7.1 Overview of approach to calculating benefits

7 CALCULATION OF BENEFITS 7.1 Overview of approach to calculating benefits This section sets out an overview of the approach taken to estimate the economic benefits which could arise by introducing higher power for licence-exempt WBA in Great Britain. These benefits need to be balanced with the costs in order to determine the overall benefit to the UK of permitting higher power of licence exempt bands. The assessment of costs of interference is discussed in sec. 8 7.1.1 Overview of the methodology There are three elements of the approach taken to estimate benefits which, when combined, produce a representative business model of the demand for rural broadband services using WBA, and an estimate of the consumer surplus associated with such a service. These elements are: 1. A model of the demand for WBA; 2. A geographical model of the likely target market for WBA; and 3. A WBA network model of the cost of providing WBA services. As the aim is to provide an estimate of the benefits arising from WBA (i.e. the consumer surplus), the models centre around the demand model. This section sets out the demand model for WBA and describes how it is used for calculating benefits, referring as necessary to the geographical and costs models. 7.1.2 Calculation of economic benefits The benefits from WBA were estimated as the consumer surplus enjoyed by those consumers who take-up the service. The consumer surplus of higher power licence-exempt WBA represents the additional value (in GBP) for UK consumers and businesses that would be created under the following scenarios: Frequency (GHz) Higher power scenario 2.4 1W EIRP 2.4 10W EIRP 2.4 80W EIRP 5.8 4W EIRP 5.8 25W EIRP 5.8 200W EIRP 3.5 23 125W EIRP Table 8 Higher power scenarios 23 3.5GHz is a licenced band for WBA. We have used the same methodology for estimating benefits at 3.5GHz for comparison with the licence-exempt bands. C13400 013 1510 Final Report v1-0.doc Page 51

The benefits obtained from permitting the higher powers shown in Table 8 are assessed using current regulatory limits as the status quo and calculating the additional benefits to consumers of new services made possible under the higher power scenario. For the purposes of the calculation it has been assumed that the benefits of higher power WBA are realised by enabling customers (residential and business) to receive a broadband service more cost effectively by WBA than by alternative means available to them. This has two important consequences for the modelling approach: 1 No benefits from WBA arise where customers are able to receive a DSL or cable broadband service of acceptable quality. Industry experience shows that WBA is invariably more expensive than an equivalent DSL or cable broadband service, hence there is no economic benefit to be gained by supplying WBA in areas where DSL or cable broadband are available. 2 In areas where DSL or cable broadband are not available, satellite is likely to be the only option available to consumers. Thus, benefits to customers were calculated as the consumer surplus enjoyed by broadband subscribers outside DSL and cable broadband areas by taking WBA services. For those consumers who currently subscribe to satellite broadband services this is simply the money saved by switching to WBA. In addition, lower price broadband services will attract new customers who would not be willing to pay the additional premium for satellite services. These consumers do not benefit by the full amount of the cost saving of WBA over satellite broadband, rather they benefit by the additional amount they would be willing to pay for broadband than they have to pay for WBA. Service choice The WBA services to be modelled have implications for the cost of providing services, which in turn feed into the price that can be charged for a viable WBA service. While different consumers subscribe to broadband services with different features, it is possible to model demand and service costs using representative typical services. In this report the following WBA services have been adopted: Residential service Business service 1Mbit/s SDSL 50:1 contention 2Gbit/month download limit 24 4Mbit/s SDSL 10:1 contention no download limit 24 Whilst this is a typical limit it does not affect our cost model. C13400 013 1510 Final Report v1-0.doc Page 52

In our view these are appropriate WBA services which will be relevant over the next 3-6 years. The key assumptions underpinning this choice are: 1 Within the 6 year timescale of our model, target WBA users are those who either cannot get DSL at all, or who can only obtain DSL of a degraded quality. Thus for residential users a speed of 1 Mbit/s will be acceptable. 2 The main broadband application for residential users in rural areas over the next 6 years will remain web browsing and downloading of small to medium sized files, and is unlikely to include services such as high quality video streaming that require download speeds in excess of 1 Mbit/s. 3 Business users will require speeds higher than residential customers, because they typically need to transfer larger files, handle multiple users and are more sensitive to delays in sending and receiving information. A 4 Mbit/s connection will be adequate for most typical business users. 4 In rural areas where DSL is available, speeds are likely to remain limited due to the limitations of the local loop. Although BT may enable most or all of its exchanges in the near future many residences and businesses are too far from their local exchange to be able to obtain a DSL connection of sufficient quality. 5 Actual WBA services are likely to be symmetric, and (where there is footprint overlap) will mostly compete with asymmetric DSL services. WBA operators will need to sell the benefits of symmetric services to compensate for the likely higher price of WBA (even in areas where DSL is unavailable or low quality). 6 Demand of broadband services is not dependent on network architecture, rather demand is common across all potential consumers, therefore residential and business users will require contention ratios and download limits comparable to equivalent DSL services currently provided. 7.1.3 Geographic modelling the addressable market for WBA As WBA services are likely to cost more than the equivalent DSL or cable services the target market for WBA suppliers will be rural consumers and businesses who either cannot get DSL or cable broadband at all, or who can only get these services at unacceptably low-speeds. Since in Great Britain DSL coverage exceeds and encompasses cable coverage, this means that the target markets are: Residences beyond the range of 1Mbit/s DSL from the local BT exchange; and Businesses beyond the range of 4 Mbit/s DSL from the local BT exchange. C13400 013 1510 Final Report v1-0.doc Page 53

The addressable market for WBA was addressed using the following geomodelling approach: 1 Great Britain was overlaid with a set of tangential interlocking hexagons, each representing a radio cell with the WBA base station assumed to be at the centre 2 The cell radius was dependent on the power scenario (higher power equated to a larger cell radius). Thus fewer cells were required to cover the UK for higher power levels 3 For each cell the total number of residences and businesses lying within the boundary of the cell was determined 4 The residences and businesses in each cell which were within range of a BT exchange for provision of 1Mbit/s and 4 Mbit/s DSL service, respectively, were excluded. This provides an estimation of the addressable WBA market (numbers of residences and businesses) for each cell, in each power scenario. As pointed out above, higher powers relate to large cell sizes, and therefore fewer cells required to cover Great Britain. The power scenarios and corresponding cell sizes used were as follows: Frequency (GHz) Scenario Cell radius (km) Number of cells 25 2.4 100mW EIRP (Current situation) 1.75 20,475 2.4 1W EIRP 3.50 6,841 5.8 2W EIRP (Current situation) 3.50 6,841 5.8 4W EIRP 4.25 4,918 2.4 10W EIRP 7.25 1,922 5.8 25W EIRP 7.25 1,922 2.4 80W EIRP 16.50 451 3.5 125W EIRP 16.50 451 5.8 200W EIRP 16.50 451 Table 9 Scenarios (detail) With this method it is important to note that: Since the line length (BT exchange to customer) over which 4Mbit/s can be provided is much shorter than for 1Mbit/s service, the area covered by DSL services and therefore not considered part of the addressable WBA market, on the basis that cheaper DSL will be available, is greater for residences than for businesses. The addressable market does not include the possibility of providing WBA to customers who can receive DSL. In practice, WBA operators would be able to compete with DSL or cable modem where network footprints overlap, and 25 Number of hexagons required to cover GB at the specified cell radius (i.e. excludes Northern Ireland) C13400 013 1510 Final Report v1-0.doc Page 54

