Geographic Sharing in C-band Final Report

Transcription

1 Geographic Sharing in C-band Final Report Transfinite Systems Ltd Tel: +44 (0) C Rathbone Square Fax: +44 (0) Tanfield Road Croydon CR0 1BT Web:

2 Geographic Sharing in C-band Page 2 Documentation Control Project: Geographic Sharing in C-band Reference: MC/339 Authors: Transfinite Systems Ltd Date: 31 st May 2015 Status Final Version: 1.6 Version History: 1.2 First release 1.3 Updated with editorial changes 1.4 Updated with editorial changes 1.5 Updated with editorial changes 1.6 Updated with editorial changes This report was commissioned by Ofcom to provide an independent view into geographic sharing in C-band (3.6 to 4.2 GHz). The assumptions, conclusions and recommendations expressed in this report are entirely those of the consultants and should not be attributed to Ofcom. The International System of Units is used throughout unless otherwise specified and all maps are orientated with north towards the top.

3 Geographic Sharing in C-band Page 3 Brief Table of Contents 1 Executive Summary Introduction and Background Assignment Data and Modelling Parameters Summary Of UK Spectrum Availability Interference Zone Analysis Adjacent Channel Interference Issues Summary Of High Resolution Detailed Analysis Mitigations Conclusions and Recommendations Annex: Fixed Links Interference Zone Analysis Annex: NFD and Spectrum Masks Annex: High Resolution Analysis Annex: National Maps Full Set of Data Annex: Acronyms and Abbreviations Input Data...178

4 Geographic Sharing in C-band Page 4 Full Table of Contents 1 Executive Summary Work Undertaken Key Findings Recommendations for Additional Study Document Structure Introduction and Background Purpose and Scope International Background Assignment Data and Modelling Parameters Protection Criteria Baseline IMT Parameters Satellite Earth Station Data Fixed Link Data Geotype Data Summary Of UK Spectrum Availability Methodology Available Spectrum Separate Analysis of Satellite Earth Stations and Fixed Links Conclusions Interference Zone Analysis Satellite Earth Station Interference Zone Analysis Fixed Links Interference Zone Analysis Adjacent Channel Interference Issues Adjacent Channel Interference into Fixed Links NFD Calculations for Satellite Earth Station and IMT Interference Notes and Conclusions on NFD and ACI Summary Of High Resolution Detailed Analysis Overview Factors Specific to the High Resolution Analysis Earth Station Test Case BT Tower Test Case Power and Traffic...64

5 Geographic Sharing in C-band Page Conclusions from High Resolution Detailed Analysis Mitigations Mobile Network Based Mitigation Earth Station Mitigation Fixed Link Based Mitigation Interference Constraints in Mobile Licence Frequency or Geographic Migration Conclusions and Recommendations Conclusions Recommendations Annex: Fixed Links Interference Zone Analysis Single-Entry Interference Transmitter Power Receiver Interference Threshold Antennas Annex: NFD and Spectrum Masks Net Filter Discrimination Spectrum Masks Annex: High Resolution Analysis Overview Propagation and Surface Data Earth Station VSAT Test Case BT Tower Test Case Power and Traffic Conclusions Annex: National Maps Full Set of Data Sharing with Satellite Earth Stations and Fixed Links Combined Sharing with Satellite Earth Stations only Sharing with Fixed Links Only Annex: Acronyms and Abbreviations Input Data...178

6 Geographic Sharing in C-band Page 6 1 EXECUTIVE SUMMARY This is the Final Report of a study by Transfinite Systems into geographic sharing in C-band (3.6 GHz to 4.2 GHz). The study aimed to give an assessment of the potential to share spectrum at C-band between incumbent users (fixed links and satellite Earth stations) and new mobile services, in particular IMT-A. 1.1 Work Undertaken The following bullets outline the work undertaken. We reviewed studies within the ITU-R JTG to select suitable parameters to use in sharing analysis. We analysed licensing data provided by Ofcom for point to point fixed links and satellite Earth stations. Licensing data were imported into our standard Visualyse simulation tool and into an optimised version customised to handle the production of UK-wide maps of available spectrum. A calculation methodology for assessing whether a specific location is suitable for sharing between mobile networks and incumbent services was defined. We undertook three levels of analysis: o o o Spectrum Availability Analysis, maps and statistics of UK wide availability of spectrum categorised by geo-type, using a 5 MHz channel raster and 1 square kilometre pixelation. This is based on the total spectrum available net of the allocation to UK Broadband a total of 430 MHz. This analysis is described and summarised in Section 4 and the full set of results given in Section 13. Interference Zone Analysis, which used terrain and land use databases to identify the area around selected fixed links and satellite Earth stations where deployment of mobile networks could cause interference. This analysis is described in Section 5. High Resolution Analysis, which used a high-resolution surface database to examine, in detail, a scenario in which fixed links or satellite Earth stations were located in the types of dense urban areas likely to be the key locations for mobile small cell deployment. This analysis is described in Section 7. We investigated options for mitigation including accounting for network traffic variabililty, selecting locations to deploy mobile network base stations, use of larger dish antenna by the fixed links or satellite Earth stations and directional antennas at base stations. A combination of mitigation techniques would provide extra margin in the I/N calculations. There are multiple possible combinations, and these run variations are set out in Section The potential impact of these mitigations on sharing scenarios was assessed in a general way as discussed in Section

7 Geographic Sharing in C-band Page 7 The analysis concentrates on the scenario where the IMT-A system in these bands is used for capacity enhancement in areas of higher population density. This in turn means we have modelled indoor and outdoor small cells and, where results are related to population numbers, these have been deployed in urban, dense urban and hotspot areas (see Section 3.5 for discussion of this). Other variants of IMT-A deployments have different parameters and could be deployed in rural areas. The potential for sharing for this type of system has not be addressed in this report. 1.2 Key Findings The principal conclusion is that there is scope for sharing spectrum in this band. The reason we believe this is that, as Figure 1-1 shows, even in the baseline case, 80% of the GHz spectrum band is available to 50% of the urban+ population (urban+ means areas that are classified as urban, dense urban or hotspot as discussed in Section 3.5). This represents a massive potential economic value. Half the existing spectrum is available to 65% of the urban+ population, and this increases to around 90% if 20 db of mitigation is applied. On the other hand, even on the application of 20 db of mitigation, potential interference cannot be ruled out in some populous areas. Whether 20 db mitigation is possible in all cases is not clear, but our simulations indicate that improved modelling based on higher resolution surface data could, in most cases, result in an additional 10 db (see Section and ). The additional loss due to more accurate modelling depends on the details of the local environment, 10 db is a conservative figure for the cases we have studied and for a built-up environment. The implication is that a managed approach to shared access of the band based on geographic sharing has great merit. Spectrum is available but the possibility of interference remains in some locations under all assumptions. We find little difference between the whole frequency range and a separate analysis of GHz and GHz. The constraint on operation of mobile in C-band is dominated by the need to protect fixed links. Protection of satellite Earth stations is much less constraining. Fixed links in general present a more difficult geometry (See Figure 1-4) and, in addition, some fixed links are deployed across urban areas. In particular, we see that the fixed links across central London cover large, densely populated areas, hence potential deny use of the spectrum to a large number of mobile users. Some Earth stations may also experience interference from large areas, but, generally speaking, these Earth stations are in less populated areas, and the impact is consequentially lower Baseline Spectrum Available The primary numerical result is the amount of spectrum available in areas of high population density, for example in hotspots, urban and dense urban areas (accounting for million people in our model). We have concentrated on areas of high population density where small cells are more likely to be deployed. The definition of population geotypes is discussed in Section

8 Geographic Sharing in C-band Page 8 Figure 1-1 shows a cumulative distribution function (CDF) of the availability of spectrum across the UK by percentage of population in urban, dense urban and hotspot areas. Figure 1-2 shows a colour coded representation of the availability of spectrum across the UK as a whole. Figure 1-1: Availability of Spectrum in MHz by Percentage of Urban+ Population under Baseline Assumptions

9 Geographic Sharing in C-band Page 9 Figure 1-2: Colour Coded Map of Spectrum Available in Sharing with Satellite Earth Stations and Fixed Links in GHz Effect of Mitigation Detailed analysis presented in Section 7 shows that various mitigations can be applied in many cases. There are lots of combinations of mitigation that

10 Geographic Sharing in C-band Page 10 could be applied, and for each mitigation technique the additional margin that results is variable (dependent on assumptions and detailed modelling). One way to assess the affects of mitigation is to relax the I/N threshold by a fixed value and look at the effect on spectrum availability. In our analysis we have looked at 5, 10 and 20 db relaxations, Figure 1-3 shows the spectrum available when 5, 10 and 20 db of mitigation is applied to the baseline case. Figure 1-3: Availability of Spectrum in MHz by Percentage of urban+ Population under Baseline Assumptions and with 5, 10 and 20 db Mitigations Applied An overview of the effect of mitigation can be seen in Figure 1-4 below. The interpretation of colour-coded maps is discussed in detail in Section In the case shown below, areas coloured blue-green have over 50% of the spectrum in the band available. The picture improves significantly from the baseline case to the case where 20 db mitigation is applied, 80% of the population have access to around 295 MHz of spectrum compared to 60 MHz. Even so, sharing does not become a trivial issue, 10% of the population would have access to less than 50% of the spectrum and sharing would still need to be managed.

11 Geographic Sharing in C-band Page 11 Baseline Case 5 db Relaxation 10 db Relaxation 20 db Relaxation Figure 1-4: The Available Spectrum in London area in the Baseline Case and with 5, 10 and 20 db of Mitigation, in GHz Sharing with all Satellite Earth Stations and Fixed Link Carriers More detail on the available spectrum is given in Section 4 and the full background findings are in Section Other Findings Other key findings of the project are:

12 Geographic Sharing in C-band Page 12 Sharing with fixed links is harder than with satellite Earth stations because of the geometry involved, as shown in Figure 1-5. Earth Station Elevation > 0 Fixed Link Receiver Fixed Link Elevation < 0 Lower diffraction and decreased relative gain Higher diffraction + increased relative gain Earth Station Figure 1-5: Comparison of Earth Station vs. Fixed Link Geometry Base Station The operation of mobile network base stations in channels adjacent to fixed link and Earth station receivers is possible with minimal constraints as there is significant net filter discrimination (NFD). However, the potential for adjacent channel interference cannot be wholly ignored in exploring the potential to share the spectrum. The spectrum availability maps produced in this study are likely to be conservative. High-resolution analysis suggests that there is the potential to operate mobile services within the interference zones with some realistic constraints on deployment and taking into account the local built environment. Mitigations available include the use of directional antennas at the mobile network base station and pointing away from the fixed link or Earth station receiver. This reduces interference relative to the omni-directional antenna assumed for ITU-R JTG Studies. The methodology used to calculate the average EIRP of the mobile network base station was identified as being conservative. It did not take account of the variation in traffic levels during the day. This could reduce interference by 2.3 db. The methodology used by the ITU-R JTG to assess clutter loss using the model in Recommendation ITU-R P.452 produces conservative results compared to using a surface database and diffraction. This is particularly important given that mobile network base stations are likely to be deployed in dense urban+ areas below building height.

13 Geographic Sharing in C-band Page 13 In addition, with mobile network base stations deployed below building height there could be significant antenna discrimination for a radio path heading over buildings at high elevation angles. (see Figure 7-3, for an illustration of the direct path vs the radio path in urban environments) The size of the interference zone could be reduced by employing mitigations at the Earth station such as using a larger antenna or site shielding. It was in general harder to use such mitigations for fixed link receivers. Even with the use of mitigations, there would be some geometries where it would not be feasible to deploy mobile network base stations. In particular, it would be problematic to deploy on streets orientated along azimuths that point directly at fixed link receiver or Earth station receivers, as there would be no clutter loss in that direction. 1.3 Recommendations for Additional Study This project has highlighted areas that could potentially improve the situation further and would benefit from further study, including: Short Term Thresholds: The analyses were run using the longterm interference criteria used by Ofcom in their frequency assignment/coordination procedures. For small cells, lower power base stations resulting in relatively small separation distances mean this is a good measure. However, for macro cells with higher power levels and greater antenna heights, larger separation distance will become more important. As a result anomalous propagation mechanisms will become significant. It would be beneficial to consider the short-term interference threshold in future analysis. Analysis: For the UK-wide maps, it would be useful to extend the detailed analysis for more scenarios and use smaller pixels for a more detailed assessment. We have seen that this can give a different picture and that better data usually leads to improved decision making Propagation modelling: Significant differences were noted between the local clutter loss model in P.452 and the same calculation using diffraction over local buildings. There would be significant benefits in further study, potentially including measurement of diffraction loss in dense urban areas. Net Filter Discrimination: the NFD for the satellite Earth station case was low due to conservative assumptions about the receiver filter characteristic in the absence of real data. On the other hand using standardised spectrum masks for fixed links showed that adjacent channel interference should not be neglected. There would be benefits in studying this further. International support: It would be useful to build international support for a more detailed approach that takes into account actual assignments and high-resolution surface databases.

14 Geographic Sharing in C-band Page 14 Pricing: The study raised issues regarding the pricing of licences and recognised spectrum access (RSA) for assignments in this band. Given that this frequency band could be used for mobile applications, the area denied and area type (urban / dense urban) could be the basis of a revised pricing policy. Regulatory options: One option would be to incorporate in mobile spectrum licences the need to protect existing assignments to a given I/N threshold. The implications of this could be analysed further with feedback from the operators on the operational constraints involved. Mitigations: A number of possible mitigations were identified that would require changes to Earth station and fixed link receivers. These could be analysed further with feedback from the licence holders as to feasibility and cost. 1.4 Document Structure This document is structured as follows: Section 2: Introduces the project, covering its objectives, purposes and scope. There is a description of the background, in particular the ITU-R JTG and the work undertaken is described. Section 3: This describes assignment data and modelling parameters used in the project, including, most importantly, the protection criteria. It also describes the licensing data provided by Ofcom. Section 4: This describes the UK-wide maps of spectrum availability including statistics by land use code and population, by band and existing licence type (fixed link or satellite Earth station). Section 5: This section summarises the Interference Zone Analysis undertaken around selected fixed link and Earth station locations. Section 6: Discusses technical issues relating to non co-frequency operation and how this impacts the results from Section 5. Section 7: High Resolution Analysis, in which a 3m resolution surface database to examine in detail deployment of mobile networks, is summarised. Section 8: This discusses possible mitigations to reduce the size of the excluded areas and hence increase the spectrum available. Section 9: This sets out project conclusions and recommendations. Sections 10 to 14: These annexes give the technical details of the analysis undertaken and its outputs.

15 Geographic Sharing in C-band Page 15 2 INTRODUCTION AND BACKGROUND 2.1 Purpose and Scope The study aimed to develop an assessment of the potential for spectrum sharing at C-band (3.6 GHz to 4.2 GHz) between incumbent users and mobile services. In the UK the frequencies are used for satellite downlinks, fixed links and 2 x 84 MHz is licensed to UK Broadband. The frequencies licensed to UK Broadband; 3605 to 3689 MHz and 3925 to 4009 MHz, were not considered in this study. The mobile services considered are those generally labelled IMT- A. Within this range, the GHz portion is considered separately from the GHz portion. There is a European Decision supporting harmonised use of GHz for mobile applications, and there is support within the European Conference of Postal and Telecommunications Administrations (CEPT) for a primary mobile service allocation and identification of this band for IMT. This is in contrast to the GHz band, where there is active opposition to a co-primary allocation to mobile and the band is not harmonised for mobile use. IMT-A is a broad definition and the two main candidate implementations are capable of deploying networks with different topologies, physical characteristics and radiofrequency parameters. For sharing studies, this range of possibilities has been codified by ITU-R WP 5D into a limited set of options covering macro-cell in urban and suburban environments and small-cell indoor and outdoor deployments. In this study, we have focussed on small cells because the C-band is considered ideal for providing additional, high capacity service to multiple users in hotspot areas and areas of dense population. (see Section 3.5 for definitions and discussion) Small cells support this type of use in a number of ways, including deployment below building level, which means the local environment aids frequency reuse and enhances capacity. The analysis undertaken in this study was driven by three important considerations: It was based on a real world and practicable perspective It was focussed on protecting incumbent users in a live band, based on UK national planning criteria It was required to explore ways to maximise the spectrum available for mobile broadband. Recognising also that harmonised use of spectrum brings many benefits, the study keeps in mind that the outputs are also intended to feed into discussions in Europe. With this in mind, the next section discusses the international background and summarises the current situation.

16 Geographic Sharing in C-band Page International Background The ITU-R has studied the issue of additional spectrum for IMT services for many years. Studies have been focussed on the protection of incumbents and, although conducted using a range of inputs and assumptions, they have aimed to deliver definitive separation distances required to protect satellite Earth stations. In general, the inputs to these studies are defined by experts associated with the incumbent services and there is little scope to challenge these. Hence, the studies are weighted in favour of the incumbents and, despite the wide range of outputs available, the ITU does not provide practical guidance to those administrations with a progressive approach to the study work and seeking opportunities for spectrum sharing. The studies presented here address the fact that there is a large grey area and no definitive answers. Sharing is possible in some situation but not others and more data helps in informed decision making.

17 Geographic Sharing in C-band Page 17 3 ASSIGNMENT DATA AND MODELLING PARAMETERS This section describes: A derivation of the interference criteria we have used in the project The parameters used by JTG studies 3.1 Protection Criteria This study was focused on planning at a UK-wide level with an emphasis on understanding the constraints on sharing spectrum. The protection criteria discussed here was taken from the frequency assignment and coordination procedures used by Ofcom when running: Frequency assignment requests for microwave fixed links; Frequency coordination procedures for satellite Earth stations. Practical decisions have been made by Ofcom in the past that mean satellite Earth stations and fixed links are treated on equal terms in these procedures. That is, the intra-service frequency assignment criteria designed to protect a microwave fixed link receiver from excess interference sourced from another microwave link is also used to protect the receiver from excess interference sourced from an Earth station. Our use of the criteria follows this precedent and is applied to the spectrum sharing problem considered in this study. For both Earth stations and fixed links, the protection criteria used in this study is expressed as the ratio of interfering signal power to noise denoted by I/N expressed in decibels and illustrated in Figure 3-1. N (dbw) I/N (db) I (dbw) Figure 3-1: Application of Protection Criterion The study has considered long-term interference tests only. The protection criteria used in the simulation work for both satellite Earth stations and fixed links are summarised in Table 3-1. Service Earth Station Fixed Link Test I / N 10 db, 20% time I / N db, 50% time Table 3-1: Protection Criteria Here I represents the single-entry interferer. In the Ofcom criteria, this is defined as a source of interference from one radio station on one frequency. However, in this sharing study we define single-entry interference to be the interference sourced from a single location (pixel) where the entire bandwidth of the victim receiver is populated with interferers.

