network. It is obtained from the forecasts provided from the analysis of Service Demand (Section 4). This provides the likely traffic requirements of a network at the aggregate level Projected requirements will be converted into data rates so that they can be cross referenced with the capacity limits of the network. Geographical data to derive the required level of spatial disaggregation into neighbourhood types (also referred to in the model as geotypes ) for population and area covered in km 2. Network traffic distribution calibrated to expectations of busy hour traffic, providing an estimate of the distribution of traffic across the week or year, with adjustments for additional headroom required in order to provide an acceptable level of service to customers. The Spectrum Demand module will, where possible provide a consistent structure across each service. Some differences will be required e.g. for broadcast networks vs. communications networks, but these will be kept to a minimum. The categories listed above will be tailored to the detail required for each service. G5.2 PROCESSING: NETWORK ANALYSIS AND MODELLING Sub-models representing the demand for spectrum will be developed for the services described in Section G2.3. The inputs and outputs will for the most part be comparable across services, however the calculation process will be selectively tailored to each service as appropriate. EZM R_A Ofcom 7 April 2009 G5-1

160 G: Model Specification Figure G6: Spectrum Demand module - indicative structure EZM R_A Ofcom 7 April 2009 G5-2

161 G: Model Specification a. CELLULAR An outline of the Spectrum Demand modelling process for the cellular service is illustrated in Figure G6. The components are comprised of the following functions: Inputs Process Outputs Service demand (by technology, application and year) for cellular, subscribers and traffic are imported into this module from the Service Demand module. Conversion factors to get Demand expressed in common units. The incoming demand data has different metrics for each application, e.g. voice traffic in minutes of use, messaging is in number of messages, is in MB, these raw numbers cannot simply be added together. Relationships between services, technologies, and applications, and Bands, e.g. which technologies use which band, which applications use each technology. Network planning (by technology and year) includes the current and future availability of sites for base stations, frequency reuse, and spectral efficiency. Geographical information, to map the national demand down to neighbourhood types. 8 Peakiness of the traffic, expressed as a ratio of traffic in the busy hour of the week as a multiple of the traffic in the average hour (over the week or year). Headroom required in order to provide an acceptable Quality of service. For example, voice traffic can accept a modest probability, of the order of 1%, of a busy hour call being blocked due to congestion in the cell. This imposes a requirement for headroom over and above what is needed to just carry the busy hour traffic and these factors can range from a few percent to a factor of 2 or more. Convert the traffic from the various different applications into standardised units. Scale up from the traffic offered, to the required capacity, allowing for busy hour / peakiness and for additional headroom requirements. Allocate this requirement to neighbourhood types and compare it against the availability of sites to determine the capacity requirement per site and the spectrum requirement in MHz. Spectrum demand (by technology and year) capacity required as calculated in the process provides an estimate of spectrum demand which can in turn be compared to current spectrum allocated to give an indicated spectrum shortage. Sense checks, such as the total traffic offered by year. 8 Note that a set of simple rules will be implemented to govern the distribution of demand across geotypes, e.g. allocating demand pro rata to population rather than land area. EZM R_A Ofcom 7 April 2009 G5-3