would be expected to capture some customers in DSL areas with targeted services (e.g. serving a business park or providing a specialist gaming service). However, unless WBA can be offered at a price lower than DSL, consumers would not benefit from any such competition. Furthermore, unless WBA providers are able to differentiate consumers who can obtain DSL services from those who cannot and are able to offer each type of consumer a different price, then attempting to compete for DSL consumers on the basis of price is likely to be highly unprofitable. This latter point indicates that the modelling approach taken is conservative, in that the WBA market opportunity in practice is likely to be greater than that modelled. Determination of WBA subscribers WBA pricing is set by applying a premium to benchmark DSL pricing, which is justified by the fact that DSL is not available to these subscribers and is in line with market practice: Residential Non-residential 35 per month for 1MBit/s residential service (symmetrical) 26 150 per month for 4Mbit/s business service (symmetrical) These prices maximise consumer surplus and the number of economic subscribers within the accuracy of the model. This approach can be justified because a competitive market can be expected to maximise consumer surplus for those areas where WBA is financially viable. Given the proposed licence exempt status of WBA and the relative freedom of access to equipment, one can expect the market to be competitive. However, care should be taken in interpreting any benefits where benefits are either close to, or exceed, the cost of interference. The projected number of WBA subscribers in each cell was determined using the calibrated demand curve from the demand model to estimate the maximum number of subscribers at full penetration then applying a take-up curve to determine the addressable market for each year considered 27. 26 This includes installation, i.e. there is no separate installation charge 27 Further details on can be found in Annex E: Methodology for cost benefit analysis C13400 013 1510 Final Report v1-0.doc Page 55

7.1.4 WBA cost modelling The cost of providing WBA service to the projected WBA subscribers in the geomodel cells was estimated using a simple WBA network roll-out model. This was developed and then applied to the geo-model on a cell-by-cell basis. This model considers two main systems: 2.4GHz 5.8GHz Using 802.11g equipment. This is likely to be the most common 802.11 standard over the modelling period, with worldwide shipments reaching ca. 48 million units per year 28. Using WiMAX equipment (802.16d). This is likely to provide the most cost-effective solution for WBA in the medium term, as it is an accepted standard for operation of fixed wireless access in the 2-11 GHz bands. Worldwide shipments are forecast to reach 230 million units per year 29. The topology used for the radio system in the model was point-to-multipoint, which is the most prevalent WBA system architecture. Whilst other architectures (particularly mesh networks) may have the potential to reduce operating costs of WBA service provision in certain circumstances, PmP is the most widely used system today and is assumed the most appropriate for sparsely populated communities and hence is considered to be more appropriate for the base case. As a comparison to the licence exempt 2.4GHz and 5.8GHz systems a 3.5GHz licenced point-to-multipoint system was also considered. 28 Fixed Wireless, WiMax, and WiFi Market Opportunities, Market Forecasts, and Market Strategies, 2005-2010, Wintergreen Research 29 Ibid C13400 013 1510 Final Report v1-0.doc Page 56

The network model defines the radio network equipment requirements for each cell, to provide service. In order to do this assumptions are made covering the following key areas: Assumption Notes Main source(s) Bandwidth requirement The cell bandwidth needed by each subscriber (residential & business) Growth in bandwidth requirement ISP benchmark figures Radio equipment Cell backhaul Transit Network operation ISP operation Base Station and CPE costs Number of sectors per base station Sector capacity Point to Point radio backhaul from each cell, dependent on the total cell bandwidth requirement: 5.8GHz licence exempt 18 GHz licenced Average cost of transit from an aggregation point (multiple cells) to a backbone network Includes peering (i.e. an uncontended connection to the internet) Base Station site rental Network operating cost (including maintenance) Web hosting Billing Staff & office costs Marketing costs Interviewees Industry price benchmarks Industry operating benchmarks Industry price benchmarks BT Central Internet pricing ISP benchmark figures ISP benchmark figures Table 10: Network model assumptions The model sizes the radio network based on the number of subscribers (and hence bandwidth requirement) after 6 years operation, the end point of our business case scenario. The bandwidth requirement for each cell determines: Number of Base Stations and Sectors Cell backhaul type 5.8GHz licence exempt (10 Mbit/s) or 18 GHz licenced (100 Mbit/s) Transit type (4 Mbit/s, 34 Mbit/s or 100 Mbit/s leased line). C13400 013 1510 Final Report v1-0.doc Page 57

The WBA cost model allows some headroom in cell backhaul and transit. For cell backhaul, cell bandwidth is rounded up to determine the backhaul type. For transit, a conservative degree of aggregation is assumed (1-4 cells), again rounding up to provide a safety margin for transit capacity. All costs are allocated throughout on a per-cell basis (including ISP operating costs). Thus capex and opex per cell can be determined. Depreciating the capex in line with industry practice allows comparison of annual revenue with annual operating costs for each cell, which in turn allows the economic viability of each cell to be determined. 7.1.5 Derivation of overall economic benefit In a competitive environment with profit motivated firms one can only expect roll out of WBA to those cells that provide a positive return for investors. The financial viability of cells was assessed on a cell-by-cell basis and only the cells, for which there was a positive return, were included in the calculation of benefits. Then, aggregating demand across all financially viable cells provides an estimate of the total demand for WBA. The total additional consumer surplus from WBA can then be estimated from this total demand, the price of WBA and the price of satellite broadband services, giving the benefit realised under each scenario. Since the model covers scenarios at the current power regulations, the impact of increasing power can be quantified and plotted as net benefit (increase in consumer surplus) versus power increase (EIRP). C13400 013 1510 Final Report v1-0.doc Page 58

7.2 Geographic modelling 7.2.1 Overview The objective of the geo-modelling is to calculate the addressable market, used for calculating benefits, for a set of theoretical scenarios for WBA service over Great Britain. The addressable market is dependent upon: The capacity of ADSL to provide the services beyond a specified distance from an exchange. The coverage achievable by a single WBA basestation. The coverage is related to the maximum power that may be transmitted by basestation and CPE. The approach taken has been to build cellular based coverage over the whole country for each scenario. The cost of provision is then evaluated for each scenario along with the coverage and revenue obtained. The major competitor to this service is wired ADSL. An important assumption behind the model is that WBA cannot compete with ADSL services. The basis of this assumption is that over the past 10 years there have been several wireless operators aiming to provide fixed wireless broadband services on a commercial basis (e.g. Invisible Networks, Atlantic Telecom). These companies are no longer offering WBA but adoption of broadband has continued to rise primarily through adoption of ADSL. The following sections describe the approach taken to: Identifying those residences and business that are unlikely to obtain broadband services via ADSL Modelling the coverage of WBA for the various scenarios Note that the model will not necessarily only identify rural residences and businesses that could benefit from WBA. A benefit may arise where WBA can provide broadband to any residence or business that cannot obtain the service via ADSL wherever they may be located. The areas where the benefits arise, including identifying rural areas that benefit, are described in sec. 7.5.4. 7.2.2 Identifying the unserved residences and business BT have provided figures for the proportion of UK properties it thinks should be able to obtain ADSL broadband following its Extended Broadband Reach trials in Milton Keynes, Fort William and Dingwall in 2004. It removed its 6km 30 restriction on ADSL as a result of the trials. BT expected that it should be possible to serve 30 BT do not state whether this is crow flies or line length although from our analysis it appears to be line length C13400 013 1510 Final Report v1-0.doc Page 59