18 Geographic Sharing in C-band Page Earth Stations For the long-term interference test considered in these studies, I/N = -10 db; here we model the interfering signal I exceeded for 20% of time. N is a calculated value using ktb (dbw) where k = dbw/hz/k is Boltzmann s Constant, T is noise temperature specified by the satellite Earth station operator (source: Ofcom data) expressed in Kelvin and B is the bandwidth of the victim receiver. N varies across the Ofcom data set used in this study Fixed Links Frequency assignment procedures for microwave fixed links are specified in Ofcom s OfW446 Technical Frequency Assignment Criteria for Fixed Point-to- Point Radio Services with Digital Modulation 1. This document sets out the Wanted-to-Unwanted ratios, denoted by W/U and used by Ofcom in its frequency assignment procedures. W/U are derived for individual radio systems. For the long-term interference test, the wanted signal is modelled fully faded (receiver sensitivity) and the unwanted signal at its median level (exceeded for 50% of time). Figure 3-2 shows the noise-interference budget used to derive W/U for the long-term test. Here R med is the receiver s median signal level, M is the fade margin, R ref is receiver sensitivity, C/(N+I) is the carrier to noise plus aggregate interference ratio, M I is the interference margin, N is total noise, I is the aggregate interference threshold, L + NF are fixed system losses and noise figure, I is the single entry interference threshold and ktb is the inescapable thermal noise level in the receiver s bandwidth. Figure 3-2: Noise-interference Budget 1

19 Geographic Sharing in C-band Page 19 In this study, we map W/U to an I/N value on the following basis. Ofcom, employ a 1 db interference margin in this frequency band which means that for the long-term interference test, the aggregate interference threshold is 5.9 db below noise and, assuming four equal single-entry interferers contributing to aggregate interference, the single-entry interference threshold is 6 db below the aggregate threshold, giving an I/N = db. However, we should note that the W/U are integers, rounded with a non-conservative bias and, in order to remain consistent with this practical approach, we obtain I/N = db for the fixed link radio system considered in this study. An analysis of the fixed link data allows for a mapping of each fixed link in the data to the radio system 140/155 in 30 (Mbit/s in MHz). From Ofw446, this system has R ref = -97 dbw, N = dbw and W/U = 37 db. Therefore, Equation 3-1 is used to obtain an I/N consistent with the Ofcom criterion: I / N N ( R W / U ). med Equation 3-1: Calculating Protection Criterion 3.2 Baseline IMT Parameters Our baseline parameters for the IMT base stations are given in Table 3-2, extracted from document JTG which is the basis for a Draft New ITU-R Report. 2 See Annex 17 from the Chairman s report of the final meeting, dated 19 August 2014 of

20 Geographic Sharing in C-band Page 20 Base station characteristics / Cell structure Cell radius / Deployment density Small cell outdoor 1-3 per urban macro cell (<1 per suburban macro site) Small cell indoor depending on indoor coverage/capacity demand Antenna height 6 m 3 m Sectorization single sector single sector Downtilt N/A N/A Frequency reuse 1 1 Antenna pattern Recommendation ITU-R F.1336 omni Antenna polarization linear linear Indoor base station deployment Indoor base station penetration loss Below rooftop base station antenna deployment Maximum base station output power (5/10/20 MHz) Maximum base station antenna gain Maximum base station output power/sector (EIRP) Average base station activity Average base station power/sector taking into account activity factor N/A 100% N/A 20 db 100% N/A 24 dbm 24 dbm 5 dbi 0 dbi 29 dbm 24 dbm 50% 50% 26 dbm 21 dbm Table 3-2: Deployment Related Parameters for IMT-Advanced Systems between 3 and 6 GHz Bands One point to note is that the maximum output power per base station is not dependent on the bandwidth. Our baseline results relate to a 10 MHz channel. We take the reference to Recommendation ITU-R F.1336 Omni to mean equations 11a et seq. in that Recommendation. The antenna pattern is illustrated in Figure 3-3.

21 Geographic Sharing in C-band Page 21 Figure 3-3: IMT Base Station Antenna Pattern A simple omnidirectional antenna was deployed for the indoor Base Station runs. 3.3 Satellite Earth Station Data A comprehensive set of satellite Earth station data, operating in the GHz band, was provided by Ofcom for the purposes of this study. 3.4 Fixed Link Data A comprehensive set of fixed link data, operating in the GHz band, was provided by Ofcom for the purposes of this study. 3.5 Geotype Data For the purpose of this study, we have considered five geotypes which are defined by population density thresholds following the Plum report for Huawei on The Economic Benefits of the use of C-band for Mobile Broadband in the UK 3. The geotypes are: Hotspots (13,910 or more persons per sq. km) Dense urban (between 10,910 and 13,910 persons per sq. km) Urban (between 4,290 and 10,910 persons per sq. km) Suburban (between 202 and 4,290 persons per sq. km) Rural (between 0 and 202 persons per sq. km) Our analysis of the pixelated population data gives the following figures in Table accessed 5th March 2015

22 Geographic Sharing in C-band Page 22 Geotype Area sq. km Population Rural ,732,733 Suburban ,647,678 Urban ,886,506 Dense Urban 90 1,118,787 Hotspot ,820 Table 3-3: Area and Population of each Geotype

23 Geographic Sharing in C-band Page 23 4 SUMMARY OF UK SPECTRUM AVAILABILITY In this Section, we report the key points of the calculation of the amount of spectrum available across the UK for hotspot, dense urban and urban geotypes. Data is presented as maps and tables and where appropriate, cumulative distribution functions (CDFs). The full data set is in Section Methodology The methodology is based on performing an I/N calculation against all carriers in the satellite Earth station and fixed link databases on a grid of points covering the UK. For each grid point the calculation is performed every 5 MHz from the low end of the band (3.6 GHz) to the high end (4.2 GHz), excluding the UK Broadband licensed frequencies (a total of 168 MHz in 2 x 84 MHz pairs within this band). Therefore, the results presented in this report are for the total spectrum available to the human population to a resolution of 5 MHz rather than contiguous spectrum. If the calculation at a point in a given 5 MHz is below the threshold for all carriers in the database then that 5 MHz is considered usable at that location. A count of the number of available 5 MHz slots (N) is kept for each point. Therefore 5*N is considered to be the spectrum available at that location in MHz. The calculation is performed for the whole GHz range and separately for the GHz and GHz bands. The calculation is performed considering sharing with satellite Earth stations and fixed links together and each separately. The output data grid is calculated using, as input, the results of the interference calculation which create the matrix: Where: And: 0 i N p is the pixel index 0 j N c is the channel index μ(i, j) μ(i, j) = 1 if the channel is available at that pixel μ(i, j) = 0 if the channel is not available at that pixel The channel would be available at a pixel if the worst single entry I/N was below the threshold for all potential victim systems for that channel: T ( I N ) > I N (RX l) Furthermore, each pixel would have a geotype code identified by: λ(i) The number of pixels having a particular geotype k would be N (k).

24 Geographic Sharing in C-band Page 24 The number of channels available at each pixel would then be: j=n c N a (i) = μ(i, j) j=1 Then the percentage of pixels of a particular geotype code k having N channels available would be: And: i=n p P(k, N) = 100 N λ (k) η(i) η(i) = 1 if λ(i) = k and N = N a (i) η(i) = 0 otherwise Once this grid of data has been generated, it can be processed in several ways plotted on a map, analysed against pixel geotype, related to population covered etc Run Variations In our detailed analysis, we have found a number of mitigation and calculation options that could effectively reduce the interference levels. For example, additional losses due to accurate local clutter modelling, large antenna elevations and indoor losses could all be legitimately added. These can all be modelled as single values in the interference zone simulations and there are many possible values and combinations. Rather than explicitly model each possible combination, the approach we have taken is to reproduce the spectrum availability grid based on relaxing the threshold by 5, 10 and 20 db. This is an established practice in satellite Earth station coordination, where the generation of auxiliary contours is common Interpretation of the Maps, Tables and Statistical Distribution The data in this section is presented as colour-coded UK-wide maps, as tables of spectrum bandwidth available vs population and as cumulative distribution of bandwidth available vs population. The colour-coded maps each have a key that looks like the following figure: i=1 The map is divided into pixels that reflect the key from the example above, red pixels have 0 MHz of available spectrum, the darkest blue between MHz etc.

25 Geographic Sharing in C-band Page 25 It is important to note that this scale of spectrum available may vary from map to map based on the size of the spectrum band being considered, for example if we are considering the GHz band, the scale will not be the same as for the GHz band. The tables have a simple interpretation. An example for the GHz band is given below: Spectrum Available (MHz) Population Population Percentage In this example, a total population of million has been considered (this is the sum of urban, dense urban and hotspot areas). Of that relevant population, it is possible to see how many have X MHz of bandwidth available. In the example around 951,509 have access to 40 MHz, million have access to 115 MHz, which is the full band in this case. The cumulative distribution function curves are presented as at least Y MHz of spectrum available vs. percentage of relevant population. In all the cases below, the relevant population is the million people in the urban, dense urban and hotspot areas. An example CDF is shown in Figure 4-1 below:

26 Geographic Sharing in C-band Page 26 Figure 4-1: Example Cumulative Distribution Function Each CDF allows for the comparison of different scenarios on the same picture and most of the plots show four cases as described in above. To understand the information in the curves consider the following interpretation: For a point on the X-axis (X %) read up to the curve and across the Y axis (Y MHz). This means X % of the population have access to at least Y MHz of spectrum. The power in the graphical presentation is in the comparison of the different assumptions corresponding to the four coloured curves. Blue is the baseline case and the others are the progressive improvements by 5, 10 and 20 db. Suppose the economic viability of the band for mobile use requires a minimum amount of spectrum to be available. You can read this number on the Y-axis across to each curve in turn and down to X axis to see what percent of the relevant population is denied access to this amount under the different assumptions. The rest of this section gives a summary of the main results and a commentary on the implications of these. 4.2 Available Spectrum Any mobile network will be constrained by the necessity to co-exist with both Earth stations and fixed links. Therefore, the most important results are those that include both services as potential victims in the calculation and these are the results we concentrate on here.

27 Geographic Sharing in C-band Page 27 The results are presented for the whole range GHz and for GHz and GHz separately. In all cases the UK Broadband licensed frequencies are excluded. The complete set of results are given in Section Analysis of GHz The total amount of spectrum available in the whole C-band, excluding the UK Broadband licensed frequencies, is 430 MHz. The maps below show how much of this is available across the UK all geotypes are displayed on the maps (i.e. urban, suburban and rural). Red pixels are absent from the maps showing there are no areas where all spectrum is denied. Figure 4-2: Baseline Spectrum Availability: Frequency Range is GHz, Sharing with Satellite Earth Stations and Fixed Links Large parts of the populated areas of the south east UK, particularly London and areas to the east and west of London, have between 5-75 MHz of spectrum available. This is a useful amount of spectrum but means over 80% of the total available spectrum is denied at these locations.

28 Geographic Sharing in C-band Page 28 Figure 4-3: Baseline Spectrum Availability: Frequency Range is GHz, Sharing with Satellite Earth Stations and Fixed Links Figure 4-4: Baseline Spectrum Availability: Frequency Range is GHz, Sharing with Satellite Earth Stations and Fixed Links

29 Geographic Sharing in C-band Page 29 Figure 4-3 and Figure 4-4 shows that in other parts of the country 300+ MHz of spectrum is generally available, except in the north east of Scotland, where fixed links connecting the islands mean there is less than 300 MHz in some locations. As discussed above the maps relate to the whole of the UK and all pixels have been considered, independent of geotype. It is also useful to consider how the spectrum available relates to populations in the most populated parts of the country. Figure 4-5 below shows this. Figure 4-5: Availability of Spectrum in MHz, in the Range GHz by Percentage of Urban+ Population under Baseline Assumptions Keeping in mind that the spectrum available is counted in 5 MHz chunks, we can see the following from the figure: For 60% of the urban, dense urban and hotspot population there is just under 300 MHz of spectrum available. For 80% this falls to 65 MHz. Analysis of the underlying numbers shows that the amount of spectrum available falls rapidly for 97%+ of the urban+ population. 97% having access to 40 MHz but 98% having access to less than 10 MHz Analysis of GHz The pattern of available spectrum is broadly repeated in the GHz range as shown in the figures below. We see some difference in the overall picture but in the most populous areas the underlying numbers are similar.

30 Geographic Sharing in C-band Page 30 Note that the range in the colour scale is different here, because there is only 115 MHz of spectrum available. However, the colours represent approximately the same percentage of spectrum as in the previous section. Figure 4-6: Baseline Spectrum Availability: Frequency Range is GHz, Sharing with Satellite Earth Stations and Fixed Links

31 Geographic Sharing in C-band Page 31 Figure 4-7: Baseline Spectrum Availability: Frequency Range is GHz, Sharing with Satellite Earth Stations and Fixed Links Figure 4-8: Baseline Spectrum Availability: Frequency Range is GHz, Sharing with Satellite Earth Stations and Fixed Links The maps relate to the whole of the UK across all geotypes. It is useful to consider how the spectrum available relates to populations in the most populated parts of the country. The CDF at Figure 4-9 below shows this.

32 Geographic Sharing in C-band Page 32 Figure 4-9: Availability of Spectrum in MHz, in the Range GHz by Percentage of Urban+ Population under Baseline Assumptions Keeping in mind that the spectrum available is counted in 5 MHz chunks, we can see the following from the figure: For 60% of the urban, dense urban and hotspot population there is around 85 MHz of spectrum available in this part of the band. This falls rapidly between 75-80%, 76% having access to 40 MHz but 79% having access to 15 MHz Analysis of MHz The pattern of available spectrum is broadly repeated in the GHz range as shown in the figures below. We see some difference in the overall picture but in the most populous areas the underlying numbers are similar. Note that the colour scale is different here, because there is only 315 MHz of spectrum available. However, the colours represent approximately the same percentage of spectrum as in the previous section.

33 Geographic Sharing in C-band Page 33 Figure 4-10: Baseline Spectrum Availability: Frequency Range is GHz, Sharing with Satellite Earth Stations and Fixed Links One thing to note in this plot is that there are areas with red pixels close to satellite Earth stations. Recall that the red pixel means 0 MHz of spectrum available in the MHz band.

34 Geographic Sharing in C-band Page 34 Figure 4-11: Baseline Spectrum Availability: Frequency Range is GHz, Sharing with Satellite Earth Stations and Fixed Links Figure 4-12: Baseline Spectrum Availability: Frequency Range is GHz, Sharing with Satellite Earth Stations and Fixed Links The maps relate to the whole of the UK across all geotypes. It is useful to consider how the spectrum available relates to populations in the most populated parts of the country. The CDF at Figure 4-13 below shows this.

35 Geographic Sharing in C-band Page 35 Figure 4-13: Availability of Spectrum in MHz in the Range GHz by Percentage of Urban+ Population under Baseline Assumptions Keeping in mind that the spectrum available is counted in 5 MHz chunks, we can see the following from the figure: For 60% of the urban, dense urban and hotspot population there is around 205 MHz of spectrum available in this part of the band. At 80% of the urban+ population this has fallen to 60 MHz. This falls rapidly above 97%. 97% having access to 35 MHz but 98% having access to only 5 MHz in this part of the band Impact of Mitigations Various mitigation techniques have been identified in this study. Whilst many of these are site specific, it is still useful to see the effect that a generally applied improvement would bring. We have modelled this by changing the I/N threshold in the simulation methodology by 5 db, 10 db and 20 db. Graphically we can see this in Figure 4-14 below:

36 Geographic Sharing in C-band Page 36 Baseline Case 5 db Relaxation 10 db Relaxation 20 db Relaxation Figure 4-14: The Available Spectrum in London area, in the Baseline Case and with 5, 10 and 20 db of mitigation, in GHz Sharing will all Satellite Earth Station and Fixed Link Carriers The pink pixels have less than 75 MHz of spectrum available. Light blue pixels represent areas that have more than 230 MHz available (more than half the available spectrum). Green pixels have access to the entire spectrum.

37 Geographic Sharing in C-band Page 37 A key feature of this graph is that even with 20 db of mitigation there are areas in a relatively narrow band that are pink. This reflects the impact of the fixed links, as discussed in Section , which provide low latency, high data rate connections across London. The CDF at Figure 4-15 below summarise the statistics of spectrum availability under baseline assumptions and relaxations by 5, 10 and 20 db, showing the impact of mitigations on the spectrum available per head of urban+ population. Figure 4-15: Availability of Spectrum in MHz, in the range GHz by Percentage of Urban+ Population under Baseline Assumptions, 5, 10 and 20 db Relaxations 4.3 Separate Analysis of Satellite Earth Stations and Fixed Links Section 13 contains full details of the spectrum that would be available if sharing with only the satellite Earth stations or fixed links. The analysis shows that sharing with fixed links results in the largest constraints on the spectrum. It is apparent that, whilst a satellite Earth station may experience interference from a large area this has a much smaller impact on practical spectrum availability than the fixed links we have modelled. This is partly because the fixed link receivers tend to be higher up, with low elevation main beams. Another factor is that the fixed links are in the densely populated areas where the small cell mobile networks are most likely to be deployed. The dominant role of the fixed links masks the fact that the satellite Earth station use of the band is only a minor constraint on mobile usage in urban areas.

38 Geographic Sharing in C-band Page Conclusions The following table summarises the spectrum available. The principal conclusion is that there is scope for sharing spectrum in this band. The reason we believe this is that, as the table shows, even in the baseline case, 80% of the GHz spectrum band is available to >50% of the urban+ population which represents a massive potential economic value. Half the existing spectrum is available to 65% of the urban+ population, and this increases to 88% if 20 db of mitigation is applied. However, even on the application of 20 db of mitigation, potential interference cannot be ruled out for all frequencies and all geographic locations. The implication, therefore, is that a managed approach to shared access of the band on a geographic basis has great merit, and should be able help facilitate spectrum sharing. Percent of total spectrum in GHz (430 MHz) % of urban+ population this is available to Baseline case 5 db relaxation 10 db relaxation 20 db relaxation 20% 79% 83% 91% 99% 50% 65% 68% 75% 88% 80% 52% 55% 60% 75% 100% 45% 48% 52% 64% Table 4-1: Breakdown of Percentage of Urban+ Population that has access to X% of Spectrum The results show that the fixed links are a much larger constraint on the use of C-band for mobile than the satellite Earth stations are. Even under the baseline case, useful amounts of spectrum could still be used by large percentages of the urban+ population. The minimum useful amount of spectrum for IMT-A is considered to be 5 MHz,and in many populated areas there is significantly more spectrum available. Applying mitigations up to 20 db improves the situation, but areas of significant constraint on the mobile network remain our study shows a band across central London where 50% or more of the available spectrum is denied. This reflects the presence of low latency links on high towers across the most populous area of the South East. The same 20 db improvement would mean sharing of mobile services with satellite Earth stations also appears possible, though the same constraints in London would apply. No major difference between GHz and GHz ranges is seen in the analysis.