162 G: Model Specification b. WIRELESS (SHORT RANGE) The calculation components for short-range wireless resemble those for cellular, with some slight variations to reflect the differences between the two services. The input data is of course completely different the volumes, the mapping of applications to technologies, the network dimensioning parameters, and the network technical parameters such as spectral efficiency, all take values to reflect the way that short-range wireless services and technologies operate. The outputs take the same form as those for cellular, i.e. spectrum demand in MHz by service x band x year. c. BROADCAST (RADIO AND TV) Broadcast is one-way transmission (i.e. downlink only) and there are therefore a few differences in the way that the model handles these services. The main difference is the demand is expressed in terms of number of channels, i.e. from a broadcasting operator viewpoint rather than a consumer viewpoint. The demand for each channel is in essence divorced from considerations of number of people consuming the channel, i.e. demand is considered as demand for the capacity to transmit to the population, rather than (as for cellular) a demand for pairwise two-way communication. The outputs take the same form as those for cellular, i.e. spectrum demand in MHz by service x band x year. d. FIXED (FIXED WIRELESS ACCESS (FWA) AND BACKHAUL) Demand will be calculated separately for each of the technologies (Fixed WiMAX, TD- CDMA, etc) using network planning parameters for fixed link networks. For FWA technologies the calculation process will be similar to that for cellular, again with different parameters used to reflect differences in network architectures. For Backhaul, the distribution of the number of links and/ or nodes in each capacity band will be estimated (e.g. number of E1s) and this will be converted into a demand for spectrum. The outputs take the same form as those for cellular, i.e. spectrum demand in MHz by service x band x year. G5.2.1 Spectral Efficiency and Frequency Re-Use Spectral efficiencies (bits/s per Hz) will be taken as input parameters which can vary over time and by technology. A parameter indicating the frequency reuse factor will quantify the effect on spectrum demand due to the need to de-conflict neighbouring sites. A standard 1/K approach will be implemented where K (typically 7 or 9) is the number of cells which cannot share the same frequencies. There will be scope for K to vary between services and by technologies. EZM R_A Ofcom 7 April 2009 G5-4

163 G: Model Specification G5.3 OUTPUTS From the spectrum demand analysis, the projections of spectrum available per service per year (e.g. total MHz) are provided. In addition, a summary of the traffic in each year, in common units, will be calculated as a sense check on the assumptions behind the demand data. An indicative list of outputs is as follows: Spectrum demand by service per year Breakdown of spectrum demand into a broad set of bands Breakdown of spectrum demand by technology and year Statistics on the total traffic and other measures of how much spectrum is utilised, by service and neighbourhood type and year. The results determined for each service will then be fed into the spectrum capacity component to evaluate spectrum shortages. EZM R_A Ofcom 7 April 2009 G5-5

164 G: Model Specification G6. QUANTIFICATION OF SPECTRUM SHORTAGES This section presents the calculations performed by the model in order to support the scenarios and sensitivity analyses. The model will examine the potential impact of a range of events and future business possibilities on spectrum, such as: Changes in spectrum availability Changes in the demand for spectrum, driven by changes in the demand for services and by the efficiency of the networks in delivering that demand within the spectrum available. Each scenario or sensitivity analysis comprises a combination of as many of the above factors as required. G6.1 INPUTS There are three inputs: Demand for spectrum, in MHz, is provided from the previous stage in the calculations Current allocations of spectrum Changes in spectrum availability, e.g. spectrum release, reassignment, licence renewals or withdrawals. G6.2 PROCESS There are two steps: Determine the spectrum available, by service and year, by taking the initial allocations and applying each change as it falls due Determine the net surplus of shortage, by comparing the demands for spectrum against what is available, and provide various breakdowns of this data. G6.3 OUTPUTS A global view of when and where capacity shortfalls will occur will be the key result of this module. The spectrum capacity will be measured through the capacity of the network against the spectrum demand for each service. From this spectrum shortage can be derived. Results will be presented to the user in the form of tables and charts. An example of this (with fictional data) is given in Figure G7 where total spectrum demand is presented with an indication of current spectrum allocated for active technologies. EZM R_A Ofcom 7 April 2009 G6-1

165 G: Model Specification The key results include: The demand required for spectrum, by service (with further breakdowns where appropriate) The configuration of spectrum available to meet demand, and the net surplus or shortage Spectrum demand (MHz) Total spectrum allocated Year Figure G7: Example output Demand, supply and net surplus / shortage of spectrum for cellular mobile in one scenario (example output ONLY) Examples of potential outputs subject to assessment of priorities - include: Total spectrum demand per year Breakdown of spectrum demand per year into a set of bands Total spectrum shortage per year Breakdown of spectrum shortage per year into a set of bands Breakdown of spectrum shortage per year by neighbourhood types, i.e. which types are short of spectrum Breakdown of spectrum shortage per year by services Approximate percentage of total Service Demand not met due to spectrum shortage per year by services. EZM R_A Ofcom 7 April 2009 G6-2