99.8% of properties with ADSL at 512kbit/s and higher 31 It also expected that 96% of homes would be able to receive ADSL at 1Mbit/s and higher. However, for the purposes of modelling the benefits of high power WBA across the UK it is necessary to know where the properties, both residential and business, are located. To produce the database of Unserved properties the following source information was used. SEERL database of GB postal delivery points. This identifies for every postal delivery point in Great Britain its geographic location and classification between residential and non residential (business). A database of the postcode of every BT telephone exchange in the U.K. Technical specification of the DSL standards on the maximum usable speed at different line lengths. 32 The summary results of the BT extended reach ADSL trials conducted in 2004. Note that Hull was excluded from the GB postal delivery points database because the exchange database does not include Hull The first stage was to obtain from the SERRL database and the telephone exchange postcodes the location of all the telephone exchanges in Great Britain. Then the distance from every address to the nearest telephone exchange was determined. This is the crow flies distance. A probability distribution function of the distance of an address from its nearest telephone exchange was calculated. This probability distribution was compared BT s information regarding line length and proportion of population served. For residences the desired speed is 1Mbit/s downlink. From the ADSL Forum information for Extended Reach ADSL this corresponds to a line length of 6km. As stated above BT estimated that 96% of the population can obtain 1Mbit/s or higher. From the crow flies probability distribution it is estimated that 96% of the population live within 4kms of an exchange. Therefore, the link length to crow flies relationship at 6km from an exchange is 1.5:1. For businesses, the desired speed is 4Mbit/s downlink. Using the ADSL Forum information for ADSL2 this corresponds to a line length of 3.3km. ADSL2 is beginning to be employed and hence it is appropriate to use technical data relating to this technology over our study period. Although there is no data from BT about the proportion of the population that can achieve 4Mbit/s ADSL, as there is for 1Mbit/s, it is assumed that the line length to crow flies relationship will be 10% higher at the shorter distance i.e. 1.65:1. Therefore, at 3.3kms line length 31 Source: MK Extended Broadband Reach Trial Results, Questions and Answers from BT, August 18 th 2004 32 ADSLforum website C13400 013 1510 Final Report v1-0.doc Page 60

the crow flies distance will be 2kms. The higher ratio between the line length and distance is justified by the fact that the telephone lines generally lie alongside streets and roads. The route between an exchange, which will usually be located in the middle of a town, and a residence or business on the edge of a town or in the countryside is more likely to more direct than the route to a residence or business within the town. These results have been tested against a small number of addresses and telephone numbers taken from the telephone directory that are approximately 4km crow flies from the exchange. All were offered 1Mbit/s and not 2Mbit/s by the BT online DSL availability checker (all checks were performed using the telephone number rather than the postcode). It is believed that the BT system uses a database of actual lengths lines where a telephone number is entered but uses a geographic estimate where a postcode is entered. A database has been created that contains all businesses are that are unserved i.e. more than 2km from their nearest exchange and all residences that are are unserved i.e. more than 4km from their nearest exchange. Using this technique the proportion of GB residences that can obtain 512kbit/s has also been calculated. This result and the calculation of the proportion of residences that can obtain 1Mbit/s ADSL have been compared with BT s statements about ADSL coverage (assuming all exchanges are broadband enabled) and other information from Analysys 33. See Table 11. Calculated result BT Statements Analysys Report 512kbps 99.8% 99.8% 99% 1Mbit/s 98.9% 96% 92% Table 11 ADSL availability as reported by different sources. Note: Some approximations have been made to perform this comparison. Only 99.6% of BT exchanges are enabled for DSL. Therefore, the calculated result includes 0.4% of subscribers as served when they are not. Not all addresses are connected to their nearest telephone exchange and so some premises may be further away from their exchange than predicted, which would result in excluding more premises than would be appropriate. However, this is most likely to affect premises on the boundary between two exchanges, rather then those close to an exchange. Therefore this should have little impact on our modelling. 33 Rural Broadband: New Fault Lines Appearing?, Analysys, Aug 2005 C13400 013 1510 Final Report v1-0.doc Page 61

Figure 5 Unserved Businesses C13400 013 1510 Final Report v1-0.doc Page 62

Figure 6 Unserved Residences The unserved area, Figure Figure 5 and Figure 6Figure, is greater for businesses than residences because the access speed is higher. However, there are fewer unserved businesses (272,547) than residences (343,942). Hence, the business density population is lower. 7.2.3 Identifying the difficult to serve areas The next phase of the modelling approach was to identify what are termed difficult to serve areas. These are geographic areas where the population that can be served by one broadband wireless radio basestation is very low and consequently the business case for installing such a basestation is poor. The purpose of this undertaking was to examine the range of scenarios (cell sizes and powers) that should be considered when calculating benefits. A model of a uniform radio network that covers the whole country was constructed. This is modelled as a uniform array of fixed size hexagons. By changing the size of the hexagons it was possible to model different radio parameters. C13400 013 1510 Final Report v1-0.doc Page 63

The first step was to determine roughly the number of residence and business subscribers that need to be served by a basestation to justify the costs of that basestation. The assumptions (in italics) made regarding revenues and costs of service provision are shown in Table 12. It was estimated from these assumptions that a basestation must support 48 residences or 15 businesses to cover its costs. Years Residential Business Base Station Revenue Revenue per month 28 70 Revenue (p.a.) 336 840 Costs Peering (per customer p.a.) 50 100 CPE Capex 400 500 Amortised 4 100 125 Net Revenue (p.a.) 186 615 BS Capex 30,000 BS Backhaul Capex 4,000 BS Opex 4,500 Amortised 8 8,750 Table 12: Revenue and costs assumptions made for estimating difficult to serve areas Assuming a typical take up of 30% this becomes a coverage requirement of 150 residences or 45 businesses (or a linear combination of the two) required to be covered by a radio basestation for it to be to be financially viable. This assessment was performed before the detailed network modelling (sec. 7.3) and demand modelling (sec. 7.4) had been undertaken in order to estimate the range of cell sizes that needed to be modelled later. As the intention is to provide coverage of as many properties as possible, the cell size was increased until only a small proportion (<10%) of the hexagons cover too few addresses to be financially viable. These hexagons are our difficult to serve areas. This is repeated for residential only, business only and the combination of the two to determine if the type of customer influences the geographic areas that are difficult to serve. This resulted in a target cell radius of 16.5kms. C13400 013 1510 Final Report v1-0.doc Page 64

Figure 7 Difficult to serve areas (business only). Key: Yellow hexagons indicate financially unviable cells. Using the geographic model to find the least viable areas on this basis showed that the highlands of Scotland dominated the 10% of ground area that is difficult to serve. The analysis was repeated 4 times using a random hexagon array origin to ensure that the model is not too sensitive to the origin of the hexagon overlay. Although some of the specific areas that became financially viable at 16.5kms cells size changed no new regions of the country were added to or removed from the results. C13400 013 1510 Final Report v1-0.doc Page 65