39 Geographic Sharing in C-band Page 39 5 INTERFERENCE ZONE ANALYSIS This section is a description of the work done looking at specific satellite Earth station and fixed link sites. In this section, we look at interference zones around selected satellite Earth station and fixed link sites. We explore how these zones are affected by different assumptions and potential mitigations. 5.1 Satellite Earth Station Interference Zone Analysis This part of the study considered Area Analyses for two carriers at the Chalfont Grove site and looked at the area south-east of the site, which includes greater London. The Chalfont site has a number of antennas including some operating at low elevation angle towards satellites in the east. The site has licensed carriers across a large part of the C-band. The area analysis splits a large area into small pixels and performs the interference calculation at each pixel. When the I/N threshold is exceeded, the pixel is considered to be part of the interference zone. The output is as a colour code map overlay that indicates that I/N would be exceeded, under stated assumptions. This type of analysis is commonly seen internationally in JTG and prior to that in Report As such, it provides us with a test bed for comparing JTG methods with our baseline and with the variations to the baseline that are supported by the detailed analysis in Section Baseline System Parameters The IMT parameters are described in Section 3.2. The licensing data for the site at Chalfont Grove contains 58 licensed carriers with differing parameters that will affect the calculation of the interference zone. Parameters that influence the I/N calculation are: 1. Antenna azimuth and elevation. Low elevation operation in the direction of a proposed IMT network will generate the worst case. If we are considering London then low elevations to the southeast will be the most representative case. 2. Antenna height above the local terrain. Generally, this is important because height provides clearance from local shielding. All antennas in the Chalfont Data are at the same height. 3. Antenna gain and sidelobe performance. In practice, the sidelobe gain is important. Even for the lowest elevation angles and for the smallest dishes the interferer will not impact the main lobe. All the Chalfont antennas are specified as per REC-580. Recommendation ITU-R S gives a sidelobe gain performance that is independent of peak gain. 4. Link noise. Lower noise means that there is a higher I/N for a given value of I. The lowest elevation links have a temperature of 79K and the next lowest are 60K.

40 Geographic Sharing in C-band Page 40 We would add a further assumption that the calculation is essentially independent of bandwidth and frequency. Based on these considerations we propose to model two sets of parameters for Chalfont as in Table 5-1 below: Parameter Carrier 1 Carrier 2 Antenna Elevation 8, Antenna Azimuth Antenna Height 10m 10m Peak Gain Sidelobe Performance Rec 580 Rec 580 Link Temperature 79 K 60 K Table 5-1: Modelling Parameters for Chalfont Earth Station The area analyses are configured such that ITU-R P is used to model losses on the interference path and predict the interfering signal power of the median interferer incident to the fixed link receiver. The Ofcom terrain and clutter database are deployed. The baseline case uses 50m terrain data and a complementary clutter database Simulations The simulations aim to show the size of the interference zone under a number of assumptions. These can be categorised as improvements on the baseline and degradations from the baseline as discussed here: Section 7 presents detailed analysis using very high fidelity modelling of the local environment, looks at aggregation and suggests that our baseline area analysis modelling could justifiably be modified. Our baseline uses detailed data that are not always available, for example to other administrations. So JTG has presented a variety of other analyses and defines a generic case. It is very useful to compare this to our results. Baseline runs have been made for the small cell indoor and outdoor cases, run variations concentrate on outdoor cases only. So we have performed simulations (Table 5-2) ranging from the non-sitespecific to baseline case and to baseline modified to account for dense urban geometries, power management and traffic levels.

41 Geographic Sharing in C-band Page 41 Case 1 Case 2 Case 3 Case 4 Case 5 JTG non site specific case - no terrain data, no clutter losses Terrain but no clutter loss Our baseline case Our baseline adjusted to reflect extra losses from dense urban cells and aggregation Table 7-1: Diffraction vs. Clutter Loss gives a baseline additional loss of 17.3 db but Section 7 indicates that aggregation from dense small cells could be at least 10.9 db. This gives a net advantage of 6.4 db. Case 4 with additional reductions when considering power, polarisation and traffic. Section 7.5 suggests an additional 3.8 db advantage Table 5-2: Five cases studied The modification we introduce in Case 4 reflects the average values found in Section Outdoor Small Cells Figure 5-1 below shows comparative results obtained around the Chalfont site, which is marked with the black star. Case 1 - smooth earth case (shown by the black contour). All points inside the contour are above the interference threshold and hence spectrum cannot be used by IMT. All points outside the contour are below the threshold Case 2 - with terrain database but no local clutter loss (shown as red pixels). All red pixels are above the interference threshold. Case 3 - with terrain and local clutter (baseline case shown as blue pixels). All blue pixels are above the interference threshold. It should be noted that there are many red pixels covered by blue pixels close the Chalfont site.

42 Geographic Sharing in C-band Page 42 Figure 5-1: Comparison Interference Zones of the Baseline Case, Case with Clutter Loss Removed and JTG Non Site-Specific Case i.e. Smooth Earth. Table 5-3 shows the largest separation distance and the area of denied pixels for the three cases. Case Largest separation distance km Area of denied pixels 4 km2 Case Case Case 3 (baseline) Table 5-3: Separation Distance and Area of Denied Pixels for Cases 1, 2 and 3 Some interesting points to note are: 1. The JTG generic case is not necessarily the worst case. When terrain is added, the effect of higher ground can be clearly seen. 2. Inclusion of a clutter database does not have a great effect on the maximum separation distance, but greatly reduces the total area of denied pixels. 4 Note this calculated area is presented for comparative purposes only. It is a function of the size of the Area Analysis which is arbitrary.

43 Geographic Sharing in C-band Page 43 Figure 5-2 shows the additional impact of Case 4 (shown in yellow) and Case 5 (shown in red), which take a simplified account of the findings of our detailed analysis. Note that there are yellow pixels beneath all the red pixels, due to the additional improvement introduced in Case 5 Case Figure 5-2: Additional Impact of Case 4 and Case 5. Largest separation distance km Area of denied pixels 5 km2 Case 3 (baseline) Case Case Table 5-4: Separation Distance and Area of Denied Pixels) for Cases 3, 4 and 5 Note that when the additional advantages of cases 4 and 5 are taken into account both the largest separation distance and the area of denied pixels decreases Indoor Small Cells Our baseline indoor parameters introduce 25 db of extra advantage due to the lower power operation and a 20 db indoor/outdoor loss. Lower antenna height assumptions also have a small effect. 5 Note this area is presented for comparative purposes only. It is a function of the size of the Area Analysis which is arbitrary.

44 Geographic Sharing in C-band Page 44 When we simulate this against the two carriers at Chalfont Grove we find a separation distance of 3 km but very few affected pixels. Even the local area is essentially unaffected as can be seen from Figure 5-3 the interference zone around Chalfont Grove, for baseline indoor small cell case. The Blue pixels show the affected areas, around the Chalfont site and a further affected point around 3 km to the south east. Figure 5-3 Interference Zone around Chalfont Grove for Baseline Indoor Small Cell Case Conclusions from Satellite Earth Station Interference Zones The interference zone analysis provides a quick way to get an overview of a local situation that may be useful in coordinating a specific site or just as a way to get a high level feel for the problem. It is also a good tool for testing variations in parameters and mitigations. We have found the following to be the case: The non-site specific or generic case used in many JTG studies is not always the worst case. In some sense, this is the furthest from a generic study that we could get as almost no locations can be accurately modelled as smooth Earth. The value of generic studies to the UK context is very limited. In the international context studies that use no terrain, or use terrain without a clutter database, or studies that result in a recommended minimum separation distance are also of limited value. We find that whilst the use of clutter data does not reduce the maximum required separation very much, it does dramatically reduce the area of the interference zone. When taking into account in a simple way the results from our detailed analysis in Section 7, the interference zone shrinks dramatically.

45 Geographic Sharing in C-band Page 45 For indoor small cells, under the baseline assumptions, the interference zone around Chalfont is very small the maximum separation required is 3km, although there are plenty of locations closer to the site that show no problems. 5.2 Fixed Links Interference Zone Analysis This study considered Area Analyses for a sub-set of the fixed links data. Sixteen receivers were selected in the London area. Each receiver was considered in isolation and subject to two simulations where it was exposed to interference from outdoor and indoor small cell base stations. Each pixel of 0.25 km 2 in the area analysis is considered to be a potential location for base stations, therefore each pixel is modelled as an interference source and the interfering signal power incident to the fixed link receiver is calculated. The entire bandwidth of the 30 MHz fixed link receiver is populated with 10 MHz IMT interferers. Hence, in these analyses, single-entry interference is defined as multiple interfering signals sourced from a single location Simulations Table 5-5 shows the schedule of simulation runs for fixed links.

46 Geographic Sharing in C-band Page 46 Station name Request Runs BT Tower /1 BT Tower /1 Hillingdon /1 Hillingdon /1 Hillingdon /1 Hillingdon /1 LSE /1 LSE /1 LSE /1 Royal Free /1 Royal Free /1 Royal Free /1 Royal Free /1 Wanstead /1 Wanstead /1 Wanstead /1 Outdoor base station, Indoor base station Outdoor base station, Indoor base station Outdoor base station, Indoor base station Outdoor base station, Indoor base station Outdoor base station, Indoor base station Outdoor base station, Indoor base station Outdoor base station, Indoor base station Outdoor base station, Indoor base station Outdoor base station, Indoor base station Outdoor base station, Indoor base station Outdoor base station, Indoor base station Outdoor base station, Indoor base station Outdoor base station, Indoor base station Outdoor base station, Indoor base station Outdoor base station, Indoor base station Outdoor base station, Indoor base station Table 5-5: Run Schedule: Fixed Links Area Analyses The area analyses were configured such that ITU-R P was used to model losses on the interference path and predict the interfering signal power of the median interferer incident to the fixed link receiver. The Ofcom terrain and clutter database were deployed Example Simulation: Outdoor Base Station Figure 5-4 shows the results of an area analysis for request /1 (BT Tower) considering interference from an outdoor IMT Base Station. For the outdoor Base Station run, the area where denied pixels (red) are present corresponds to a representation of the fixed link antenna pattern. If we

47 Geographic Sharing in C-band Page 47 switch off the terrain and clutter database as shown in Figure 5-5 the pattern is more comprehensive, suggesting a frequency coordination zone reminiscent of the classical keyhole coordination procedures. This is also a useful illustration of the impact of terrain and clutter data on the modelling and on frequency coordination in general. Figure 5-4: Area Analysis for Request /1 (BT Tower, London): Outdoor Base Station (Denied Area = 511 km 2 ) Figure 5-5: Area Analysis for Request /1 (BT Tower, London): Outdoor Base Station, Smooth Earth (Denied Area = km 2 ) With terrain and clutter included in the analysis, 2044 pixels of 0.25 km 2 and a total area of 511 km 2 are denied to the IMT base station; that is, interfering signal power exceeds the protection criterion for the fixed link receiver when the base station is located in one of these pixels. With terrain and clutter switched off, 3007 pixels and a total area of km 2 are denied. For this example, the area populated by the denied pixels extends around 50 km east from the fixed link antenna and the widest dimension of the denied area is around 32 km using terrain plus clutter and around 27 km with these features switched off. Although, in general, the use of terrain and clutter will

48 Geographic Sharing in C-band Page 48 reduce the count of denied pixels, use of this data will sometimes result in a more potent model of the radio interference path Example Simulation: Indoor Base Station Figure 5-6 shows the results of an area analysis for request /1 (BT Tower, London) considering the indoor IMT Base Station. In this case, the interference is attenuated by 20 db to account for building loss (indooroutdoor). Figure 5-6: Area Analysis for Request /1 (BT Tower, London): Indoor Base Station (Denied Area = 27.5 km 2 ) Here, the denied area is reduced to 27.5 km 2 when the IMT Base Station is indoors with 110 denied pixels located in the main beam of the fixed link antenna i.e. at locations where the interfering signal is subject to the maximum antenna gain available from the fixed link Set of Visualyse Simulations The area analyses are a useful preliminary investigation into the sharing problem. The results are very similar for all of the fixed link receivers investigated and they all show, very clearly, that there are very significant constraints on sharing between fixed links and mobile base stations when the two services are operating co-frequency. These constraints are greatly reduced but still significant when the mobile base station is located indoors and a reasonable assumption made with regard to building loss. The entire set of fixed link area analyses simulation files are designated as a project deliverable.

49 Geographic Sharing in C-band Page 49 6 ADJACENT CHANNEL INTERFERENCE ISSUES This section reports on our investigations into adjacent channel interference. Some supplementary technical information on net filter discrimination calculations and Out-of-Band spectrum masks is provided in Section Adjacent Channel Interference into Fixed Links Net Filter Discrimination (NFD) can be defined as the advantage obtained on the radio interference path when the interferer is offset in frequency from the victim receiver, relative to a scenario where victim and interferer are cofrequency. NFD is expressed in db and can be subtracted from the interferer s power at a suitable point on the radio interference path. The calculation of NFD requires spectrum masks for the transmitter and the receiver. ETSI TR sets out a well-established method for calculating NFD where spectrum masks are convolved in frequency; this method is used by Ofcom in its frequency assignment and frequency coordination work. Spectrum masks for fixed links are easily available (although the technical standards can be complex and difficult to interpret) and a basic IMT mask can be modelled based on discussions in international fora. Spectrum masks for satellite Earth station receivers are much harder to acquire. Simple default masks are used in professional practice including by Ofcom. We have derived the results of some NFD calculations using a Gaussian curve to represent the satellite Earth station receiver mask NFD Calculations Convolving the Out-Of-Band (OOB) spectrum masks and exercising Equation 11-1 at each discrete frequency offset between carrier frequencies, we obtain a graph of the NFD available; this is shown in Figure 6-1. Figure 6-1: Net Filter Discrimination We consider a specific interference scenario where two IMT interferers, operating in 10 MHz bandwidth, are tuned such that they operate in the first 10 MHz channel adjacent to the fixed link receiver bandwidth, one either side of the fixed link s 30 MHz receiver channel.

50 Geographic Sharing in C-band Page 50 This scenario results in a 20 MHz frequency offset between the fixed link carrier and each of the IMT carriers. Using the method described here, we obtain db of NFD. The NFD can be applied at a convenient point on our model of the radio interference path and in these calculations, we make adjustments at the source of interference by reducing the IMT transmitter (TX) power. Table 6-1 illustrates calculations for aggregate transmitter power from the two IMT adjacent channel interferers in 1 MHz of bandwidth. Description Value Unit TX power in 10 MHz 24 dbm NFD for 20 MHz frequency offset db TX power in 1 MHz dbm Activity Factor 3 db Adjusted TX power in 1 MHz dbm Adjusted TX power in 1 MHz dbw Two adjacent channel interferers dbw Table 6-1: NFD and Adjusted Transmitter Power Using the value dbw/mhz, we revisit some area analyses to assess whether adjacent channel interference exceeds the protection criterion at the fixed link receiver Analyses: Outdoor Base Stations Figure 6-2 shows an area analysis for request /1 (BT Tower, London) where the fixed link receiver is exposed to adjacent channel interference only. 164 pixels and a total area of 41 km 2 are denied to the IMT Base Station.

51 Geographic Sharing in C-band Page 51 Figure 6-2: Area Analysis for Request /1 (BT Tower, London): Outdoor Base Station Operating in Adjacent Channels (Denied Area = 41 km 2 ) A further example is given in Figure 6-3 where we expose request (Hillingdon, London) to adjacent channel interference. Here, 96 pixels and a total area of 24 km 2 are denied. Figure 6-3: Area analysis for Request (Hillingdon, London): Outdoor Base Station Operating in Adjacent Channels (Denied Area = 24 km 2 ) Both of these requests were considered again, modelling adjacent channel interference from IMT base stations located indoors. Figure 6-4 and Figure 6-5 show the results of the area analyses.

52 Geographic Sharing in C-band Page 52 Figure 6-4: Area Analysis for Request (BT Tower, London): Indoor Base Station Operating in Adjacent Channels (Denied Area = 0 km 2 ) Figure 6-5: Area Analysis for Request (Hillingdon, London): Indoor Base Station Operating in Adjacent Channels (Denied Area = 0.25 km 2 ) For request (BT Tower, London), no pixels are denied and for (Hillingdon, London), one pixel is and a total area of 0.25 km 2 are denied. We can see from these investigations that adjacent channel interference cannot be neglected and that there is potential for harmful interference incident to the fixed link receivers from the outdoor base station. Clearly, the results for the indoor base station is far more optimistic and, depending on the exact sharing objectives, some further study may be appropriate including consideration of alternative interference scenarios over a larger set of receivers.

53 Geographic Sharing in C-band Page NFD Calculations for Satellite Earth Station and IMT Interference This section has focused on a case study where NFD is applied to a scenario where the victim fixed link is exposed to adjacent channel interference (ACI) from IMT. Here we report on some calculations for NFD when a satellite Earth station receiver is victim to a 10 MHz IMT interferer. Spectrum masks for permanent earth station receivers are not easily available and it is common practice to use default spectrum masks. Here, we model a Gaussian curve extending two times the transponder bandwidth and attenuating to -30 db at the lower and upper bounds of the curve. Figure 6-6 shows a graph of this default permanent earth station mask Attenuation (db) Frequency offset from carrier frequency (MHz) Figure 6-6: Default Gaussian Spectrum Mask for Earth Station Using the default satellite Earth station mask in Figure 6-6 and the IMT mask shown in Figure 11-2, we calculate NFD over a range of frequency offsets. The results are presented in Figure 6-7. Here we see NFD calculated up to the point where the masks are overlapping by 1 MHz in the frequency domain; at this point around 95 db of NFD is available.