166 G: Model Specification G6.3.1 Other Considerations a. OPERATORS There are several operators with license to use spectrum, and the number of operators with licenses to use spectrum and the allocation of that spectrum across operators is likely to change significantly in the coming years. We do not propose to analyse demand by operator within the model, but will parameterise the efficiency of spectrum usage so that the effects of industry structure can be represented in a reasonably fair manner. b. SENSITIVITIES The model will be tested for sensitivity to key parameters and data assumptions. This will involve setting up some test scenarios that concentrate on evaluating the effect of one key parameter in turn in order to gauge its effect on the results. EZM R_A Ofcom 7 April 2009 G6-3

167 G: Model Specification G7. MODEL INTERFACES AND ENVIRONMENT The main components of the model, described in sections 4-6, will be developed and maintained as independent modules, but will be linked in the calculation process. The user(s) 9 will be able to be run each of these three components in what is effectively a stand-alone environment and well as running the components jointly in a fully integrated environment. The model structure, parameters, forecasts and results will be implemented in MS Excel for the evaluation of spectrum capacity. The results will be presented to the user in the form of tables and charts and will include: The demand for spectrum, by service The configuration of spectrum available to meet demand. An accompanying user guide will be provided that describes the main components of the model and how to use them. Users will not need anything more than a basic familiarity with MS Excel and MS Office in order to navigate through the model and browse the inputs and results. Users may need a moderate degree of familiarity with Excel in order to explore the details of the calculations. In some steps in the calculation, a basic familiarity with VBA may be required in order to follow the logic. The model will comprise a set of core base data and parameters. From the core data, users will be able to develop a range of scenarios by configuring the data and parameters to their needs. This framework provides a means for comparison of alternative scenarios. Once defined, scenarios can be saved for future development, reporting and comparison with others. The final version of the model handed over to Ofcom will follow the structure outlined in section 3, but will include some pragmatic fine tuning, for example: Adding some high-level splits to the calculation process, such as splitting the Service Demand module into its two components (the source data, and the analysis/preprocessing), and splitting the Spectrum Demand module into two (the source data, and the analysis) Incorporating the Spectrum Shortage module with the Spectrum Demand calculations, on the grounds that the final calculations of spectrum shortage do not warrant the additional overhead of a separate module and interface. 9 The model will be developed on the assumption that there will be one user at any one time, but there may be different users sharing access to the same copy of the model within Ofcom, and these users may use variant copies of the model for ad hoc analysis. If there are multiple users, they may need to institute some tracking and logging of their interactions with the model, e.g. version control, and to avoid distributing it outside Ofcom. Like most Excel-based models, the model will not have the hard boundaries distinction between data and functionality that would be needed to support a fully fledged multi-user implementation. EZM R_A Ofcom 7 April 2009 G7-1

168 G: Model Specification The diagram below shows how this particular fine tuning of the Excel implementation would affect the relationship between the modules. There is no difference in the functionality or the results, just some improvements in the usability of the model (e.g. separation of data and functionality). Figure G8: Fine tuning of the design to work efficiently within Excel (Design on the left; Excel implementation on the right) G7.1 FITTING THE MODEL WITHIN THE IT ENVIRONMENT The model will be developed in MS Excel 2003 in MS Windows XP. The modelling system is a stand-alone application. Provided that a medium/ high specification and fully functioning 10 PC capable of operating under Windows XP is used, the model should operate under any standard hardware configuration. It is advisable that the PC running the model has a minimum of 1GB of RAM and a processor speed of at least 2 GHz. 10 Fully functioning means that e.g. macros are enabled, and that the overhead from the load set and other applications on the PC does not have a materially detrimental impact on its performance, functional capability, or reliability. EZM R_A Ofcom 7 April 2009 G7-2