Figure 8 Difficult to serve areas from the 4 separate runs of the model To achieve a cell radius of 16.5km would require a significant increase in power over current licence exempt use although it is the typical operating range of the current 3.5GHz BWA licence predicted using the Okumura propagation model. This model is likely to be optimistic in the Scottish highland environment suggesting that even the current licenced powers may not provide economic coverage in these areas. 7.2.4 Modelling different RF radio regulations The previous stage of analysis provided an upper bound on the cell size. The final stage of the modelling provides the information needed to calculate the size of the addressable market for each frequency and power scenario. A database was constructed for each scenario containing a list of radio basestations along with the number of residences and business addresses served by each one. These tables are obtained using the same type of hexagon array used for the difficult to serve area determination. The size of hexagon is chosen to match various power limits. The relationship between cell size and power is discussed in.the following section. For each scenario a table has been generated with a list of basestations locations and separate numbers for the C13400 013 1510 Final Report v1-0.doc Page 66

residential and business addresses served by that basestation. For a selection of the scenarios the results have been plotted, Figure 9 2.4GHz scenarios of 100mW EIRP and 10W EIRP Figure 10 5.8GHz scenarios of 2W EIRP and 4W EIRP C13400 013 1510 Final Report v1-0.doc Page 67

In the plots above, the colours represent ranges of numbers of delivery points within each cell. Where there is no cell outline shown there are no delivery points which can be served. This is because either there are no delivery points within the cell (e.g. the Scottish Highlands, northern Scotland, parts of mid-wales) or all delivery points within the cell are served by ADSL (e.g. central London). 7.2.5 Modelling the relationship between cell size and power Various assumptions have been made in order to estimate the range of operation of a PmP wireless broadband access scheme. These include the antenna gains and the sensitivity. The sensitivity of the system has been calculated from first principles assuming no benefit from processing gain. 17MHz bandwidth and 12dB S/N ratio have been assumed. The use of directional antennas to connect the home user to the base station will reduce the overall power needed and may also act to partially alleviate multipath problems. Due to the shorter wavelength, it is assumed that slightly more directional antennas can be used at 5.8GHz than 2.4GHz. The assumptions made in the system design therefore already include a relatively conservative system sensitivity which may be exceeded depending on the final technology choices. A further 10dB fade margin has been included in the range estimations. It is assumed that CPE antennas are externally mounted. The model assumes single carrier modulation although it is recognised that OFDM modulation can provide some benefits over single carrier in multipath environments. These assumptions have been made to support the generalised model used for the cost benefit analyis and is not intended as a basis for actual network rollout. C13400 013 1510 Final Report v1-0.doc Page 68

The overall system link budget is given by L MAX (db) = P + G T + G R (RX SENSITIVITY + F M ) where L MAX (db) P G T G R F M RX SENSITIVITY Maximum link budget, db Power transmitted, dbm Gain of transmitting antenna, db Gain of receiving antenna, db Fade Margin, db Receiver sensitivity, dbm The following table shows the figures used for the analysis. Parameter 2.4GHz System 5.8GHz System Base Antenna Gain 10 db 12 db CPE antenna Gain 10 db 12 db Device Bandwidth 17 MHz 17 MHz Device S/N requirement 12 db 12 db Device overall N/F 2 db 2 db Fade Margin 10 db 10 db Derived Sensitivity -87.7 dbm -87.7 dbm Example Link Budget for 0dBm conducted power at the transmitter 97.7 db 101.7 db Table 13: System parameters used in propagation analysis The transmitted power required for a given range of communication and reliability depends on the link budget. Many approaches are possible to estimating radio link budgets and all include wide variations between estimated and actual results because radio propagation varies widely with terrain and conditions. C13400 013 1510 Final Report v1-0.doc Page 69

Propagation Models The most accurate radio propagation models are based on knowledge of actual terrain data and take the shadowing and diffraction caused by the terrain into account. However, these models provide an estimation that is only valid for the particular terrain covered by the data. Other models take a theory based approach being purely mathematical. Finally there are results-based models which are the result of significant amounts of research. As such they represent typical behaviour over generalised terrain. Radio propagation is such that over any two similar distances the link attenuation will vary significantly both due to large signal variations, for example due to hills or buildings in the way, and due to small signal conditions such as multipath. Okumura s model is an appropriate model to use for achieving the objective of generating an overall understanding of the benefits from a regulation change. It is a model which gives a generalised conclusion based upon data taken in a range of environments. It is of the results-based type and has been used in this analysis to indicate the range of operation, and hence predict the overall benefits and costs arising from a particular regulation change. The basic Okumura equation is represented below: L 50 (db) = Lf + A MU (f,d) G(H TE ) G(H RE ) G AREA (f) The terms are explained in the table below. L 50 (db) Lf A MU (f,d) G(H TE ) G(H RE ) G AREA (f) The Okumura 50 th percentile path loss prediction in db The free space path loss for the frequency and range The Median Attenuation Adjustment for that frequency and range A term represented by an equation to adapt the path loss for different transmitter antenna heights A term represented by an equation to adapt the path loss for different receiver antenna heights An adjustment factor for the type of area, Suburban, Open or Quasi Open. Table 14 Terms in Okumura equation The value of each term can be obtained from look-up tables. The values in the tables have been derived from measurements. C13400 013 1510 Final Report v1-0.doc Page 70

Results The power required vs link range has been evaluated using Okumura, and extrapolated to 5.8GHz since Okumura data is only available to 3GHz. Pow er Requirement Versus Range 80 70 60 Transmit power, dbm 50 40 5.8GHz Suburban 30 2.4GHz Suburban 5.8GHz Quasi open 20 2.4GHz Quasi open 10 5.8GHz Open 2.4GHz Open 0 0 5 10 15 20 25 Range, Km Figure 11: Power vs range relationship The chart shows the transmit power required versus range as predicted by Okumura, at both 2.4GHz and 5.8GHz and for Suburban, Quasi Open and Open terrain. The following tables show the results for the chosen scenarios including comparison with high powered 3.5GHz. The quasi-open terrain Okumura data is most appropriate for modelling the areas that are likely to benefit from WBA. Range figures without fade margin show what should typically be achievable, The fade margin adds reliability to the performance of equipment installed within that range. C13400 013 1510 Final Report v1-0.doc Page 71

Power Level EIRP 2.4GHz Okumura Quasi Open Range prediction 10dB Fade margin Okumura Quasi Open Range prediction with no Fade margin, for comparison 100mW 1.75 3.5 1W 3.5 7 4W 5.75 11.25 10W 7.25 16 80W 16.5 26 Power Level EIRP 5.8GHz Okumura Quasi Open Range prediction 10dB Fade margin Okumura Quasi Open Range prediction with no Fade margin, for comparison 200mW 1.75 3.5 2W 3.5 6.75 4W 4.25 8.25 10W 5.5 11.25 50W 8.75 19.5 125W 12.25 24 200W 16.5 26 Power Level EIRP 3.5GHz Okumura Quasi Open Range prediction 10dB Fade margin Okumura Quasi Open Range prediction with no Fade margin, for comparison 125W 16.5 26 Table 15 Power vs link range for chosen scenarios The modelling work outlined above generates a figure that can be considered to be an average maximum range of communications for a given maximum power limit. C13400 013 1510 Final Report v1-0.doc Page 72