54 Geographic Sharing in C-band Page 54 Figure 6-7: NFD: 72 MHz Satellite Earth Station and 10 MHz IMT These results can be used to consider the adjacent channel interference problem. If we consider a similar scenario to that modelled for the fixed link case, where 10 MHz bandwidth mobile interferers are located in the first 10 MHz channel either side of the satellite Earth station bandwidth, then the frequency separation between satellite Earth station and mobile carriers is 41 MHz and 9.51 db of NFD is obtained. Use of a Gaussian default mask is likely to be conservative relative to a mask based on measurements or even one defined in the manufacturing standards. Returning to the fixed link case study, if we replace the ETSI OOB mask with a Gaussian curve then the NFD obtained for the specific scenario investigated is reduced to db (from db). Clearly, the NFD calculations are highly dependent on the inputs; that is, the spectrum masks. 6.3 Notes and Conclusions on NFD and ACI The case study examined here shows that adjacent channel interference cannot be neglected. A significant number of locations are denied when the IMT Base Station is outdoors. The indoor case appears to be favourable but a more comprehensive investigation could be considered. NFD calculations are dependent on the spectrum masks and when default masks are utilised, these tend to be conservative, leading to spectrum denial when spectrum access is possible. The Gaussian curve is a typical example of a convenient default spectrum mask. This is used by Ofcom in its frequency assignment and frequency coordination procedures to model the satellite Earth station emissions. However, it may be very difficult in practice to replace these with masks based on the real or standardised performance of the radios. In general, modern frequency assignment/coordination procedures will model adjacent channel interference in order that a candidate transmitter satisfies protection criteria at potential victim receivers that are co-frequency and in adjacent frequencies.

55 Geographic Sharing in C-band Page 55 7 SUMMARY OF HIGH RESOLUTION DETAILED ANALYSIS 7.1 Overview One of the key challenges in undertaking interference analysis is to balance the need to protect incumbent services without being over conservative by taking a series of worst-case assumptions. One way to facilitate sharing is to use models that are more detailed. This can reduce the number of assumptions and generalisations that are required. The following approaches to modelling propagation paths can be considered improvements due to the increasing level of detail: 1. Smooth Earth, no terrain or clutter 2. Use of a terrain database 3. Use of terrain and land use database 4. Use of a surface database The majority of the study was undertaken with the third of these, as this was the most detailed level information available on a UK-wide basis. However, for central London a higher resolution 3m surface database was used. This surface database allowed us to analyse a small number of cases in highresolution, considering the impact of the need to protect existing fixed links or satellite Earth stations on the ability of a mobile network operator to deploy small cell base stations in a dense urban environment. Further information on the high resolution analysis is given in Section Factors Specific to the High Resolution Analysis The basis of the high-resolution study was a selective qualitative analysis using a 3m resolution surface database within central London. We consider this representative of the type of high-density urban areas where mobile networks are likely to be deployed. From the assignment database provided by Ofcom the following were selected for high-resolution analysis: Earth station (ES): a very small aperture terminal (VSAT) in central London. Fixed links: BT Tower, with link connecting to the LSE. The analysis methodology in each case was as follows: Deployment: Calculate the number of small cell base stations within a circle centred on the Earth station or fixed link receive station, taking into account the ratio of small cells to macro cells [1, 2, 3]. Position each base station at random within the circle. Where necessary (e.g. random location results in base station on top of a building), move base station to nearest street.

56 Geographic Sharing in C-band Page 56 Analysis: Databases Make limited adjustments for location (e.g. avoid very low separation distances between base station and relocate some of the base stations that were deployed within parks). For the directional case, orientate antenna to point along the street. 1. Calculate aggregate I/N at Earth station or fixed link receive station from all base stations. 2. If the aggregate I/N > T[I/N] db then: a. Identify the worst single entry case b. Remove that base station from the group of interfering base stations c. Continue at Step 1 3. Output final base station deployment The analysis was based upon terrain and surface databases for the areas where satellite Earth stations are located. The following databases were available: Ofcom s standard 50 m terrain and 50 m land use code database High resolution 3 m surface database of central London The following figures show central London in 50 m terrain (Figure 7-1) and 3 m surface resolution (Figure 7-2). Even from these small pictures, it is clear that the 3 m data captures much more about the central London environment than the 50 m data. Streets and individual buildings are clearly visible and can be explicitly modelled.

57 Geographic Sharing in C-band Page 57 Figure 7-1: 50m Terrain Data of Central London Figure 7-2: 3m Surface Database of Central London

58 Geographic Sharing in C-band Page Propagation Models The databases described in the previous section were used as inputs into the propagation models, in particular Recommendation ITU-R P.452. This model includes a number of propagation modes that are selected based on an analysis of the path profile created from the terrain or surface database. The relevant terms are then calculated and merged mathematically to generate a path loss. A terrain database, as the name implies, represents the height of the terrain underlying the stations, built environment and vegetation that we are trying to simulate. If we use a terrain database, the model can then be extended to include clutter loss based on local clutter types. This requires a database, such as the 50 m data, which maps on to the clutter codes of the Recommendation as described in Section A surface database, on the other hand, represents the height of the terrain plus anything substantial that sits on top of the terrain. As mentioned above, a surface data of sufficiently high resolution will represent buildings and streets. If a surface database is used then it is not necessary to include the clutter loss as an extra term, it will be accounted for by P.452 path loss using the diffraction model applied to the local environment as additional obstacles on the path. While the two approaches (diffraction and local clutter loss) are based upon common concepts, they are implemented differently and this leads to significantly different results as can be seen in Table 7-1. Obstruction Close Baseline Far Frequency (GHz) Transmit antenna height (m) Obstruction height (m) Distance to obstruction (m) Distance from obstruction to receiver (m) Clutter loss (db) Diffraction loss (db) Table 7-1: Diffraction vs. Clutter Loss In particular, the clutter loss is capped at around 20 db while the diffraction loss can be much greater in the right hand column for the case where the transmitter is closer to the obstruction. As the land use database does not know the actual case involved, it must use simplifying assumptions and generate a typical value that might not be applicable for most actual deployments. Analysis presented in detail in Section suggests that the interference can become either: An aggregate of multiple interferers all significantly attenuated due to large diffraction losses

59 Geographic Sharing in C-band Page 59 Dominated by a single interferer that has significantly lower diffraction loss (e.g. due to street being aligned with the victim) Antenna Gains Another factor that will have a significant impact on the interference calculation is the gain at the transmit antenna, and this will vary depending upon the terrain or surface database used. If a terrain database is used then it is likely that for short paths the direct path is used to calculate the transmit gain, but if a surface database is used the radio path can be very different and hence the gain calculated also differs, as in Figure 7-3. Figure 7-3: Radio Path vs. Direct Path This would lead to large differences in the calculated aggregate interference level, particularly if the base station is modelled using measured data from a directional antenna. One output from the detailed analysis in Section 12 is the combined effect of using diffraction in place of clutter and of accounting for the elevation dependence of the base station Summary Two mechanisms were identified by which the aggregate interference calculated using a high-resolution surface database could be significantly different from that derived using terrain and land use codes: 1. The clutter loss calculation uses fixed parameters and is capped at around 20 db. 2. The transmit gain calculated using a terrain database is likely to use an inaccurate direct path. This suggests that the lower aggregate interference levels calculated using a surface database are based upon real phenomena and can be accepted for use in further analysis. 7.3 Earth Station Test Case Elements of the Analysis of the VSAT case The previous section identified benefits in analysing sharing between mobile network base stations and incumbent systems using a high resolution surface database. This section considers the impact on satellite Earth stations using a specific VSAT station on a roof in central London.

60 Geographic Sharing in C-band Page 60 The density of small cell outdoor base stations was derived from: Small cell outdoor: 2 per urban macro cell Urban macro cell every 0.6 km 50 small cell outdoor base stations were deployed over an area calculated from the above to be a circle of radius 1.69 km based upon the victim Earth station as shown in Table 7-2. Macro base station separation distance (km) 0.6 Number of Small Cell / Macro base station 2 Area / Macro base station (km 2 ) 0.36 Area / Small Cell (km 2 ) 0.18 Number of Small Cell in simulation 50 Area of Simulation (km 2 ) 9 Radius of deployment zone (km) 1.69 Table 7-2 : Deployment Density for IMT-Advanced Small Cell Outdoor Systems As described in Section 12 two mobile scenarios were modelled one using omni-directional antennas and one using measured base station performance. Locations of the 50 small cell base station were randomised within a circle of radius 1.69 km around the Earth station. They were then moved to the nearest street. For the measured base station data the antenna was pointed along the street. No attempt was made to analyse interference between base station or the coverage / service provided. The resulting initial deployment locations are shown in Figure 7-4.

61 Geographic Sharing in C-band Page 61 Figure 7-4: 50 Small Cell Base Stations Located within 1.69 km of Victim Earth Station in Central London A full description of the simulations and the analysis is given in Section 12 including: 1. How interfering base stations were identified and removed to give a final deployment consistent with the I/N threshold 2. Single entry and aggregate interference 3. Analysis with 1,2 or 3 small cells per macro cell 4. Operation in adjacent channel 5. Mitigation options and their effectiveness Summary Results The analysis suggests that the exclusion zone around an Earth station where a mobile small cell base station would not be permitted could be small, possibly in the range km 2 without mitigation.

62 Geographic Sharing in C-band Page 62 The small size of exclusion zone is due to the mobile network base stations operating on lower power below the clutter in a dense urban environment where there will be significant off-axis gain and diffraction loss. The zone can be reduced using various mitigation methods and could be as small as 0.02 km 2. One possible approach to facilitate shared use of this band by mobile networks would be to include in the licence terms and conditions the necessity of protecting a list of existing satellite Earth station. A suitable threshold would be either: Single entry I/N suitably adjusted for aggregation Aggregate I/N threshold The mobile operator could also offer to the Earth station operator: Site shielding around the Earth station; A larger replacement antenna with higher gain and/or lower far-off-axis gain values. Either of these methods would significantly reduce the average area excluded. Aggregation interference was observed to be between 9-11 db higher than single entry levels. The analysis was undertaken for 1, 2 and 3 small cell base stations per macro cell. Similar behaviour was observed in each case, though the omni antenna deployment appeared to be reaching an interference driven limit in the number of base stations that could be deployed. 7.4 BT Tower Test Case This section describes analysis interference into a fixed link receiver from small cell mobile network base stations in a dense urban environment to identify the degree to which the deployment would be constrained by its presence. The objective was to gain an understanding of a single specific sharing scenario located where high-resolution surface data was available by modelling it in detail. A critical case was considered to be the BT Tower as: 1. It is located at the centre of a dense urban area (central London). 2. The receiver height is large, increasing the difficulty of sharing. The receive antenna at the BT Tower was therefore used as the basis of the analysis Elements of the Analysis of the BT Tower Case The mobile deployment was specified using 100 small cell outdoor base stations deployed over an area calculated in Section to be a circle of radius 2.39 km around the BT Tower as shown in Table 7-3.

63 Geographic Sharing in C-band Page 63 Macro base station separation distance (km) 0.6 Number of Small Cells / Macro base station 2 Area / Macro base station (km 2 ) 0.36 Area / Small Cells (km 2 ) 0.18 Number of Small Cells in simulation 100 Area of Simulation (km 2 ) 18 Radius of deployment zone (km) 2.39 Table 7-3 Deployment Density for IMT-Advanced Small Cell Outdoor Systems Locations of the 100 small cell base station were randomised within a circle of radius 2.39 km around the BT Tower. They were then moved to the nearest street and for the measured base station data case their antenna was pointed along the street. No attempt was made to analyse interference between base station or the coverage / service provided. The resulting initial deployment locations are shown in Figure 7-5 with terrain colour coded so that lighter shades are higher. Figure 7-5: 100 Small Cell Base Station Located Within 2.39 km of Victim Fixed Link Station in Central London

64 Geographic Sharing in C-band Page 64 The same 3 m resolution surface database was used as for the central London VSAT site analysis. In this case an adjustment was made to the surface database to remove the BT Tower which would be interpreted (incorrectly) as an obstruction by P.452. Note that the propagation model was configured with a percentage of time = 50% to be consistent with the fixed link receiver threshold. Section 12 give a full description of the analysis and includes: 1. Details of the fixed link parameters 2. Consideration of appropriate threshold and interference apportionment in this specific case 3. Identifying interfering base stations to develop a deployment consistent with the I/N threshold 4. Analysis of average values of gain + diffraction losses 5. Concept of an average excluded zone base on smooth Earth propagation plus the average value of gain and diffraction. 6. Potential mitigations and their impact Summary Results and Conclusions BT Tower Case The interference zone around the fixed link receiver on the BT Tower was found to be on average km 2 in size. This zone was calculated using average gain + diffraction plus smooth Earth losses. The zone can be reduced using a number of mitigation techniques, discussed in detail in Section However, the average interference zone contained many locations where a base station could be located due to higher diffraction losses. This could facilitate operation very close to the fixed link station: indeed for the deployment of 100 base stations considered, between 94% and 96% were found to be located in positions that had sufficient diffraction to protect the fixed link receiver. In particular, a key factor was the fixed link antenna receive gain, and hence in locations not in its main beam there was high likelihood of ability to deploy small cell base stations. It is also noted that for the omni antenna case the aggregate I/N = -0.3 while for the directional case it was -2.3 db. With an adjacent band discrimination or NFD of 23.5 db, this means that all of the base station locations considered could operate without causing harmful interference when transmitting non-cofrequency. 7.5 Power and Traffic This section describes results from using Monte Carlo analysis to convolve: Variations in base station transmit power due to possible traffic variation over the whole day Use of P.2001 propagation model using full percentage of time range [0, 100] Polarization loss with range [0, 3] db

65 Geographic Sharing in C-band Page 65 The results, as presented in detail in Section 12.5, suggest that the mean power that could be used with P.452 and a percentage of time = 20% would be lower at 6.1 dbm / MHz or dbw / MHz. Alternatively, the power could be reduced by a further 2.3 db to take account of the time of day variation using the alternative traffic profile. When combined with average polarization loss of 1.5 db this would give a total reduction of 3.8 db. Note that this reduction in power relates to traffic variation and the power per user would not necessarily be affected. 7.6 Conclusions from High Resolution Detailed Analysis The spectrum availability maps we have developed in this analysis are likely to be conservative in dense urban areas. This is despite the fact that our method already improves on the one commonly used internationally in JTG There is significant potential to deploy low power low height base stations very close to satellite VSAT type Earth stations. The fixed link case is harder but operation is still possible close to point to point fixed link receivers located in dense urban areas. The key is to ensure there isn t line of sight between interferer and victim stations, and preferably a significant degree of diffraction loss due to buildings. This would impose constraints on where the mobile operator could deploy their base station, requiring them to (for example), choose the side of the street with the greatest radio shadow. However there would be significant benefits in terms of the number of locations that could be served. There are likely to be some geometries for which this approach would not be viable, in particular streets that point directly at receiver stations. Monte Carlo analysis was also undertaken that suggested that the average power used in studies could be reduced further to take account of time of day variation of traffic.

66 Geographic Sharing in C-band Page 66 8 MITIGATIONS This section describes mitigations that could be employed to facilitate operation of mobile network base stations within parts of C-band. Mitigation is discussed in the context of: Changes required to mobile networks Changes required to satellite Earth stations Changes required to fixed links Regulatory approach to mobile licensing Frequency or geographic migration 8.1 Mobile Network Based Mitigation The following mitigations were considered during this study: Traffic management, including time of day profiles, indoor operation and deployment density base station antenna performance base station deployment These are discussed in the following sub-sections Traffic / Coverage Management During the high resolution analysis consideration was made of the impact of the time of day variation of traffic carried by a mobile network. It was noted that the EIRP levels used in the ITU-R JTG studies were based upon the busy hour only. However, the interference threshold is averaged over the whole year. Therefore the average EIRP should take account of low traffic times of the day, such as early morning. An example traffic model was identified that would reduce the average EIRP by 2.3 db (see Section 7.5). Other diurnal models may be available, but it is certainly clear that JTG have over-estimated the EIRP level to be used. There is also significant location variation in traffic carried by mobile networks. One of the key assumptions was that this band would be used for small cell deployment, primarily aimed at urban and dense urban locations. This in itself brought significant advantages in all three of the analysis phases in that the base station was below the clutter line and hence in most cases the interfering signal was reduced by either clutter or diffraction. Consideration was also made in the high-resolution analysis of the ratio of small cells to macro cells. Interference constraints made it increasingly difficult to deploy additional base station within the area around the test satellite Earth station, as shown in Table 8-1 below.

67 Geographic Sharing in C-band Page 67 Scenario 1 small cells per macro cell Omni Antenna 20 out of % Directional Antenna 24 out of % 2 small cells per macro cell 37 out of % 45 out of % 3 small cells per macro cell 50 out of % 67 out of % Table 8-1: Active Base Stations that can be Operating while Meeting Satellite Earth Station I/N Threshold As the density increases, the ratio of the worst single entry interferer to aggregate interference increases. It reaches 11 db for the omni antenna case and 8.9 db for the directional antenna. This means a larger exclusion zone is required around the Earth station. Another deployment option is suggested by the fact that a significant reduction in interference would occur if the band were used solely for indoor base stations. This is a less attractive option for the mobile operator as it would involve significant constraints on the service. Reducing the traffic carried by C-band mobile network base stations would decrease interference but would similarly reduce the value of the band to mobile operators. However, analysis undertaken in this study suggests that a package of mitigations (such as the base station deployment option discussed below) would allow 3 small cells per macro cells to operate with a small exclusion zone Antenna Performance There are a number of options to modify base station antenna performance to support interference mitigation including: Antenna downtilt MIMO / beam forming Use of multiple sectors, sector planning and sector disabling MIMO antennas, beam forming and sector disabling are all examples of how detailed design can be used to concentrate radiated energy towards the intended users. Theoretically this should improve sharing significantly in some cases, but relies on specific detail of the deployment. In particular, these techniques are more applicable for macro base stations than small cells which typically use a single sector. The high resolution analysis considered two antennas, the baseline omni antenna and a directional antenna using manufacturer s measured data. The measured directional antenna could be used to point along a street or at a traffic hot spot such as stadium. It is small enough to be deployed on street furniture. The benefit for sharing is that it can be pointed away from a specific victim in both azimuth and elevation (i.e. downtilt). The increased directivity also meant that the far off axis gain was lower than the omni case, decreasing the gain towards the victim.

68 Geographic Sharing in C-band Page 68 In the satellite Earth station case, detailed analysis showed only a sub-set of base stations within the deployment zone need be switched off to reduce the exclusion zone from 3.28 km 2 to 0.91 km 2. This could be improved further by combining the pointing selection with deployment, as discussed further in the mitigation option below. There was a similar improvement for the fixed link case, with only a handful of base station in the excluded locations where they would have to be switched off, shown in figure 8-1 below. There was a similar reduction in the exclusion zone size, from 77.5 km 2 to 18.2 km 2. Figure 8-1: Excluded Base Station Locations Directional Antenna Base Station Deployment The majority of this study was undertaken using a 50m terrain and land use database that made assumptions about clutter loss in the direction of the victim. In the high resolution analysis in Section 7, use was made of a surface database that could identify the impact of deploying base station at specific locations. The analysis focussed on urban and dense urban locations where

69 Geographic Sharing in C-band Page 69 small cell base stations were likely to be deployed where there was also likely to be significant building diffraction loss. It was found that even within the interference zones identified in Section 4 there was significant potential to deploy base stations if the locations were selected to ensure there was a high diffraction loss in the direction of the victim satellite Earth station or fixed link. For example, the area around the central London VSAT Earth station was analysed under the assumptions in Table 8-2 and locations where deployment of mobile network small cell base stations would not be feasible were calculated, as shown in Figure 8-2. The analysis also considered the reduction in the interference zone to include in the analysis two potential mitigations: time of day variation and average polarisation discrimination totalling 3.8 db. Average transmit power of small cell base stations Antenna pattern of small cell base stations Peak gain of small cell base stations Threshold I/N at Earth station Aggregation factor Adjusted threshold I/N -19 dbw / MHz Omni 5 dbi -10 db 10 db -20 db Table 8-2: Parameters for Deployment Exclusion Zone Analysis Figure 8-2: Deployment Exclusion Zones for VSAT / Permanent Earth Station Case Key: Red: locations where I/N would exceed threshold even with use of mitigations

70 Geographic Sharing in C-band Page 70 Yellow: locations where I/N would be less than threshold taking into account 3.8 db of mitigation Similar analysis was undertaken for the BT Tower fixed link case as shown in Figure 8-3. Figure 8-3: Deployment Exclusion Zones for Fixed Link Case It was observed that in both cases: Deployment was difficult very close to the satellite Earth station or fixed link receiver (under a kilometre) or in open spaces such as parks (e.g. Kensington Gardens/Hyde Park for the central London VSAT case). Deployment was extremely difficult in streets pointing directly at the satellite Earth station or fixed link receive station (e.g. the A400 or Hampstead Road which points directly at the BT Tower). In most other cases there were locations available that could be used to deploy small cell base stations (e.g. roads on the map not shaded in red). Even though the locations in Figures 8-2 and 8-3 would be identified as unavailable if using terrain data and larger pixels, high-resolution analysis such as this suggests that small cell base stations could be deployed at 80 90% of positions. This included indoor locations. Further work would be required to identify the breakdown solely for outdoors. This could be considered a constraint, in that in many cases, deployment could only occur if the base station was located on certain sides of streets but not others. However, it can also be considered a mitigation, as using this technique it was possible to operate small cell base stations very close to the victim station without causing harmful interference.

71 Geographic Sharing in C-band Page Earth Station Mitigation The following mitigations were considered during this study: Introduce site shielding around the Earth station Increase the size of the antenna Under existing licensing arrangements, both of these would most likely require agreement from the permanent Earth station operator and involve additional costs. They were assessed in terms of the technical impact on the exclusion area around the Earth station Site Shielding Recommendation ITU-R SF.1486 suggests values of site shielding in the range 2-33 db considering a combination of natural and artificial shielding and by judicious choice of a ground location. In Section we have considered a value of 15 db additional shielding (close to the median suggested by Rec. SF. 1486) and calculated the average excluded area. Table 8-3 shows the excluded area in square kilometres, when considering omni directional IMT base stations. Results are given for the cases where there are 1, 2 and 3 small cells per macro cell. Scenario 1 small cell per macro cell 2 small cells per macro cell 3 small cells per macro cell No site shielding db additional site shielding Table 8-3: Average Area Excluded with 15 db Site Shielding (km 2 ) If such a shielding isolation is taken into account, up to the 33 db suggested by Rec. SF. 1486, then the required separation distance to protect satellite Earth station receivers from IMT transmitters can be reduced even further. However, the required distance separation between IMT transmitter and a receiving Earth station using site shielding has to be evaluated on a site-bysite basis and is dependent on characteristics and location of each site. The possibility of applying site shielding is not guaranteed for all sites Earth Station Antenna Upgrades It is always possible to improve a sharing scenario by upgrading antenna performance for any directional links in the scenario. In the case of the Earth station analysis a larger dish improves the margin in the satellite link and this extra margin can be in theory be allocated to additional interference from IMT. Though we might expect a larger dish to result in a smaller link temperature and hence a more sensitive link, manufacturers data indicate that this is a minimal effect at 10, 20 and 40 degree reference elevations

72 Geographic Sharing in C-band Page 72 The high-resolution analysis in Section 12 shows the impact of replacing a 2.4m dish with a 3 m dish a scenario relevant, for example, to VSATs in urban areas. The results are summarised in Table 8-4 below. The table shows the excluded area in square kilometres, when considering omni directional IMT base stations. Results are given for the cases where there are 1, 2 and 3 small cells per macro cell. Scenario 1 small cell per macro cell 2 small cells per macro cell 3 small cells per macro cell 2.4 m satellite Earth station dish (current size) m satellite Earth station dish (upgrade) Table 8-4: Comparison of Area (km 2 ) Excluded around the VSAT Site when using a 2.4m Dish Compared to a 3.6m Dish 8.3 Fixed Link Based Mitigation The geometry involved makes it harder to employ interference mitigation techniques for the fixed link case because: It is less practicable or beneficial to add site shielding to sites such as the BT Tower. In the direction in which the antenna points there must be a clear line of sight in order to provide the wanted service. In other directions there will be attenuation of interference due to the gain pattern without the need for site shielding. Using a larger antenna would improve the wanted signal but also the interfering signal. There would be some benefit due to relative reduction in the noise, but the benefit would be less than for the Earth station case. The geometry involved is shown in Figure 8-4. Fixed Link Receiver No potential for site shielding Larger Earth station antenna increases wanted signal but not interference Larger fixed link antenna increases both wanted and interfering signals Site Shielding Earth Station Base Station Figure 8-4: Difficulty of Mitigation for Fixed Link Receiver

73 Geographic Sharing in C-band Page 73 Relocation of some fixed link receivers is also a potential mitigation, discussed as an important option below in Section Interference Constraints in Mobile Licence It was noted that for deployments of fixed links and satellite Earth stations in urban areas it is still feasible to locate small cell base stations close by as long as there is shielding in the form of a building to provide diffraction loss. This means that it could be difficult to define an exclusion zone simply via a distance. One option could be to define a coordination contour based upon analysis using for example, terrain plus land use data. However, this could lead to a regulatory burden on the mobile operator given the number of base stations that will need to be deployed. An alternative would be to specify in the mobile operator s licence that they must ensure that the aggregate I/N at a specified number of fixed link and Earth stations is within a defined threshold. The mobile operator would then have the flexibility to: Deploy base stations Relocate base stations Balance density with how close to the victim station deployment is feasible Identify the best combination of locations to provide their customers service With this framework the mobile operator would be incentivised to undertake the additional work to identify those locations where there is sufficient diffraction that the fixed link or satellite Earth station would be protected. The mobile operator would also have access to high-resolution databases identifying buildings etc. in urban areas plus know what traffic hot-spots they wish to serve. The methodology in this study could be the basis of a sharing algorithm which could be published and agreed in the same way that the 3G and 4G coverage obligations were documented. 8.5 Frequency or Geographic Migration Another mitigation method would be to re-locate the satellite Earth station or fixed link to either an alternative geographic location or different frequency band, as discussed below. This could be encouraged by use of a pricing mechanism as discussed in Section or be part of a band segmentation regime as discussed in Section Earth Station Migration The principle market for mobile networks in this band is likely to be dense urban locations, and geographic separation can be increased by ensuring satellite Earth station are sited in remote rural areas. In some circumstances, this could require re-location, the difficulty of which would depend upon the user type. For example:

74 Geographic Sharing in C-band Page 74 The VSATs in central London can be assumed to be fixed locations with no opportunity to be moved. The Chalfont Grove site involves a large number of satellite Earth station antenna, representing a significant investment. For example, in the data provided by Ofcom, 8 antenna at this site were identified as operating at C-band. It is technically feasible that even a large teleport such as Chalfont Grove could be relocated, though this would also involve associated infrastructure being moved (e.g. connecting fibre, control stations etc.). Industry sources have suggested a cost of 15m, which would need to be compared with the value of the spectrum released. There is also the potential in some circumstances to move to alternative, higher frequencies, though again there could be limitations. For example: If the band is being used for telemetry, tracking, and command (TT&C) for existing satellites it will not be feasible to move to alternative bands as TT&C systems on the satellite cannot be re-tuned. However in these cases it could be feasible to move operation geographically to more remote sites. There could be constraints at the satellite dictating the bands it can use to provide services. For example it is possible that the uplink is at a location where C-band is required and the satellite used requires the downlink also to be in C-band, in particular for international services. If very high availability is required then this generally implies use of lower frequencies such as C-band where the rain fade is lower. In particular, if the target satellite is at a very low elevation angle, then the rain fade can become excessively large at higher frequencies bands, e.g. Ka band. This can also be driven by fade on the uplink which may be outside of the UK. However, migration in frequency of some satellite communication services should be feasible if there is sufficient time to identify technical solutions including transponder availability, though there are likely to be cost implications Fixed Link Migration The fixed links were found to create greater constraints on the deployment of mobile network small cells base stations than Earth stations due to their larger interference potential and the number located in dense urban areas. In theory, there is the possibility for migration of these links to higher frequency bands using one of two methods: In dense urban areas there is high availability of optic fibres, which could be used to provide the required connectivity. In other locations it could be feasible to create the same connections by moving to higher frequencies where hop lengths would be shorter but using additional hops. In practice, there could be resistance from existing licensees to this migration due to their requirement for low latency links. The original motivation for them

75 Geographic Sharing in C-band Page 75 selecting C-band for these point to point links was that the path length was long, reducing the need for additional hops that could increase delay. In addition, wireless services were selected over wired due to the need for low latency. If this requirement continues, then alternative low latency spectrum outside C-band would need to be found to incentivise migration. The other category of fixed link was to provide connectivity to the Scottish islands. Here the key requirement is path length, which can exceed 80 km for the more remote islands such as the Shetlands. In this case it would not be feasible to move to higher frequencies, but this is unlikely to constrain development of mobile networks in this band due to the lower traffic levels in these locations Pricing Mechanisms Migration could be encouraged via a spectrum pricing mechanism. Spectrum regulators have the job of balancing conflicting requirements from multiple stakeholders, and one solution is to define a framework that aims to optimise economic spectrum efficiency. One of the key tools in this framework is price, which applies to fees set for: Licences, for example for fixed links or transmit satellite Earth station. Recognised Spectrum Access (RSA), for example for receive satellite Earth station. Factors that can be included in pricing strategies are the demand for spectrum, density of deployment and alternative uses. In particular, pricing can take account of the spectrum opportunity cost of one service denying access to another. This cost can be quantified by considering the net benefit that would be gained by a service for it having access to additional spectrum 7. For mobile services, the benefit of access to spectrum can be very large, and so bands that could be used for mobile services have high valuations. In particular, valuations for mobile spectrum typically are significantly higher than that for bands used by satellite applications or most fixed links. Therefore, it could be argued that the opportunity cost of C-band should be taken into account when determining the prices of site licences for satellite Earth station and fixed links. This approach could be used to encourage migration of incumbent services (location or frequency), as discussed above. A pricing mechanism would likely have to reflect the differing value of spectrum in urban and dense urban areas compared to other locations. The economic spectrum opportunity cost of a fixed link in central London will be greater than that of the equivalent link in the Highlands and Islands of Scotland. It may also need to consider the type of use of the service, as some fixed link uses may be higher value than others (e.g. low latency trading links in London and across the south coast). However, it is noted that the motivation for migrating the incumbent is largely based on wider social economic arguments i.e. that the new service adds more value to the UK economy. Hence spectrum pricing would allow a balanced approach which is able to take account of these considerations: 7 We note that Ofcom published its approach to spectrum pricing in 2010.

76 Geographic Sharing in C-band Page 76 For fixed links: if the economic benefits due to low latency links exceed the benefit of mobile applications, then the licensees will continue to operate in this band. If not they will be motivated to move to other frequencies or technologies (such as fibre). For satellite earth stations: similarly, if the economic value of using existing satellites and infrastructure exceeds the cost of migrating (either to a new site location or frequency band) then they would be prepared to pay the additional costs. If there were a price structure that takes account of location, e.g. charging greater amounts for urban and dense urban locations, then this could incentivise site migration to other locations. The area denied to mobile services calculated using the methodologies described in this report could be used to derive the true spectrum opportunity cost to other services. The fixed link and satellite Earth station licensee have the option to trade their licence to the mobile operator, which would effectively remove that constraint Band Segmentation Frequency or geographic migration could be employed as part of a band segmentation policy. Band segmentation as a mitigation technique would rely on: 1. Neither sharing service requiring use of the full band for operation 2. Feasibility of operation with a small frequency separation. This study did not investigate the spectrum requirements for satellite Earth stations or fixed links. However, we have shown in Section 6 that net filter discrimination can give 23.5 db advantage in sharing with fixed links and 9.5 db when sharing with satellite Earth stations. The analysis in Section 6.3 suggests that this should be sufficient to permit operation in adjacent bands with minor constraints very close to an Earth station.

77 Geographic Sharing in C-band Page 77 9 CONCLUSIONS AND RECOMMENDATIONS 9.1 Conclusions The aim of this study was to develop and assess the potential for sharing in C-band between incumbent users such as fixed links and satellite Earth stations and mobile services, in particular IMT-A Results of Analysis As part of this project, Transfinite undertook three levels of analysis: Spectrum Availability Analysis Interference Zone Analysis High Resolution Analysis These are discussed in the following sections Spectrum Availability Analysis The principal conclusion is that there is scope for sharing spectrum in this band. The reason we believe this is that, even in the baseline case, 80% of the spectrum is available to 50% of the urban population which represents a massive potential economic value. Half the existing spectrum is available to 65% of the urban population, and this increases to around 90% if 20 db of mitigation is applied. On the other hand, even on the application of 20 db of mitigation, potential interference cannot be ruled out in some populous areas at all frequencies. Whether 20 db mitigation is possible in all cases is not clear, but our simulations indicate that improved modelling based on higher resolution surface data could easily find an extra10 db. The implication is that a managed approach to shared access of the band based on geographic sharing has great merit. Spectrum is available but the possibility of interference remains in some locations. It was also important to quantify the effects of mitigations. These included, for example, the additional losses due to accurate local clutter modelling, large antenna elevations in urban areas and building penetration losses. To generate flexible outputs and avoid having to explicitly model each possible combination of multiple mitigation techniques, our approach was look at relaxing the I/N threshold 5, 10 and 20 db. This is an established practice in satellite Earth station coordination, where the generation of auxiliary contours is common. The primary numerical result of the study is the amount of spectrum available in areas of high population density, under the different assumptions. The figure below shows a CDF of the availability of spectrum across the UK by percentage of population in urban, dense urban and hotspot areas.

78 Geographic Sharing in C-band Page 78 Figure 9-1: Availability of Spectrum in GHz band in MHz by Percentage of Urban+ Population under Baseline Assumptions Detailed analysis presented in summary in Section 7 and in full in Section 12 shows that various mitigations can be applied in many cases. To see some idea of the potential impact, Figure 9-2 shows the spectrum available when 5, 10 and 20 db of mitigation is applied to the baseline case. Further informamation on the impact of mitigation can be seen in Figure 9-3 below.

79 Geographic Sharing in C-band Page 79 Figure 9-2: Availability of Spectrum in GHz Band in MHz by Percentage of Urban Population under Baseline Assumptions and with 5, 10 and 20 db Mitigations applied Interference Zone Analysis In the analysis phase of the project, interference zones around specific Earth station and fixed link receivers were developed to gain an understanding of the issues involved. It also allowed rapid analysis of what-if cases and to crosscheck the outputs of the spectrum availability analysis. The satellite Earth station analysis was based around the Chalfont Grove site. A range of scenarios were modelled including the JTG generic case, inclusion of terrain, addition of clutter loss via local clutter codes and mitigations that are derived from the detailed analysis. The fixed link analysis was based around several sites in the South East and a similar range of analysis was performed. The analysis also considered the effect of operating non-co-frequency via use of the net filter discrimination and multiple carriers. It was noted that: The area excluded due to the fixed link was greater than that required to protect the Earth station.

80 Geographic Sharing in C-band Page 80 The fixed link interference zones also tend to cover more populated areas than the satellite Earth station greatly increasing their impact on spectrum availability per head of population. Assuming 20 db of mitigation removed most of the interference zone around the Chalfont Grove site. The net filter discrimination for the satellite Earth station case was significantly less than that for the fixed link receiver, leading to greater potential for interference in the non-co-frequency case. However, even for the fixed link, the out of band interference cannot be neglected High-Resolution Analysis The basis of the high-resolution analysis was a selective qualitative analysis using a 3m resolution surface database within central London. We consider this representative of the type of high density urban areas where mobile networks are likely to be deployed. From the assignment database provided by Ofcom the following were selected for high-resolution analysis: Earth station: VSAT based in central London Fixed link: BT Tower, connected to the LSE The number of base stations that could be deployed for the Earth station case is given in Table 9-1. Scenario 1 small cells per macro cell Omni Antenna 20 out of % Directional Antenna 24 out of % 2 small cells per macro cell 37 out of % 45 out of % 3 small cells per macro cell 50 out of % 67 out of % Table 9-1: Active Base Stations that meet Earth Station I/N Threshold It was noted that: It became increasingly hard to deploy small cell base stations as the density increased due to interference aggregation effects. Sharing was facilitated via use of a directional antenna that could have higher discrimination towards the Earth station / fixed link receiver. It was in general harder to share with the fixed link than the Earth station due to the geometry involved, as shown in Figure 1-4. A number of possible mitigations were considered, including: Analysis of traffic profiles and polarisation in Section 7.5 suggested an average attenuation of interference of 3.8 db could be used. Use of a larger dish antenna or site shielding were found to be effective ways of facilitating sharing with the satellite Earth station, as discussed in Section It was in general harder to use these mitigations for the fixed link.