7.3 Network Modelling 7.3.1 Overview The network model determines the cost of providing WBA service to each geomodel cell, based on a point-to-multipoint architecture. The model inputs the number of addressable users for each cell, applies an economic demand model to determine the number of subscribers per cell, and then calculates the cost of providing service to those subscribers. This is compared with the revenues generated by those subscribers to test whether it is financially viable to provide service to each cell. For viable cells, subscribers in those cells are aggregated to calculate the gain in consumer surplus arising from the introduction of WBA. The network model and key assumptions are described in more detail below and additional outputs of the model and sensitivity analyses are set out in Annex D. 7.3.2 Bandwidth requirement The network is sized using bandwidth requirement per customer, increasing at 5% pa to 2010: Residential Business 2005 bandwidth requirement 0.035 Mbit/s per customer 0.230 Mbit/s per customer 2010 bandwidth requirement 0.045 Mbit/s per customer 0.294 Mbit/s per customer Table 16 Bandwidth Requirements The 2005 figures are adapted from current ISP figures in order to provide the required contention ratios for business and residential services. For residential subscribers the allocation has been increased from a benchmark figure of 0.020 Mbit/s to allow for the inclusion of an expected significant proportion of high usage subscribers taking residential service (e.g. sole traders, teleworkers, SOHO users). For business users it has been assumed that there are around 6-7 employees per business in the target areas, with each employee having the same requirement as a high use residential subscriber. The resulting business and residential bandwidth requirements are aggregated to provide a total cell bandwidth figure for each individual cell. C13400 013 1510 Final Report v1-0.doc Page 73

7.3.3 Point to multipoint architecture The network architecture applied to the cells is simplified in order to provide per cell costs and to enable the same network model to be applied to all of the power scenarios (which range from 450 to 20,500 cells to cover Great Britain). The architecture consists of three core elements: Radio network The radio equipment for each cell consists of: One or more base stations, situated at the centre of each cell One or more sector radios (capacity 10Mbit/s for 5.8GHz and 4Mbit/s for 2.4GHz), attached to the base station(s) One CPE per subscriber. The number of sector radios required is calculated according to cell bandwidth. Each base station is assumed to support a maximum of 6 sectors at 5.8GHz and 3 sectors at 2.4GHz; if more sectors are needed then additional base stations are added. Cell backhaul Traffic from each cell requires radio backhaul (via a point to point link) to an aggregation point. It has been assumed that: Base station infrastructure is located at a single site, which needs to be backhauled The basestation site is unlikely to be located at a BT exchange, hence a radio link is needed The link must be capable of delivering the specified services (i.e. at least 4 Mbit/s) The actual backhaul link used is sized according to the bandwidth requirement in each cell 2.4GHz backhaul is not used due to the sensitivity to interference of PtP links in this band The link length must be capable of reaching an adjacent cell, i.e. 2x cell radius, so that backhaul traffic can be aggregated. C13400 013 1510 Final Report v1-0.doc Page 74

Transit Aggregated traffic from several cells must be delivered to an uncontended internet peering point via a backbone network connection: The model assumes an average cost of transit from an aggregation point (multiple cells) to a backbone network and routing to a peering point The degree of aggregation depends on the bandwidths in the cells to be aggregated A degree of headroom is required, since in practice cells will be aggregated on a geographical basis rather than in a way which minimises transit capacity requirements. 7.3.4 Radio equipment The model is based on mid-range equipment (e.g. the Motorola Canopy system), with a capacity of 10 or 4 Mbit/s per sector, and a maximum of 6 or 3 sectors per base station, at 5.8GHz and 2.4GHz respectively. CPEs are outdoor type. Cost assumptions are based on industry interviews, available benchmark information and equipment list prices, and are structured to allow scalability in the network model. Assumptions are as follows: C13400 013 1510 Final Report v1-0.doc Page 75

Frequency 2010 price Notes Base station 2.4GHz 3,400 Includes antenna, ancillary equipment & installation. This represents the cost for the site equipment and first sector. Base station 5.xGHz 4,300 As above Base station 3.5GHz 4,730 3.5GHz equipment is based on 5.x pricing increased by 10% to allow for lower equipment volumes and higher powers at this frequency. Sector 2.4GHz 1,500 Cost per additional sector/distribution node, including installation. Sector 5.xGHz 2,200 As above Sector 3.5GHz 2,420 3.5GHz equipment is based on 5.x pricing increased by 10% to allow for lower equipment volumes and higher powers at this frequency. CPE 2.4GHz 240 Outdoor, including antenna and installation. CPE 5.xGHz 3.5GHz 335 368.50 Table 17 Cost Assumptions As above 3.5GHz equipment is based on 5.x pricing increased by 10% to allow for lower equipment volumes and higher powers at this frequency. In addition, pricing for the highest power scenarios (16.5km cells) is increased by 20%, to allow for equipment modification beyond a simple software change. C13400 013 1510 Final Report v1-0.doc Page 76

7.3.5 Cell backhaul A point-to-point radio link is required to backhaul each cell. This is selected from the following options, depending on cell bandwidth requirement: Speed Mbit/s 2010 price 34 Notes 5.8GHz Licence exempt 18 GHz Licenced 10 5,750 Point to point link cost including antenna, ancillary equipment & installation. 100 20,400 Point to point link cost including antenna, ancillary equipment & installation. Table 18 Backhaul Options The approach in this part of the model is designed to generate scalable, representative costs figures, not to model detailed deployment architectures. 7.3.6 Transit Transit capacity is required to connect cell backhaul traffic from an aggregation point to a backbone network and hence to the internet. This depends on the distances of the aggregation points to the core network and the traffic carried by the specific cells aggregated. This will vary greatly depending on cell size and geography. Therefore, an average approach based on a typical transit link deployed by rural WISPs has been taken. Rural WISPs may be able to link into existing regional networks to reduce their transit costs. In view of these factors, the following approach has been adopted: List pricing is according to available industry data (BT Central internet) Overall cost per link is in line with WISP benchmarks Headroom is built in to the aggregation of cells, i.e. total available capacity exceeds required capacity by some margin Overall transit cost as a % of total opex higher than capex, in line with industry experience The impact of reduction in transit costs is assessed in a sensitivity analysis. Adopting this approach results in the following transit options: 34 extrapolated from 2005 prices C13400 013 1510 Final Report v1-0.doc Page 77

Speed Mbit/s 2010 annual rental Notes Leased line 4 25,800 Minimum transit to enable the required business services. Price includes setup costs (amortised over the period). Leased line 34 82,000 Dedicated circuit & uncontended port connection to the internet via major service provider. Price includes setup costs (amortised over the period) Leased line 100 134,000 Dedicated circuit & uncontended port connection to the internet via major service provider. Price includes setup costs (amortised over the period) Table 19 Transit Options This results in transit costs which amount to 40% - 60% of WISP annual opex, depending on the scenario. Headroom in transit capacity ranges from 60% for 1.75km cells (most difficult to serve) to 5% for 16.5km cells. C13400 013 1510 Final Report v1-0.doc Page 78

7.3.7 Network operation Network operating costs are shown in the following table. Cost type Cost Notes Basestation site rental 11,000 p.a. in 2010 Operations and maintenance: 2.4GHz 5.8GHz 10% of capex 20% of capex Web hosting: Residential Business Billing cost 10/subscriber p.a. 18/subscriber p.a. 10/subscriber p.a. Table 20 Network Operating Costs 1. Additional opex is allocated to 2.4GHz scenarios to allow for higher potential interference suffered by operators from WLANs particularly 2. includes staff costs for network maintenance These figures are based on industry benchmarks, and comprise around 25% - 45% of WISP annual opex, depending on the scenario. 7.3.8 ISP operation The costs of a WISP operation include staff cost, office expenses and marketing. In line with typical rural WISP operation it has been assumed that these are relatively modest at around 70,000 per ISP per year. This figure was derived from ISP benchmarks. In order to allocate these costs on a per cell basis the broad assumption has been made that operating 50 cells requires one ISP. For the higher power scenarios (16.5km and 7.25km) additional ISP operating cost per cell has been allocated to compensate for the significantly lower number of cells in these scenarios. Results show that overall ISP operating costs contribute around 5% of annual opex. 7.3.9 Determination of economic viability Capital costs for each scenario are the total of the base station, sector, CPE and cell backhaul costs per cell. These are depreciated over 5 years to calculate an annual depreciation charge in the viability determination year (2010). This is combined with annual opex per cell (comprising site rental, network opex and ISP opex) to arrive at a cost for operating each cell in 2010. If annual revenues from C13400 013 1510 Final Report v1-0.doc Page 79

2010 subscribers in that cell exceed these costs, then the cell is considered financially viable and contributes to the calculation of consumer surplus, otherwise it is assumed that no roll-out of WBA will occur for that cell. Therefore, no benefits from WBA will be realised for that cell. 7.4 Demand modelling and consumer surplus This section provides an overview of the methodology underpinning the economic cost/benefit analysis of higher power WBA. A detailed description of the methodology is provided in Annex E. 7.4.1 Estimating net private benefits The main benefit of higher power is likely to be availability of WBA broadband at lower prices than satellite broadband, thus net private benefits are captured by the increase in consumer surplus as a result of these lower prices. Consumer surplus is the benefit enjoyed by all consumers from paying a lower price. For those consumers who would have otherwise subscribed to satellite broadband this is represented by the cost saving of WBA over satellite. For consumers who would not subscribe to satellite, but who are attracted to WBA at the lower price, their consumer surplus is captured by the difference between the price they have to pay to WBA and how much they would have been prepared to pay for the service. Since satellite broadband is relatively expensive, there are large potential cost savings and pent up demand The magnitude of this effect is significant. Benefits are not considered to arise from WBA competing with DSL and cable broadband. Since current prices for xdsl and cable services are competitive, and WBA broadband does not offer a cost advantage, customers who currently have access to these services will not obtain significant benefits from the introduction of higher power WBA (i.e. there is no change in consumer surplus). C13400 013 1510 Final Report v1-0.doc Page 80

7.4.2 Broadband price-demand relationships In order to calibrate the price-demand functions (which allow calculation of consumer surplus) the following assumptions and data have been used: Assumption Value Main source(s) Price of DSL service Residential 16 per DotEcon price month survey, Jan 2006 Business 70 per month Price of broadband satellite service Residential 90 per month Business 500 per month DotEcon price survey, Jan 2006 Number of UK DSL connections Number of UK satellite connections Business proportion of broadband connections 8,087,000 Ofcom Communications Market August 2005 Quarterly Update 9,000 Ofcom Communications Market August 2005 Quarterly Update 8% [see discussion below] UK household premises 25,210,513 SERRL data Target household premises (unserved by DSL) 343,942 SERRL data UK business premises 1,418,570 SERRL data Target business premises 272,547 SERRL data Table 21 Assumptions made to calculate price-demand relationships The SERRL data are based on postal delivery data, i.e. sites where mail is delivered. Business sites are defined as those which receive 10 or more items of mail per day. The number of UK household premises described by this dataset is similar to that considered in previous work by Analysys 35 24.7 million based on UK statistics office data. The number of business premises in this dataset is lower than the number of UK vat registered firms (National Statistics reports just over 35 Spectrum demand for non-government services, Analysys/Mason, Sept 2005 C13400 013 1510 Final Report v1-0.doc Page 81

1.6 million for 2005); the total number of firms in the UK is greater still (2.4 2.6 million, depending on the source). Although business premises and numbers of firms are different measures, it is likely that a large number of firms are missing from the dataset. Despite this, in our model the business dataset has not been adjusted upwards from 1.4 million, thus taking a conservative approach. This means that the smallest businesses (including sole traders and micro businesses likely to be operating from household premises) which are those likely to be missing from the dataset are considered as residential users, which is representative as such businesses are unlikely to require higher speed broadband services, but rather would benefit most from a service similar to that demanded by residential users. Demand per cell is determined in two stages, as follows: 1 Calculate the maximum demand per cell, based on existing UK broadband penetration data 2 Apply an adoption curve which reaches saturation at maximum demand to determine the path of take-up over timescale considered. It has been assumed that the demand for broadband services is a linear function. Thus demand can be calibrated using two quantity/price points: Due to its high cost, demand for satellite broadband in areas not covered by DSL provides a natural point high on the demand curve where demand is curtailed by price; and DSL and cable broadband services are relatively cheaper, due to lower infrastructure costs, and therefore the demand for these services provides a natural point lower on the demand curve. The slope of this demand relationship (defined in Annex E) is determined by the relative actual penetrations and prices of DSL and satellite broadband services, for both business and residential customers. This is shown in Figure 12, which highlights low demand for the relatively expensive satellite services, and higher demand for the more affordable DSL services. C13400 013 1510 Final Report v1-0.doc Page 82

600 500 400 Price 300 Residential Business 200 100 0 0% 10% 20% 30% 40% 50% 60% Penetration Figure 12 Broadband price-demand relationship The relative slopes of the demand curves are very sensitive to the proportion of business DSL connections (i.e. how the 8 million DSL connections in the UK are split between business and residential premises). Analysts suggest that business connections are currently around 12% - 14% 36. However, since the SERRL delivery point dataset excludes smaller businesses operating from household premises it has been assumed that the relevant proportion of business connections is lower, at 8%. The level of demand for WBA at saturation, for any given price, was determined from the relevant demand curve. The quantity demanded for a given price is simply the number of subscribers that would demand the service linearly interpolated at the given price. 7.4.3 WBA take-up An adoption curve is applied to the maximum demand for WBA (determined from the price/demand model as described above) to determine the path of take-up of WBA subscribers. Since the target areas are currently under-served, there is likely to be significant pent-up demand for broadband, therefore a rather swift take-up is likely. Figure 13 shows the Gompertz adoption curve estimated using benchmarks, where take-up reaches 85% of maximum demand with 6 years. 36 DotEcon calculation using Point Topic data, and Generics calculation based on Analysys/Mason in Spectrum demand for non-government services, Sept 2005 C13400 013 1510 Final Report v1-0.doc Page 83