81 Geographic Sharing in C-band Page 81 There was in general a significant attenuation of interference due to the combined diffraction over buildings and the relative gain of the small cell base stations antenna of around 50 db. With a dense deployment of small cell base stations close to the fixed link or Earth station receiver, there could be significant aggregation effects, with 10 db observed. However, the net effect was sufficient in many cases to permit operation of small cell base stations close to the receivers. A key factor was the site chosen for each small cell base stations and in particular the geometry of the path to the Earth station or fixed link receiver. Locations where there was line of sight so that there was no significant diffraction, would not be usable sites for small cell base stations. These locations were identified for the two tests cases described in Section 7 and shown in Figure 8-2 (for the Earth station) and in Figure 8-3 (for the fixed link). It was noted that in both cases unless very close to the base station there was usually the option to deploy a base station if its location could be selected, in particular: Not on streets pointing towards the victim receiver; On the side of streets closer towards the victim receiver to maximise diffraction loss. 9.2 Recommendations This project has extended, in various dimensions, the work undertaken and methodologies used at international forums such as the ITU-R by taking into account actual assignments and a high-resolution analysis of dense urban deployments. This more detailed approach generated results that suggests that there is a greater potential for use of this band by mobile networks than would be identified using generic low-resolution studies. The work, however, also raised issues that would benefit from further study, including those described below Thresholds The analysis was undertaken as agreed with Ofcom using the long-term threshold. However, work within JTG suggested that short-term threshold could be more sensitive, leading to some very large minimum separation distances. It would be beneficial to consider further the short-term interference threshold, noting that there are differences between the study approach described here and that within the JTG. In particular: The JTG analysis was mostly based upon use of macro-cells (higher power and above the clutter line) rather than small cells (lower power and below the clutter line). Our analysis is therefore concentrated on relatively short separation distances and short interference paths. For short paths some propagation modes of Recommendation P.452, such as ducting and troposcatter are negligible. Whereas on longer

82 Geographic Sharing in C-band Page 82 paths there is potential for atmospheric ducting to greatly enhance the interfering signal Scope and Limitations of Analysis The analysis covered a range of scenarios. It concentrated on the most likely mobile deployment options and focussed on long term sharing criteria. We identified a number of mitigation techniques and related these to a general relaxation in the I/N thresholds. The results give a good characterisation of the sharing issues and a large amount of background data is available to inform strategic planning for this band. However, there are a number of ways in which the analysis could be extended and improved: Detailed modelling of alternative deployments, for example macro cells in the mobile network. Examination of additional power and traffic models. Production of more accurate maps using smaller pixels in the spectrum availability analysis. Running the spectrum availability analysis with specific mitigations derived from the high-resolution analysis rather than generic attenuations of {5, 10, 20} db. Extending the high-resolution analysis to assess more locations and fixed link / Earth station receivers Propagation Modelling NFD Significant differences were noted between the clutter loss model in P.452 and undertaking the same calculation using diffraction for a specific location. There would be significant benefits in further study, potentially including measurement of diffraction loss in dense urban areas. This could be done by using the fixed link as a reference signal and measuring the attenuation in various stages of shadow behind buildings. The high-resolution analysis could be extended to generate statistics that could be used in other studies. The NFD for the satellite Earth station case was low due to assumptions about the receive filter. There would be benefits in studying this further to identify if this was a conservative assumption, for example using alternative filters or using measured data International Support It would be useful to build international support for a more detailed approach that takes into account actual assignments and high-resolution surface databases.

83 Geographic Sharing in C-band Page 83 This could be done by developing the work in this study into a series of papers that could be forwarded to international meetings including CEPT / ECC or ITU-R as appropriate Pricing Given that the band could be used for mobile applications, the area denied and area type (urban / dense urban) could be the basis of revised pricing. Issues regarding the pricing of licences and recognised spectrum access (RSA) for assignments in this band are worthy of further study. The methods and results of the analysis presented here would be useful inputs to this study. Unavailable locations for mobile services due to continued operation of fixed link and satellite Earth station licences, would be an important factor in a pricing strategy Regulatory Options The study suggested that one option would be to incorporate in mobile spectrum licences the need to protect existing assignments to a given I/N threshold. The implications of this could be analysed further with feedback from the operators on the operational constraint this would involve. A key question would be whether the operator could maintain coverage at the required level while meeting the deployment constraint to avoid harmful interference into the fixed link or Earth station receiver. It would also be worth considering this constraint in the context of availability of suitable sites in dense urban areas The Impact of Mitigations The study identified a number of possible mitigations which would require changes to Earth station and fixed link receivers. These could be analysed further with feedback from the licence holders as to feasibility and cost. Other mitigations are site specific and/or IMT deployment specific so it is not easy to define an optimum set of mitigations that can be generally applied. However, to see the potential effect of applying mitigations we can refer to the Figures (maps and CDFs) in Section 4. Figure 9-3 below is a close up view of the effect on spectrum availability in the London area in the baseline case and the cases where 5, 10 and 20 db of mitigation are applied.

84 Geographic Sharing in C-band Page 84 Baseline Case 5 db Relaxation 10 db Relaxation 20 db Relaxation Figure 9-3: The Available Spectrum in London in the Baseline Case and with 5, 10 and 20 db of Mitigation, in GHz Sharing will all Satellite Earth Station and Fixed Link Carriers

85 Geographic Sharing in C-band Page ANNEX: FIXED LINKS INTERFERENCE ZONE ANALYSIS This section sets out some supplementary technical information related to the Interference Zone analyses for fixed links described in Section Single-Entry Interference Figure 10-1 illustrates the single-entry interference modelled at each pixel. Here, the entire bandwidth of the 30 MHz fixed link receiver is populated with 10 MHz IMT interferers. Hence, in these analyses, single-entry interference is defined as multiple interfering signals sourced from a single location. Three 10 MHz IMT interferers 30 MHz fixed link receiver Figure 10-1: Interference from IMT In the simulations, we have considered interference sourced from, and incident to, any 1 MHz of bandwidth where the IMT interferer and the fixed link receiver are co-frequency Transmitter Power Table 10-1 illustrates the derivation of IMT transmitter output power in 1 MHz of bandwidth. Description Value Unit tx power in 10 MHz 24 dbm tx power in 1 MHz 14 dbm Activity Factor 3 db adjusted tx power in 1 MHz 11 dbm adjusted tx power in 1 MHz -19 dbw Table 10-1: Transmitter Power Here we take the IMT transmitter operating in 10 MHz bandwidth as a reference point and adjust the power for 1 MHz of bandwidth taking account of the activity factor Receiver Interference Threshold The fixed link protection criterion of db is discussed in section 3 of this report. Applying this ratio to the total noise threshold N = dbw specified in Ofw446 delivers the single-entry interference threshold I = -134 dbw used by Ofcom in its frequency assignment procedures. Table 10-2 shows the adjustments made for a receiver operating in 1 MHz of bandwidth.

86 Geographic Sharing in C-band Page 86 Description Value Unit Total noise threshold in 30 MHz dbw Total noise threshold in 1 MHz dbw I / N db Single -entry threshold in 1 MHz dbw Table 10-2: Single-Entry Threshold The single-entry interference threshold of dbw/mhz is used to define the results of the fixed link area analyses. An IMT Base Station is positioned in each pixel covered by the analysis and interference calculations performed. If the interfering signal power incident to the fixed link receiver exceeds dbw, the pixel is coloured red, otherwise the pixel is not coloured Antennas As discussed in section 3, for the outdoor Base Station runs, a typical Base Station antenna was sourced for the IMT interferer with a peak gain of 5 dbi. A simple omnidirectional antenna was deployed for the indoor Base Station runs. The fixed link antenna pattern used in these area analyses was sourced from Ofcom data. All of the fixed links identified in the minimum run schedule are equipped with antenna A/04/H/00/016/AA with a peak gain of 35.1 dbi. The antenna pattern is illustrated in Figure Figure 10-2: Fixed Link Antenna Pattern

87 Geographic Sharing in C-band Page ANNEX: NFD AND SPECTRUM MASKS In this section of the report, we set out some supplementary technical material on net filter discrimination (NFD) and spectrum masks Net Filter Discrimination The ETSI Technical Report ETSI TR sets out a well-established method for calculating NFD where spectrum masks associated with the victim receiver and interfering transmitter are convoluted in frequency; this method is used by Ofcom in its frequency assignment and frequency coordination work. The method samples spectrum masks associated with the victim receiver and interfering transmitter co-frequency and the samples are summed. The process is then repeated with the interfering transmitter tuned to the desired frequency offset. Then the ratio of these two sums is obtained and expressed in decibels. Equation 3-1 captures the method: in NFD 10log i in1 ( Ti 10 i0 ( Ti Ri ) /10 ( offset ) Ri ) /10 ( db) Equation 11-1: Net Filter Discrimination Where T i and R i are the i-th samples from the co-frequency transmitter and receiver masks respectively and T i(offset) is the ith sample from a transmitter mask offset in frequency from the victim receiver Spectrum Masks ETSI specifies spectrum masks for fixed link transmitters and, although receiver masks are not standardised, there is a well-defined method set out in ETSI TR for deriving these from the transmitter mask. Figure 11-1 shows the fixed link spectrum mask associated with the radio system considered in our analyses. In general, the masks extend five times the channel bandwidth of the radio system; 2.5 times channel bandwidth either side of the carrier centre frequency. The IMT transmit spectrum mask was sourced from Section 3.1, Annex 17 of the JTG Chairman s Report which specifies the adjacent channel leakage ratio (ACLR) associated with these radio systems. The mask was constructed such that the out of band (OOB) domain was consistent with the Ofcom methodology. Figure 11-2 shows the resulting IMT spectrum mask.

88 Geographic Sharing in C-band Page 88 Figure 11-1: Fixed Link Spectrum Mask Figure 11-2: IMT Spectrum Mask

89 Geographic Sharing in C-band Page ANNEX: HIGH RESOLUTION ANALYSIS 12.1 Overview One of the key challenges in undertaking interference analysis is to balance the need to protected incumbent services without being over conservative by taking a series of worst-case assumptions. One way to facilitate sharing is to use more detailed models as this can reduce the number of assumptions that are required. The following approaches to modelling propagation paths can be considered to be increasing in detail: 1. Smooth Earth, no terrain or clutter 2. Use of a terrain database 3. Use of terrain and land use database 4. Use of a surface database The majority of the study was undertaken with the third of these approaches, as this was the most detailed level information available on a UK-wide basis. However for central London the project team had access to a higher resolution 3m surface database, as described in Section This surface database was used to analyse a small number of cases in highresolution, considering the impact of needing to protect existing fixed link or Earth station receivers on the ability of an mobile networks operator to deploy small cell base stations in a dense urban environment Propagation and Surface Data Databases As far as possible, our analysis for this study took into account the actual sharing environment, including UK specific terrain and clutter. Therefore, rather than making worst case assumptions such as smooth Earth, the analysis was based upon terrain and surface databases for the areas where satellite Earth stations are located. The following databases were available: Ofcom s standard 50 m terrain and land use code database High-resolution 3 m surface database of central London The following figures show central London in each of these databases where the height scale for the terrain and surface data is as in Figure 12-1.

90 Geographic Sharing in C-band Page 90 Figure 12-1: Height Scale Used with Terrain and Surface Data Figure 12-2: 50 m Terrain Data of Central London

91 Geographic Sharing in C-band Page 91 Figure 12-3: 3 m Surface Database of Central London Figure 12-4: 50 m Land Use Data of Central London

92 Geographic Sharing in C-band Page Propagation Models These databases were used as inputs into the propagation models, in particular Recommendation ITU-R P.452: Prediction procedure for the evaluation of interference between stations on the surface of the Earth at frequencies above about 0.1 GHz. This model, to which the UK has contributed both theoretical models and measurement data, is well established and used both UK-wide and internationally. It includes a number of modes within the core model, such as: Line of sight Diffraction Surface ducting Tropospheric scatter Elevated layer reflection and refraction These various components are selected based upon a check on whether the path is line of sight or not, using a path profile created from a terrain or surface database. The relevant terms are then merged mathematically to generate a path loss. The model then defines how the path loss can be adjusted to include a height gain variation due to clutter. This requires an additional land use database to be available and for there to be a mapping from each land use code to the parameters required by P.452 s clutter model. For the Ofcom land use database these are as given in Table If a surface database is used then it is not necessary to include the clutter loss as that will be calculated in the P.452 path loss using the diffraction model. Note that the core propagation model in P.452 has been extended: Into the point-to-area model P.1812 by including a location variability term (typically log-normal with standard deviation e.g. 5.5 db). A specific variation on P.452 is that P.1812 includes the clutter height on top of the terrain height when deriving the path profile. Into a generic wide ranging propagation model applicable for Monte Carlo analysis in P.2001 by including fading terms.

93 Geographic Sharing in C-band Page 93 Clutter type Nominal height, h a (m) Nominal distance, d k (km) Undefined 0 0 Open Fields Main Road 0 0 Buildings Urban Suburban Village Sea 0 0 Lake 0 0 River 0 0 Coniferous Deciduous Mud flats 0 0 Orchard Mixed trees Dense Urban Table 12-1: Land Use Codes and P.452 Nominal Heights and Distances Clutter and Diffraction Loss While diffraction and clutter loss models are based upon common concepts, they are implemented differently and this leads to significantly different results. JTG Document 715 Annex 2 relating to IMT parameters stated the following relating to below rooftop base station antenna deployment: When conducting sharing studies it is also important to account for how the antennas are deployed in relation to the surrounding environment, including the clutter. If the antennas are deployed below the rooftop level it might be necessary to use a different propagation model compared to the scenario when the antennas are installed above the roof top level. An alternative approach could be to add clutter loss to propagation loss calculations The propagation model in P.452 uses the diffraction loss model from Recommendation ITU-R P.526. For a knife edge diffraction of height h at distance d 1 from the transmitter and d 2 from the receiver, the loss can be calculated using: L d = log ( (v 0.1) v 0.1) where, using to denote wavelength:

94 Geographic Sharing in C-band Page 94 v = h 2 λ ( 1 d d 2 ). However the clutter loss in P.452 is where: A h = 10.25F fc e d k {1 tanh [6 ( h h a 0.625)]} 0.33 F fc = {1 + tanh[7.5(f GHz 0.5)]}. These lead to significant different values, as can be seen in Table Obstruction Close Baseline Far Frequency (GHz) Transmit antenna height (m) Obstruction height (m) Distance to obstruction (m) Distance from obstruction to receiver (m) Clutter loss (db) Diffraction loss (db) Table 12-2: Diffraction vs. Clutter Loss In particular, the clutter loss is capped at around 20 db while the diffraction loss can be significantly greater in the right hand column for the case where the transmitter is much closer to the obstruction. This highlights another difference between using a land use database and surface database: the land use database will always use the same values of distance to obstruction and obstruction height but with the surface database these values will vary depending upon geometry. As the land use database does not know the actual case involved, it must use simplifying assumptions and generate a typical value that might not be applicable for most actual deployments. In particular, small cell base stations deployed in dense urban areas are likely to use street furniture that are located close to one side or other of a street, as shown in the figure below, leading to cases where diffraction loss will be greater than that predicted by the clutter model.

95 Geographic Sharing in C-band Page 95 Base Station Figure 12-5: Higher Diffraction than Clutter Model Predictions Analysis in Section suggested that the interference can become either: Antenna Gains An aggregate of multiple interferers all significantly attenuated due to large diffraction losses Dominated by a single interferer that has significantly lower diffraction loss (e.g. due to street being aligned with the victim) Another factor that will have a significant impact on the interference calculation is the gain at the transmit antenna, and this will vary depending upon the terrain or surface database used. If a terrain database is used then it is likely that for short paths the direct path is used to calculate the transmit gain, but if a surface database is used the radio path can be very different and hence the gain calculated also, as in Figure Figure 12-6: Radio Path vs. Direct Path This would lead to large differences in the calculated aggregate interference level, particularly if the base station is modelled using measured data from a directional antenna Summary Two mechanisms were identified by which the aggregate interference calculated using a high-resolution surface database could be significantly different from that derived using terrain and land use codes:

96 Geographic Sharing in C-band Page The clutter loss calculation uses fixed parameters and is capped at around 20 db 2. The transmit gain calculated using a terrain database is likely to use an inaccurate direct path This suggests that the lower aggregate interference levels calculated using a surface database are based upon real phenomena and can be accepted for use in further analysis Earth Station VSAT Test Case The previous section identified benefits in analysing sharing between mobile network base stations and incumbent systems using a high-resolution surface database. This section considers the impact on satellite Earth stations, starting with identification of system parameters System Parameters IMT Parameters The study used IMT-A parameters that were agreed for use in sharing studies in JTG As we were concentrating on small cell deployments in urban areas Table 12-3 was extracted from document JTG /715 Annex 17 from the Chairman s report of the final JTG meeting. This Annex is dated 18 th August 2014 and is the basis for a Draft New ITU-R Report. The analysis was undertaken with a reference bandwidth of 1 MHz on the basis that the entire victim Earth station bandwidth would be used for IMT/LTE deployment. The power density of 24 dbm across 10 MHz was a median value averaging across 5, 10 and 20 MHz. The density of small cell outdoor base stations was derived from: Small cell outdoor: 2 per urban macro cell Urban macro cell every 0.6 km 50 small cell outdoor base station were deployed over an area calculated from the above to be a circle of radius 1.69 km based upon the victim Earth station as shown in Table 12-4.

97 Geographic Sharing in C-band Page 97 Small cell outdoor JTG Baseline Measured base station data Cell radius / Deployment density 1-3 per urban macro cell (<1 per suburban macro site) 1-3 per urban macro cell (<1 per suburban macro site) Antenna height above terrain 6 m 6 m Sectorization single sector single sector Downtilt N/A 5 Frequency reuse 1 1 Antenna pattern Recommendation ITU-R F.1336 Omni k = 0.5 Measured data Antenna polarization linear linear Below rooftop base station antenna deployment Maximum base station output power in 10 MHz Maximum base station antenna gain 100% 100% 24 dbm 24 dbm 5 dbi 9.32 dbi Average base station activity 50% 50% Average base station power/ 10 MHz taking into account activity factor Maximum base station output power/sector (e.i.r.p.) in 10 MHz Average base station power / 1 MHz taking into account activity factor 21 dbm 21 dbm 26 dbm dbm 11 dbm = -19 dbw 11 dbm = -19 dbw Table 12-3: Parameters for IMT-Advanced Small Cell Outdoor Systems Macro base station separation distance (km) 0.6 Number of Small Cell / Macro base station 2 Area / Macro base station (km 2 ) 0.36 Area / Small Cell (km 2 ) 0.18 Number of Small Cell in simulation 50 Area of Simulation (km 2 ) 9 Radius of deployment zone (km) 1.69 Table 12-4: Deployment Density for IMT-Advanced Small Cell Outdoor Systems Locations of the 50 small cell base station were randomised within a circle of radius 1.69 km around the Earth station. They were then moved to the nearest

98 Geographic Sharing in C-band Page 98 street and for the measured base station data case their antenna was pointed along the street. No attempt was made to analyse interference between base station or the coverage / service provided. The resulting initial deployment locations are shown in Figure Figure 12-7: 50 Small Cell Base Station Located within 1.69 km of Victim Earth Station The gain pattern for Rec. F.1336 for peak gain of 5 dbi in omni k=0.5 case is shown in Figure 12-8.