Figure 13 Broadband take-up in target areas Applying this take-up curve to the maximum demand per cell gives the number of WBA subscribers, as shown in Figure 14. C13400 013 1510 Final Report v1-0.doc Page 84

120,000 100,000 WBA Subscribers 80,000 60,000 40,000 Residential subscribers Business subscribers 20,000-2005 2006 2007 2008 2009 2010 Figure 14 WBA subscribers 37 2010 penetration for residential WBA subscribers is 20% of households, and for business WBA subscribers 40% of businesses. This reflects the difference in saturation penetrations. The total number of business subscribers is greater than residential this is because of the lower distance limit for the selected business service. 7.4.4 Calculation of consumer surplus The gain in consumer surplus for business and residential consumers from WBA is calculated by netting consumer surplus from satellite broadband from the consumer surplus for WBA, for aggregate demand in all financially viable cells for each year from 2005 2010. Business and residential consumer surplus in each year was aggregated and then discounted the annual benefits to provide the net present value of private net benefits. Consumer surplus gains calculated for the higher power scenarios considered are set out in section 7.5.3. 7.5 Benefits, results and interpretation This section sets out the main results of the application of the network and economic modelling to the geo-model cells for each scenario. Further outputs of the network model and sensitivity analyses are given in Annex D. 37 Scenario: 3.5km cells C13400 013 1510 Final Report v1-0.doc Page 85

It is important to note that the model is necessarily broad in scope, ranging from broadband deployment to calculation of economic benefits for rural WBA via radio network deployments covering the whole of Great Britain. In order to do this the radio network architecture needs to be simple and many assumptions need to be made, some of which are very sensitive in terms of the final results. In view of this the following approach has been taken: Adopted a common, conservative approach to the calculation of benefits of higher power and costs of interference Used the same modelling approach for current situation and higher power scenarios Adjust these scenarios where necessary to allow fair comparison (e.g. higher equipment costs for the 16.5km scenarios) Run sensitivity analyses for the critical parameters. 7.5.1 Network deployment Application of the network model to the geo-model cells results in an estimate of the costs of deploying a licence exempt radio network to residences and businesses beyond the reach of DSL for the defined residential (1 Mbit/s) and business (4 Mbit/s) services: Cell radius Frequency Scenario Served cells 38 Capex GBP, million Annual Opex GBP, million Current 1.75 2.4 regulations 18,495 226 517 3.50 2.4 1W EIRP 6,617 127 274 Current regulations 6,617 141 251 3.50 5.8 4.25 5.8 4W EIRP 4,789 125 206 7.25 2.4 10W EIRP 1,894 90 159 7.25 5.8 25W EIRP 1,894 98 129 16.50 2.4 80W EIRP 447 84 111 16.50 3.5 125W EIRP 447 99 79 16.50 5.8 200W EIRP 447 92 78 Table 22: Served, cells, capex and opex for each scenario Deployment of such a network would serve over 300,000 households and 250,000 business premises. Cost of deployment and operation reduces at higher power, as shown in Figure 15 and Figure 16. 38 Note that this is lower than the number of cells in the geo-model as, after applying takeup, some cells have no subscribers and hence no radio network is deployed C13400 013 1510 Final Report v1-0.doc Page 86

250 200 Capex (GBP, million) 150 100 2.x higher power 5.x higher power 2.x current power 5.x current power 50 - - 2 4 6 8 10 12 14 16 18 Cell radius (km) Figure 15 WBA roll-out capex requirements In the model capex is lower for 2.4GHz than 5.8GHz at the same cell size, due to lower equipment costs. The results show a significant reduction in capex requirement at 2.4GHz by increasing power to 1W, mirroring the large reduction in the number of cells required. Capex benefits for increasing 5.8GHz power to 4W are modest (8%). C13400 013 1510 Final Report v1-0.doc Page 87

600 500 Opex (GBP, million) 400 300 200 2.4GHz higher power 5.8GHz higher power 2.4 current power 5.8 current power 100 - - 2 4 6 8 10 12 14 16 18 Cell radius (km) Figure 16 WBA roll-out opex requirements The model indicates that operating costs decrease with cell size and are higher at 2.4GHz than 5.8GHz due to greater potential for interference. The reduction in operating costs for higher power is significant, particularly for the larger 7.25km and 16.5km cell sizes. This is mainly due to reduced base station site rental and transit costs. C13400 013 1510 Final Report v1-0.doc Page 88

7.5.2 Financially viable cells Given capex and opex per cell, and revenue per subscriber, the profit (or loss) for each individual cell can be determined. The number financially viable cells and the net profit for each scenario can be determined: Cell radius Frequency Scenario Financially viable cells 39 % of cells financially viable Net profit GBP, million, pa Current 1.75 2.4 regulations 550 3% -339 3.50 2.4 1W EIRP 923 14% -76 Current regulations 1,190 18% -56 3.50 5.8 4.25 5.8 4W EIRP 1,339 28% -8 7.25 2.4 10W EIRP 861 45% 46 7.25 5.8 25W EIRP 1,027 54% 75 16.50 2.4 80W EIRP 294 66% 95 16.50 3.5 125W EIRP 319 71% 124 16.50 5.8 200W EIRP 321 72% 126 Table 23 Economic cells and subscribers The percentage of cells which are financially viable increases with cell radius, with a significant increase at the higher power levels (7.25km and 16.5km cells). These are the only scenarios which are financially viable across the target regions as a whole. 39 As a percentage of cells where there are subscribers after applying take-up C13400 013 1510 Final Report v1-0.doc Page 89

The number of financially viable subscribers in financially viable cells is as follows: Cell radius Frequency Scenario Financially viable residential subs 40 % of target subs (res) 41 Financially viable business subs % of target subs (bus) Current 1.75 2.4 regulations 18,562 5% 24,354 9% 3.50 2.4 1W EIRP 34,426 10% 52,537 19% Current regulations 39,563 12% 63,492 23% 3.50 5.8 4.25 5.8 4W EIRP 45,477 13% 76,372 28% 7.25 2.4 10W EIRP 52,227 15% 91,459 34% 7.25 5.8 25W EIRP 57,352 17% 99,574 37% 16.50 2.4 80W EIRP 62,425 18% 105,412 39% 16.50 3.5 125W EIRP 64,115 19% 106,465 39% 16.50 5.8 200W EIRP 64,309 19% 106,562 39% Table 24 Numbers of subscribers in financially viable cells These results highlight that: Increasing power from the current regulations to 1W EIRP at 2.4GHz or to 4W EIRP at 5.8GHz gives a modest improvement in subscribers Significantly higher powers (10W or greater) are needed for a substantial improvement (to around 60% of the saturation penetration level). 40 Number of subscribers in economic cells 41 Subscribers in economic cells as a percentage of total target subscribers in the geomodel (343,942 residential, 272,547 business) C13400 013 1510 Final Report v1-0.doc Page 90

Figure 17 Financially Viable Coverage at 5.8GHz with 2W and 4W EIRP C13400 013 1510 Final Report v1-0.doc Page 91