99 Geographic Sharing in C-band Page Gain (dbi) Elevation angle (deg) Figure 12-8: Rec. F.1336 Gain Pattern Peak Gain = 5 dbi k = 0.5 For comparison, the antenna pattern of a real base station was used. The one selected was: CommScope CMAX-DM60-CPUSEi53: Cell-Max Directional High Capacity Venue MIMO Antenna, MHz and MHz. It was selected as: It was small enough to be mounted on a lamp post (30 cm x 30 cm). The upper frequency band was close to 3 GHz. The target applications included high density locations requiring some directivity, such as stadiums and traffic hot spots. The gain pattern was downloaded for 2.7 GHz where the peak gain was 7.18 dbd = 9.32 dbi with patterns as in Figure 12-9 and Figure The azimuth and elevation slices were merged together using the smoothing parameter: λ El = El 90

100 Geographic Sharing in C-band Page Gain (dbi) Gain (dbi) Azimuth (degrees) Figure 12-9: Measured Azimuth Gain Pattern Elevation (degrees) Figure 12-10: Measured Elevation Gain Pattern

101 Geographic Sharing in C-band Page Earth Station Parameters The parameters for the Earth station (ES) were taken from the data provided by Ofcom. The selection criteria was: In central London Operating in the bands targeted for use by mobile networks Believed to be active and unlikely for its status to change in the medium term The result was the ES or very small aperture terminal (VSAT) was modelled with parameters in the table below: FREQUENCY CHANNEL SPACING NAME RX STATION GHz MHz LONDON COORD_LONG 000W N EMISSION DESIGNATION ANTENNA POLARIZATION 150KG7W CR AZIMUTH 196 ANTENNA ELEVATION 29.5 T/I 10 TRANS. ANTENNA GAIN ANTENNA BEAMWIDTH 2.6 VERTICAL BEAMWIDTH 2.6 Table 12-5: Parameters of Victim Earth Station From these parameters and the project methodology the following simulation and derived values were developed: Dish size Height of antenna above building Receive temperature Polarization 2.06 m 3 m 100 K Circular Target GSO satellite longitude Gain pattern Aggregate I/N Threshold Propagation Modelling Rec. S.580 (ITU-R APL version) -10 db Percentage of time 20% Table 12-6: Additional Earth Station Parameters The objective was to analyse in detailing a real sharing scenario to identify how the constraint of protecting a satellite Earth station would restrict a mobile

102 Geographic Sharing in C-band Page 102 operator s ability to deploy base stations. This therefore used a surface database with resolution of 3m, which was sufficient to identify individual buildings and locate base station on streets. The propagation model used was P.452 with an associated percentage of time = 20%. For these short paths the principle mechanism is diffraction rather than troposcatter, ducting and layer refraction. It is also necessary to consider multipath effects and reflections, though geometry suggests the likelihood would be low of there being a surface that would reflect energy towards the Earth station: Diffraction Reflections Earth Station Base Station Figure 12-11: Geometry of Base Station below Clutter Line A value of 3 db enhancement was used on-top of the P.452 prediction. It could be worth running measurements to compare predictions using a highresolution surface database with the P.452 propagation model against actual radiowave behaviour. Note also that the angle at the base station towards the diffraction point is typically a high elevation angle. This means there is likely to be some antenna discrimination, particularly if the antenna is using some downtilt Identify Interfering Base Stations Methodology The approach taken was to calculate the aggregate I/N at the Earth station. If the I/N was above the required threshold of aggregate I/N = -10 db then the highest single entry base station would be identified and removed from the list of interferers. This process was repeated until the I/N threshold was just met. In some cases it was observed that minor modifications to the deployment would facilitate sharing, such that: Two base stations were located on elevated roads (the Westway), and were consequently moved to ground level; Two base stations were located on streets that pointed at the Earth station and were moved to nearby sites with additional protection; Five other base stations were moved to increase shielding to nearby locations. The critical output was then:

103 Geographic Sharing in C-band Page 103 The locations of the base stations that would not be feasible (i.e. an interference zone); The locations of the base stations that would be feasible. The resulting deployment was then assessed for impact in terms of constraints on the mobile operator s deployment. The process was repeated for the two sub-cases: Omni antenna base stations. Directional Antenna base stations Omni Antenna Case Of the 50 base stations in the initial deployment, 11 had to be switched off to meet the required I/N = -10 db threshold. However, 39 could continue to operate within the deployment zone, as can be seen in Figure Figure 12-12: Omni Antenna Exclusion Zone The figure shows a red polygon that contained all the base stations that had to be switched off and the circle of radius 1.69 km within which the 50 base stations (shown as white dots) were deployed.

104 Geographic Sharing in C-band Page Directional Antenna Case Of the 50 base stations in the initial deployment, 3 had to be switched off to meet the required I/N = -10 db threshold. However, 47 could continue to operate within the deployment zone, as can be seen in Figure 12-13, which also shows the red polygon enclosing those base stations (white dots) that had to be switched off. Figure 12-13: Directional Antenna Exclusion Zone Analyse Single and Aggregate Interference Methodology Having identified the set of base stations that would lead to the aggregate I/N = -10 db threshold being just met, the data was analysed further to derive: The average in absolute [Grel + Diffraction loss] for each scenario taking into account all 50 possible base stations. The ratio of worst single entry interference to aggregate interference in the case that the aggregate threshold is just met. An area analysis showing the exclusion zone taking into account the average in absolute [Grel + Diffraction loss] plus aggregation factor.

105 Geographic Sharing in C-band Page 105 From the exclusion zone, the average area that would be inaccessible to mobile network base stations. This was undertaken for both the omni and directional cases described earlier Omni Antenna Case For the omni antenna case the values calculated were as shown below Average [Grel + Diffraction] 51.7 Single entry to aggregate I/N ratio db factor for multi-path enhancements -3 Total adjustment to free space propagation loss 37.7 Table 12-7: Omni Case Adjustments to Single Entry Free Space The adjustment factor of 37.7 db was then used to create an average exclusion zone around the satellite Earth station as in Figure which also shows the circle around the ES radius 1.69 km within which the base stations were deployed. Figure 12-14: Average Exclusion Zone for Omni Antenna Case

106 Geographic Sharing in C-band Page 106 The exclusion zone was calculated to be an area = 2.93 km 2. The exclusion zone is seen to have the classic key hole shape pointing in the direction of the Earth station s satellite (i.e. towards the south-south-west) Directional Antenna Case For the directional antenna case the values calculated were as shown below: Average [Grel + Diffraction] (*) 56.6 Single entry to aggregate I/N ratio db factor for multi-path enhancements -3.0 Total adjustment to free space propagation loss 46.2 Table 12-8: Directional Antenna Case Adjustments to Single Entry Free Space (*) For consistency with the omni case, all values of Average [Grel Diffraction] are compared to the case of an isotropic antenna with peak gain of 5 dbi. The adjustment factor was then used to create an exclusion zone around the satellite Earth station as in Figure which also shows the circle around the ES radius 1.69 km within which the base stations were deployed. Figure 12-15: Average Exclusion Zone for Directional Antenna Case

107 Geographic Sharing in C-band Page 107 The exclusion zone was calculated to be an area = 0.41 km Analysis with Variable Base Station Density The analysis was repeated using an assumption of 1 or 3 small cell base stations per macro cell rather than 2. Initially the analysis considered the higher density case with 75 (corresponding to 3 small cells per macro cell) rather than 50 base stations (corresponding to 2 small cells per macro cell) within the same area of interest as in Figure Figure 12-16: Base Station Locations for Higher Density Scenario The same process was used as for the 2 small cell base stations per macro cell namely: Deployment: Position base station at random within circle. Move to nearest street. Make limited adjustments for location (e.g. avoid very low separation distances between base station and relocate some base station within parks). For the directional case, orientate antenna to point along the street.

108 Geographic Sharing in C-band Page 108 Analysis: 1. Calculate aggregate I/N at Earth station from all base stations. 2. If the aggregate I/N > -10 db then: a. Identify the worst single entry case b. Remove that base station from the group of interfering base station c. Continue at Step 1 3. Output final base station deployment The link budgets for each of the 75 base stations were then exported to Excel where the data was post-processed to analyse the lower density case. In addition the baseline case of 2 small cells per macro cell was repeated to: Ensure there was consistency in the deployment between the three densities; Provide an opportunity to analyse the sensitivity of the initial deployment. The number of active base stations that meet the Earth station I/N criteria are shown in the table below for the 1, 2 or 3 small cell base stations per macro cell cases. Scenario 1 small cells per macro cell Omni Antenna 20 out of % Directional Antenna 24 out of % 2 small cells per macro cell 37 out of % 45 out of % 3 small cells per macro cell 50 out of % 67 out of % Table 12-9: Active Base Stations that meet Earth Station I/N Threshold It can be seen that as the number of small cells per macro cell increased, the omni case had difficulty in deploying additional base stations, suggesting it was approaching an interference driven limit. However, there was greater success (89.3 % against 66.7%) in deploying the directional antenna case due to the better antenna discrimination. In addition, there were a couple of cases for the directional antenna where interference could have been avoided by changing its azimuth angle. The locations of the active and switched off base station for the 3 small cells per macro cell case are shown in Figure and Figure where: Base stations that could remain active are shown as white dots Locations where base stations had to be switched off are shown as blue circles. Note use of the terrain / surface colour scheme shown in Figure 12-1.

109 Geographic Sharing in C-band Page 109 Figure 12-17: Locations of Active and Switched Off base stations for Omni Antenna Scenario with 3 Small Cells per Macro Cell

110 Geographic Sharing in C-band Page 110 Figure 12-18: Locations of Active and Switched Off base stations for Directional Antenna Scenario with 3 Small Cells per Macro Cell The degree to which the average [gain + diffraction loss] calculated using this methodology varied by number of small cells per macro cell is shown in Table and Table Number of Small Cells per Macro Cell Average [Grel + Diffraction] Single entry to aggregate I/N ratio db factor for multi-path enhancements Total adjustment to free space propagation loss Table 12-10: Omni Case Adjustments to Single Entry Free Space

111 Geographic Sharing in C-band Page 111 Number of Small Cells per Macro Cell Average [Grel + Diffraction] Single entry to aggregate I/N ratio db factor for multi-path enhancements Total adjustment to free space propagation loss Table 12-11: Directional Case Adjustments to Single Entry Free Space It was observed that: There was good agreement in the average [Grel + Diffraction] figure, though it was noted that the same deployment was used in all cases. The single entry to aggregate I/N ratio gradually increased, reaching around 9-11 db for the 75 base station deployment scenario. The area excluded using the average values were calculated and observed to increase with the number of small cells per macros cell as in Table This is the expected result: the higher the density of base stations the greater the aggregation effect and hence the larger the area that must be excluded to protect the ES. Scenario 1 small cells per macro cell 2 small cells per macro cell 3 small cells per macro cell Omni Antenna Directional Antenna Adjacent Band Analysis Table 12-12: Average Area Excluded In Section 6, analysis was made of the mobile networks transmit mask and receive satellite Earth station filter discrimination. It was noted that if there was no overlap between the wanted carriers of these two systems but minimal frequency separation (i.e. operating in the adjacent channel) then the discrimination available (calculated by integrating the TX and RX spectrum masks) was 9.5 db. With a co-frequency threshold of I/N = -10 db this would be adjusted to an adjacent band threshold = -0.5 db. In the most demanding case of the 3 small cells per macro cell using the omni antenna, the deployment probability was raised to 72 / 75 or 96%, while for the directional antenna case only a single deployment was infeasible (i.e. 98.7% of deployments feasible). Given the analysis was based upon a conservative assumption of a Gaussian filter at the Earth station, this suggests that for the scenarios considered there would be no significant constraints in mobile network small cell base stations operating in bands adjacent to satellite Earth station in dense urban environments Options for Mitigation Two mitigations were considered to facilitate sharing:

112 Geographic Sharing in C-band Page 112 Changing the satellite Earth station antenna for a larger antenna. Including site shielding around the satellite Earth station For the former the impact of increasing the dish size from 2.4 m to 3.6 m was analysed. Assuming the gain pattern in Rec. S.580, the side lobe is defined by the equation: G = 29 25log 10 θ This is therefore independent of the dish size and hence the interference would not change by increasing the dish size. What such a change would do, however, is improve the wanted signal that would result in an increased interference margin. The principles of the link design and impact of increasing the wanted signal are shown in Figure Increase in wanted signal RSL Target C/(N+I) + Possibly other margins C Target C/(N+I) + Possibly other margins N + I(agg) Interference margin = 1 db I(Agg)/N N + I(agg) N=kTB Increased interference margin Aggregate I Apportionment Rules Single entry I N=kTB Apportionment Rules Aggregate I Figure 12-19: Link Design and Impact of Increasing Wanted Signal A typical link design is: Single entry I Identify the receiver noise based upon ktb. Include 1 db of interference margin. Calculate the target receive signal level based upon the total noise + interference, the C/(N+I) needed to achieve the BER objectives and any other margins.

113 Geographic Sharing in C-band Page 113 From the interference margin the ratio of the aggregate interference to noise can be calculated. Using apportionment rules, the single entry (service and system) interference levels are calculated. If the power from the satellite is unchanged but the ES dish size is increased (for example, from 2.4 m to 3.6 m), then the wanted signal will be stronger due to the larger peak gain. If the noise level remains unchanged then this would provide an opportunity to increase the link interference margin while continuing to meet the BER objectives, as shown in the table below. Dish size (m) Peak gain (dbi) Interference margin (db) I agg/n (db) Table 12-13: Impact on I agg/n of Increasing Earth Station Dish Size It can be seen that increasing the dish size from 2.4 m to 3.6 m suggests that aggregate interference could be increased by 8.5 db. If it is assumed that the same proportion of the aggregate interference can be used by the mobile networks, that implies there would be the potential to increase the interference threshold from I agg/n = -10 db to -1.5 db. The impact on the average excluded area is shown below. Scenario 1 small cells per macro cell 2 small cells per macro cell 3 small cells per macro cell Omni Antenna Directional Antenna Table 12-14: Average Area Excluded with Additional 8.5 db Interference Margin The impact of there being 15 db of site shielding was also considered and it resulted in significantly lower average excluded areas as shown below. Scenario 1 small cells per macro cell 2 small cells per macro cell 3 small cells per macro cell Omni Antenna Directional Antenna Table 12-15: Average Area Excluded with 15 db Site Shielding For the directional antenna 3 small cells per macro cell case, if both mitigations were used the excluded area reduced to km 2 allowing average base station deployment to within 100m of the Earth station. These mitigations would have to be agreed between the Earth station and mobile networks operators but would provide significant potential opportunities to reduce the area excluded.

114 Geographic Sharing in C-band Page Summary The analysis suggests that the exclusion zone around an Earth station where IMT/LTE small cell base station would not be permitted could be small, possibly in the range km 2 without mitigation. The small size of exclusion zone is due to the mobile network base stations operating on lower power below the clutter in a dense urban environment where there will be significant off-axis gain and diffraction loss. The zone can be reduced using mitigation methods such as: Directional antennas at the base station Reducing the base station height above terrain Pointing the base station antenna away from the satellite Earth station Locating the base station in deep clutter maximising diffraction loss The analysis used the standard ITU-R diffraction model and a high-resolution surface database to generate predicted propagation loss: there would be benefit in measurements to verify this. One limitation in the use of the surface database is that it identifies trees as solid objects to be diffracted over rather than as an attenuating factor. One possible approach to facilitate sharing of this band by mobile networks would be to include in the licence terms and conditions the necessity of protecting a list of existing satellite Earth station. A suitable threshold would be either: Single entry I/N suitably adjusted for aggregation Aggregate I/N threshold The mobile operator could also offer to the earth station operator: Site shielding around the earth station; A larger replacement antenna with higher gain and/or lower far-off-axis gain values Either of these methods would significantly reduce the average area excluded. Aggregation interference was observed to be between 9-11 db higher than single entry levels. The analysis was undertaken for 1, 2 and 3 small cell base stations per macro cell. Similar behaviour was observed in each case, though the omni antenna deployment appeared to be reaching an interference driven limit in the number of base stations that could be deployed BT Tower Test Case This section describes analysis interference into a fixed link receiver from small cell mobile network base stations in a dense urban environment to identify the degree to which the deployment would be constrained by its presence. The objective was to gain an understanding of a single specific sharing scenario located where high-resolution surface data was available by modelling it in detail.

115 Geographic Sharing in C-band Page Fixed Links and Locations The assignment data provided by Ofcom was converted into a Visualyse Professional simulation file to check the data was complete and identify station locations. The resulting simulation file is shown in Figure Figure 12-20: Distribution of UK C-band Fixed Links It was noted that the deployment was dominated by two markets: Links connecting the Scottish islands: these benefit from the lower rain fade in C band (though there can be significant multi-path fading) and so can operate longer distances such as across the Minch. Low latency links, connecting data centres in and around London with international financial markets such as Brussels and to transatlantic cables along the south coast. There is likely to be low overlap between locations for which LTE / IMT deployment is required and the Scottish islands, and so attention focussed on areas identified in the land use database as: Buildings

116 Geographic Sharing in C-band Page 116 Urban Dense Urban These locations are identified in grey in: Figure 12-21, which also shows links around for London and the South East as lines, connecting fixed link stations between Slough in the west to Basildon in the east, with a spur heading south-east towards the Channel ports. Figure for locations near Manchester and Liverpool, which also shows three stations located between those two cities. Figure 12-21: Overlap of Fixed Link Locations with Likely London and South East IMT/LTE Deployment Zones Figure 12-22: Overlap of Fixed Link Locations with Likely Manchester and Liverpool IMT/LTE Deployment Zones

117 Geographic Sharing in C-band Page 117 In particular, three stations (Royal Free Hospital (to the north), the BT Tower and LSE) were located in central London as shown together with the surface database Figure Note that all links in this zone were licensed to a single organisation, namely Optiver Holding B.V. and the colour scheme used for the terrain / surface data is given in Figure Figure 12-23: C Band Fixed Link Stations in Central London The critical case was considered to be the BT Tower as: 1. It is located at the centre of a dense urban area that would represent a key mobile network market. 2. It is very tall, increasing the difficulty of sharing. The receive antenna at the BT Tower was therefore used as the basis of the analysis System Parameters IMT Parameters This analysis used the same parameters as given in Section 3 apart from the deployment, which was specified using 100 small cell outdoor base station deployed over an area calculated from the above to be a circle of radius 2.39 km based upon the BT Tower as shown in Table Macro base station separation distance (km) 0.6 Number of Small Cells / Macro base station 2 Area / Macro base station (km 2 ) 0.36 Area / Small Cells (km 2 ) 0.18 Number of Small Cells in simulation 100 Area of Simulation (km 2 ) 18 Radius of deployment zone (km) 2.39 Table Deployment Density for IMT-Advanced Small Cell Outdoor Systems

118 Geographic Sharing in C-band Page 118 Locations of the 100 small cell base station were randomised within a circle of radius 2.39 km around the BT Tower. They were then moved to the nearest street and for the measured base station data case their antenna was pointed along the street. No attempt was made to analyse interference between base station or the coverage / service provided. The resulting initial deployment locations are shown in Figure as small white dots within the circle of radius 2.39 km around the BT Tower. Figure 12-24: 100 Small Cell base station Located within 2.39 km of Victim Fixed Link Station The small cell characteristics were calculated within a bandwidth of 1 MHz and then scaled to the fixed link receiver s bandwidth of 30 MHz. This is equivalent to modelling the sharing environment as either: 1. At each location the entire fixed link bandwidth was used to provide mobile networks services. This could be because (for example) the site is a managed resource used by multiple operators or a single operator with access to the full 30 MHz. 2. Each location used just 10 MHz but there were an equivalent set of two x 100 base stations for the other 30 MHz which would average to be the same as for case 1.

119 Geographic Sharing in C-band Page Fixed Link Parameters The fixed link parameters were as shown below. Reference /1 Licence holder Optiver Holding B.V. Frequency (GHz) 3.8 Bandwidth (MHz) 30 Transmit station name LSE Tx longitude (deg) TX latitude (deg) Tx height (m) 50 Receive station name BT Tower Rx longitude( deg) Rx latitude (deg) Rx height (m) 50 TX power (dbw) TX peak gain (dbi) TX beamwidth (deg) 3.0 RX peak gain (dbi) RX beamwidth (deg) 3.0 RX feed loss (db) 1 Polarization RX Temperature (K) RX Noise (dbw/mhz) Table 12-17: BT Tower Link Parameters The receive temperature T was converted into noise in the wanted bandwidth = BW Hz using: N = 10log 10 (T) + 10log 10 (BW Hz ) This ensured consistency with Ofcom s planning process defined in TFAC OFW 446. The baseline runs for the project were agreed with Ofcom to be based upon the standard long-term threshold. However this section considers in detail a specific sharing scenario and examines all assumptions, including issues such as interference apportionment and aggregation issues. The single entry (single location) threshold was the long term I/N = db not to be exceeded for more than 50% of the time. Most UK fixed links have a receiver sensitivity level (RSL) associated with an interference margin of 1 db, as is standard industry practice. The interference margin can be converted to an aggregate I/N or DT/T using: H

120 Geographic Sharing in C-band Page 120 I agg N = 10log 10(10 IM/10 1) DT agg = (I agg/n)/10 T This implies there is a total of DT/T = 25.9% (here rounded to 26%) to be apportioned to various sources of interference, including: Other stations of the same service (e.g. other fixed links). Other existing co-primary services (e.g. satellite downlinks, constrained via the RR Article 21 PFD limits). Secondary services and those operating in other bands (out of band emissions). The single entry limit for other fixed links is I/N = db which equates to a DT/T = 6%. Hence, two single entry systems meeting this limit would take 12% of the total 26% available. With 1% each for secondary, OOB and 2% for other contributions, that leaves an aggregate I/N = -10 db for mobile networks which corresponds to a DT/T = 10%. Hence, a possible interference apportionment that includes the new service could be as shown below. Two other fixed links 12 % Aggregate from mobile networks 10 % Secondary services 1% Out of band services 1% Other 2% Total 26% Table 12-18: Interference Apportionment It should be noted that one method to facilitate sharing is to increase the interference margin to be above the industry standard 1 db. For example, increasing this to 2 db would result in a total interference allowance of 58.5%: with the same apportionment to other services this would make the mobile networks aggregate threshold an I/N = -3.7 db, 6.3 db higher. Conventionally, this approach would require a corresponding increase in the transmit power of 1 db in order to ensure the C/(N+I) remains above the threshold for the availability requested. However, that could introduce other burdens, as an increased transmit power would cause an increase in interference into other systems. An initial assessment was made of the interference environment by calculating the long term aggregate I/N at each of the fixed links from all the others. Links that used dual polarization were excluded due to modelling complexity, and the results on the remaining links are shown below:

121 Geographic Sharing in C-band Page 121 Total links 192 I/N not calculated 27 Remaining links 165 Aggregate I/N over -6 db 15 Aggregate I/N over db 36 Table 12-19: Intra-Service Aggregate I/N It can be seen that a significant number of links are already accepting higher levels of interference, which could lead to lower availabilities. Further analysis identified that this was the result of the initial modelling assuming all links were co-polar and hence not including any polarization discrimination. A subset of links were examined in detail and, when the effects of antenna polarization discrimination were included, all of these three had aggregate interference below the threshold: / / /2 Therefore, all the links examined were consistent with the Ofcom assignment criteria. The first of these three cases is shown in Figure where it can be seen that there is almost direct alignment between the victim and interfering links, both aiming at the same station. Figure 12-25: Link /1 and Worst Interferer Geometry The antenna code was identified as A/04/H/00/016/AA with gain pattern as in Figure

122 Geographic Sharing in C-band Page Gain (dbi) Offaxis angle (deg) Co-polar pattern Cross-polar pattern Figure 12-26: Fixed Link Receive Antenna Gain Pattern Propagation Modelling The same 3m resolution surface database was used as for earlier analysis together with the propagation model in P.452. In this case an adjustment was made to the surface database to remove the BT Tower which would be interpreted (incorrectly) as an obstruction by P.452. Note that the propagation model was configured with a percentage of time = 50% to be consistent with the fixed link receiver threshold rather than 20% as for the Earth station case. Consistent with the Earth station case, the propagation loss predicted by P.452 was decreased by 3 db to include multipath enhancements Mobile Networks Omni Antenna Identify Interfering Base Stations The aggregate I/N at the fixed link receiver was calculated taking into account the 100 small cell base station deployment, and compared against the single service aggregate stations interference threshold of I/N = -10 db. Initially this threshold was exceeded and so base stations were effectively switched off until the threshold was met. It was required to switch off 6 of the 100 so that 94 remained active. The locations of the excluded base stations are highlighted in Figure using the terrain / surface colour scheme given in Figure It was noted that: One base station was on a street in direct alignment with the BT Tower (Hampstead Road).

123 Geographic Sharing in C-band Page 123 Several of the base station were close to the BT Tower. All of the remaining excluded base station were to the east of the BT Tower, the direction the fixed link antenna was pointing. Figure 12-27: Excluded Base Station Locations Omni Antenna Average Gain and Diffraction Loss The case in which the small base station is on a street directly aligned with the victim receiver was treated as a special case, and so the average gain plus diffraction loss was calculated for the remaining 99 base stations. It was determined that the average (in absolute i.e. power terms) was 41.0 db. This was less than the equivalent for the Earth station case due to the lower path loss to the high fixed link receiver antenna. The aggregation factor from worst single entry to aggregate I/N was calculated to be 9.1 db.

124 Geographic Sharing in C-band Page Mobile Networks Directional Antenna Identify Interfering Base Stations The aggregate I/N at the fixed link receiver was calculated taking into account the 100 small cell base station deployment, and compared against the single interference threshold of I/N = -10 db. Initially this threshold was exceeded and so base stations were effectively switched off until the threshold was met. It was required to switch off 4 of the 100 so that 96 remained active. The locations of the excluded base stations are highlighted as blue circles in Figure It was noted that there were similar characteristics of the excluded base station locations as for the omni case. Figure 12-28: Excluded Base Station Locations Directional Antenna

125 Geographic Sharing in C-band Page Average Gain and Diffraction Loss The case in which the small cell base station was on a street directly aligned with the victim receiver was treated as a special case, and so the average gain plus diffraction loss was calculated for the remaining 99 base stations. It was determined that the average was 46.7 db, where the averaging was undertaken in absolute rather than db terms. This was less than the equivalent for the Earth station case due to the lower path loss to the high fixed link receiver antenna. The aggregation factor from worst single entry to aggregate I/N was calculated to be 5.0 db Average Excluded Zone Areas The average in absolute [gain + diffraction loss] was used to calculate average excluded areas using the parameters below. Antenna Omni Andrews Average [Grel + Diffraction] Single entry to aggregate I/N ratio db factor for multi-path enhancements Total adjustment to free space propagation loss Exclusion zone size (km 2 ) Table 12-20: Average Exclusion Zone Parameters The excluded zone area was the typical keyhole shape for the directional antenna case as can be seen in Figure Note this shows terrain in order to identify locations and scale: it was not used in the analysis which was based on adjusted free space path loss. The contour shows the aggregate threshold, i.e. I/N = -10 db as the adjustment included a factor to convert between single entry and aggregate I/N. Hence under these assumptions the single entry I/N threshold would be somewhere in the range -15 db to db.

126 Geographic Sharing in C-band Page 126 Figure 12-29: Example Average Excluded Area Zone These were noted to be significantly larger than for the Earth station average excluded area which were in the range km 2. Whereas the single-entry to aggregated ratios were similar, there were large differences between the average [Grel + Diffraction] between the two scenarios. The main reason for the difference was the geometry as shown in Figure 12-29: The Earth station has higher diffraction, being located close to the clutter line, and greater antenna discrimination due to it using a positive elevation angle. The fixed link has lower diffraction, being located significantly above the clutter line, and lesser antenna discrimination due to it using a low, possibly negative, elevation angle.

127 Geographic Sharing in C-band Page 127 Earth Station Elevation > 0 Fixed Link Receiver Fixed Link Elevation < 0 Lower diffraction and decreased relative gain Higher diffraction + increased relative gain Earth Station Summary Figure 12-30: Comparison of Earth Station vs. Fixed Link Geometry The interference zone around the fixed link receiver on the BT Tower was found to be on average km 2 in size. This zone was calculated using average gain + diffraction excluding any terrain effects. The zone can be reduced using mitigation methods such as: Directional antennas at the base station Reducing the base station height above terrain Base Station Pointing the base station antenna away from the fixed link station Locating the base station in deep clutter maximising diffraction loss It could also be useful to analyse the impact of traffic models, fading of the interfering signal and polarisation effects. However, the average interference zone contained many locations where a base station could be located due to higher diffraction losses. This could facilitate operation very close to the fixed link station: indeed for the deployment of 100 base stations considered, between 94% and 96% were found to be located in positions that had sufficient diffraction to protect the fixed link receiver. In particular, a key factor was the fixed link antenna receive gain, and hence in locations not in its main beam there was high likelihood of ability to deploy small cell base stations. It is also noted that for the omni antenna case the aggregate I/N = -0.3 while for the directional case it was -2.3 db. With an adjacent band discrimination or NFD of 23.5 db, this means that all of the base station locations considered could operate without causing harmful interference when transmitting non-cofrequency.

128 Geographic Sharing in C-band Page Power and Traffic This section describes analysis of using Monte Carlo analysis to convolve the effects of: Traffic and Power Variation in traffic levels, typically over a day Variation in propagation, typically over a year Random polarisation loss The parameters below were used in the JTG analysis for small cell base stations. Maximum power (dbm) 24 Bandwidth (MHz) 10 Average power (dbm) 21 Table 12-21: JTG Small Cell Power Parameters It is worth re-addressing whether these are valid, in particular to assess the time period over which the power is to be averaged. The power level is used in interference analysis with propagation models such as P.452 which has an associated percentage of time. The resulting I/N values are then compared against long term thresholds, such as 20% of the time. A key question is: what does 20% of the time mean? Propagation models such as P.452 generate statistics that are valid over very long sample times taking samples at regular intervals. All times of day are sampled equally to generate models that are valid over a year, and hence the power average should also be considered over a year. In particular, 20% of the time does not mean 20% of the mobile networks busy hour during the working week. However JTG Document 715 Annex 2 on LTE parameters specified typical average activity of a base station and corresponding average output powers during busy hour as being 50% of maximum load. It is accepted that small cells have extremely low levels of traffic for significant periods of the day, in particular night-time. There has been research recently 8 into the power savings that mobile operators can achieve by completely switching off the small cell base stations for these periods, as the macro cells: a) Have to be active anyhow to handle vehicular traffic; b) Have sufficient capacity in quiet times to also handle pedestrians even in locations that are daytime traffic hot spots. It is worth considering whether the average power is taking into account this period of potentially negligible activity. 8 For example, see the paper Multiple Daily Base Station Switch-Offs in Cellular Networks by Marco Ajmone Marsan, Luca Chiaraviglio, Delia Ciullo and Michela Meo.

129 Power wrt peak exceeded for given hours Geographic Sharing in C-band Page Traffic is by nature stochastic and will be sized by expected peaks. This means the cell would only approach maximum power during local busy hour(s). This is compounded by the motivation for opening up the C-band, namely to relieve congestion and reduce the likelihood of users being unable to access mobile broadband services due to an overloaded network. Traffic is also often highly bursty and there will be other constraints (handset performance, user interactions and wider web factors including latency) that would mean that even during busy hour power is likely to reach maximum for only short periods of time. Furthermore, as noted by Report M.2241, transmitting 100% time in 100% of frequency resources (in the case of OFDMA) means saturation of the cell and service failure for many of the users. A key question is how many much of the day the traffic level is close to that of the busy hour, and conservative assumptions are that this could be either 8 or 4 hours. Two traffic profiles considered in the analysis are shown in Table and Figure These profiles were assumed to be averages and hence could be applied over all days of the week over the complete year. Hours per day Traffic Model 1 Traffic Model db to -3 db 0 db to -6 db 4-3 db to -6 db -6 db to -6 db 4-6 db to -10 db -6 db to -10 db 4-10 db -10 db to 20 db 8-20 db -20 db Table 12-22: Proposed Traffic Models Hours of day Model 1 Model 2 Figure 12-31: Proposed Traffic Models

130 Geographic Sharing in C-band Page Propagation and Polarization 140 Study of the C-band scenario in the JTG has been undertaken using P.452 as the propagation model. This is a well-established propagation model for interference analysis but is not suitable for Monte Carlo analysis due to having an upper limit percentage of time of 50%. For this reason Ofcom has in recent years undertaken significant research into developing the concepts within P.452 further so that the full range of percentages of time can be modelled, namely [0, 100%]. This resulted in propagation model Recommendation ITU-R P.2001 and a recent revision. The benefit of using this model is it can take account of the cases where the small cell base station would be at maximum power but the signal is faded due to (for example) rain fade. It should be noted that for the short paths considered for small cell base stations there is not a great variation between short percentages of time and long time. For example Figure shows the variation for a 10 km path at 3.8 GHz, and the difference between p=20% and p=0.01% of time is 6.9 db. P.452 path loss for associated percentage of time Percentage of time (%) Figure 12-32: Variation in P.452 Propagation Loss for Short Path When undertaking Monte Carlo analysis it is also worth considering other factors that could vary. An example would be polarization which is different between satellite services (typically circular polarization) and terrestrial (typically linear). Even between two linear polarized services there will be a degree of de-polarization that will lead to the interference detected being reduced compared to a fully co-polar scenario. This could (for example) lead to decrease in interference of between 0 db and 3 db.

131 Percentage of time I/N exceeded Geographic Sharing in C-band Page Monte Carlo Analysis Monte Carlo analysis was undertaken using the parameters in Table Below. Traffic Models As per Table Propagation Model P.2001 Polarization Variation Maximum Power Base station Height Separation Distance Earth Station Height Uniform [0, 3] db 14 dbm / MHz i.e. -16 dbw / MHz 6m 10 km 3m Earth Station Antenna 2.4 m Rec. S.580 Earth Station Elevation 10 Earth Station Pointing Earth Model At base station Smooth Clutter Loss 100 Table 12-23: Monte Carlo Analysis Parameters The resulting I/N CDF is shown in Figure below: I/N (db) Traffic 1 Traffic 2 Figure 12-33: Monte Carlo I/N Analysis The results are shown in the following table, which also shows the transmit power that would generate the I/N (20%) assuming static analysis using P.452.

132 Geographic Sharing in C-band Page 132 Traffic Model 1 Model 2 Percentage of time I/N = -10 db 46.7 % 44.4 % I/N exceeded for 20% of the time -4.6 db -6.4 db Equivalent power for P.452 (20%) dbw/mhz dbw/mhz Reduction compared to peak power -6.1 db -7.9 db Table 12-24: Monte Carlo Run Results This was less variation that might have been predicted with a large (20 db) difference between maximum and minimum power. However, with a relatively small variation in the propagation loss, the power levels for percentages of time around 20% become more important. With these traffic models the 20% of time power were just 3.6 db (Model 1) or 6 db (Model 2) down from peak Summary Monte Carlo analysis was undertaken that convolved: Variation in base station transmit power due to possible traffic variation over the whole day Use of P.2001 propagation model using full percentage of time range [0, 100] Polarization loss with range [0, 3] db This suggests that between 6.1 and 7.9 db of interference mitigation could in theory be achieved compared to assuming peak power by considering variations in traffic, propagation and polarization effects Conclusions The core work of this project was the development of maps and statistics relating to availability of spectrum for mobile networks derived using UK wide terrain and land use databases. This section describes work analysing in detail central London using a high-resolution surface database, which suggested that these main maps are likely to be conservative in dense urban areas as the propagation loss is likely to be higher, as described in Section There is significant potential to deploy low power low height base stations very close to satellite Earth station and reasonably close to point to point fixed link receivers located in dense urban areas. The key is to ensure there isn t line of sight between interferer and victim stations, and preferably a significant degree of diffraction loss due to buildings. This would impose constraints on where the mobile operator could deploy their base station, requiring them to (for example) choose the side of the street with the greatest radio shadow. However there would be significant benefits in terms of the number of locations that could be served. There are likely to be some geometries for which this approach would be unavailable, in particular streets that point directly at receiver stations. Furthermore, it was found to be harder to share with fixed link receivers than satellite Earth station due to their antenna height and pointing direction.

133 Geographic Sharing in C-band Page 133 Monte Carlo analysis was also undertaken that suggested that the average power used in studies could be reduced further to take account of time of day variation of traffic. Mitigation options were considered including increasing dish sizes and for the Earth station case including site shielding. Both could significantly reduce the area excluded but would require the cooperation of the victim licensee.

134 Geographic Sharing in C-band Page ANNEX: NATIONAL MAPS FULL SET OF DATA 13.1 Sharing with Satellite Earth Stations and Fixed Links Combined National Maps Figure 13-1: Colour Coded Map of Spectrum Available in Sharing with Satellite Earth Stations and Fixed Links in GHz

135 Geographic Sharing in C-band Page 135 Figure 13-2: Colour Coded Map of Spectrum Available in Sharing with Satellite Earth Stations and Fixed Links in GHz

136 Geographic Sharing in C-band Page 136 Figure 13-3: Colour Coded Map of Spectrum Available in Sharing with Satellite Earth Stations and Fixed Links in GHz Spectrum Available by Population Tables The following tables refer to the spectrum available by population in urban, dense urban and hot spot areas only (accounting for million people in our model).