7.5.3 Economic benefits calculation of consumer surplus Based on the results of the network model, and using the methodology outlined in section 7.4 to calculate consumer surplus 42, the benefits of higher power arising from the different scenarios are: Cell radius Frequency Scenario Residential consumer surplus 43 GBP, million Business consumer surplus GBP, million Total consumer surplus GBP, million Current 1.75 2.4 regulations 18 148 166 3.50 2.4 1W EIRP 35 319 354 Current regulations 40 386 426 3.50 5.8 4.25 5.8 4W EIRP 47 464 511 7.25 2.4 10W EIRP 54 556 609 7.25 5.8 25W EIRP 59 605 664 16.50 2.4 80W EIRP 65 640 705 16.50 3.5 125W EIRP 66 647 713 16.50 5.8 200W EIRP 67 647 714 Table 25 Consumer Surplus The economic benefit in each scenario is the increase in consumer surplus generated by deploying WBA in financially viable cells which can otherwise only be served with broadband satellite. The calculated benefit arises from the fact that having broadband available at a significantly lower price than satellite increases demand substantially. Since the number of subscribers increases with higher power, calculated consumer surplus also increases with power. The net benefit of each higher power scenario (calculated by subtracting the consumer surplus of the current regulation scenarios at 2.4 and 5.8GHz) is therefore as follows: 42 Consumer surplus is calculated as the NPV of consumer surplus from 2005 to 2010 using a discount rate of 3.5% 43 As a percentage of cells where there are subscribers after applying take-up C13400 013 1510 Final Report v1-0.doc Page 92

Frequency Scenario Net consumer surplus 44 Cell radius GBP, million 3.50 2.4 1W EIRP 188 4.25 5.8 4W EIRP 85 7.25 2.4 10W EIRP 443 7.25 5.8 25W EIRP 238 16.50 2.4 80W EIRP 539 16.50 3.5 125W EIRP 287 16.50 5.8 200W EIRP 288 This is shown in Figure 18. Table 26 Net benefit (consumer surplus) 600 Net consumer surplus (GBP, million) 500 400 300 200 100 2.4GHz higher power 5.8GHz higher power 0-20 40 60 80 100 120 140 160 180 200 Power (EIRP) Figure 18 Net consumer surplus at higher power Thus the increase in consumer surplus is greatest for the 2.4GHz 10W and 80W higher power scenarios. This is because consumer surplus arising at 2.4GHz under the current regulations is much lower than for the other cases, which in turn results from the poor financially viability of serving 1.75km cells. In contrast increasing power to 1W at 2.4GHz or to 4W at 5.8GHz has a more modest impact. There is also a significant increase for 25W and 200W at 5.8GHz. 44 Net of consumer surplus under current regulations C13400 013 1510 Final Report v1-0.doc Page 93

7.5.4 Areas that benefit from a regulatory change The geographic areas which become financially viable by the application of higher power (increase from 2W to 4W) at 5.8GHz is shown in Figure 19 and at 2.4GHz in Figure 20 (increase from 100mW to 1W) and Figure 21 (further increase from 1W to 10W). Details of the method used to create these plots, and plots showing the financially viable cells for other scenarios are given in Annex G. The histogram accompanying each figure illustrates the number of residences of each area type that are within coverage area of the financially viable cells. This is all endpoints, not only those that are not served by ADSL. The total rural column in the histogram is the sum of Hamlet, Village, and Rural Towns. The histogram therefore, provides an indication of the types of areas that benefit because the characterisation of areas is related to the types of residences within them. The abbreviations used for the types of area are as follows. M U A L U A O U A L M T R Town Village Hamlet Major Urban Large Urban Other Urban Larger Market Town Rural Town Village Hamlet/Dispersed A formal description of these in included in Annex G. Note that the information provided in this section excludes Scotland for which the categorisation of area types is done differently. C13400 013 1510 Final Report v1-0.doc Page 94

Endpoints 800000 700000 600000 500000 400000 300000 200000 100000 0 MUA LUA OUA LMT RTown Village Hamlet Total Rural Figure 19 Areas and their rural classifications that benefit from a power increase at 5.8GHz from 2W to 4W It should be noted that the areas that are provided with WBA coverage due to the change from 2 to 4W can be characterised as being predominantly non rural. 2500000 2000000 Endpoints 1500000 1000000 500000 0 MUA LUA OUA LMT RTown Village Hamlet Total Rural Figure 20 Areas and their rural classifications that benefit from a power increase at 2.4GHz from 100mW to 1W C13400 013 1510 Final Report v1-0.doc Page 95

Endpoints 1800000 1600000 1400000 1200000 1000000 800000 600000 400000 200000 0 MUA LUA OUA LMT RTown Village Hamlet Total Rural Figure 21 Areas and their rural classifications that benefit from a power increase at 2.4GHz from 1W to 10W At 2.4GHz it can be seen from the bar graphs of Figure 20 and Figure 21 that the step to 1W provides WBA coverage for predominantly non-rural areas. The further step to 10W provides a much higher proportion of rural area coverage. 7.5.5 Comparison of Licenced vs. Licenced-Exempt WBA In the geo-model 16.5km cells roughly correspond to the maximum power allowed for licenced WBA at 3.5GHz, i.e. 125W. This scenario has been modelled in a simple way by taking the 5.8GHz 16.5km (200W) scenario and increasing radio network capex by an additional 10%. The net consumer surplus for this scenario 45 is GBP 287 million, i.e. it is more beneficial than the 5.8GHz (or 2.4GHz) current regulation scenario. Thus the commercial incentive to set up a rural WBA operation at 3.5 GHz is greater than at 5.8GHz unless the power allowed for 5.8GHz is at least 200W. 45 Net of consumer surplus under the 5.xGHz current regulation C13400 013 1510 Final Report v1-0.doc Page 96

7.6 Other benefits arising from a power increase Nomadic services are more valuable than fixed services because users are provided with a degree of mobility in addition to providing a fixed broadband service Therefore, prices may be higher than similar ADSL services. The scenarios modelled in this report have assumed fixed services only. Alternative scenarios are possible where operators provide their customers with a degree of nomadicity. Only a few specific scenarios have been considered. It has been assumed that WBA cannot compete in areas served by ADSL. If a wireless operator provides an 8Mbit/s service for instance (required for IPTV) then they can compete over a much larger area because the ADSL coverage is considerably reduced. The introduction of ADSL2+ may boost the availability of both speed and reach of higher rate service but above approximately 4kms there is no advantage over ADSL. Benefits for point-to-point for backhaul. Operators like to have a connection which is independent of the incumbent operator. Increased power would extend backhaul link lengths and provide operators with further options which could reduce costs. SDSL coverage is extremely poor. If demand for symmetric services grows (e.g. gaming, website hosting, main office serving satellite offices then the symmetric nature of radio services will provide them with an increasing advantage over ADSL). Benefits could be eroded by activities of wireline telcos such as fitting kerb-side cabinets. The telcos capacity to erode benefits are dependent upon the distribution of un-served customers. The more evenly distributed they are the more difficult it will be for wireline telcos to serve them cost effectively. C13400 013 1510 Final Report v1-0.doc Page 97

7.7 Feedback from the stakeholders on WBA market development issues One of the main concerns expressed by interviewees is lack of credibility and high cost of equipment. Current regulations are seen by interviewees as limiting the extent to which the business case can be made more credible the main issues cited are access to spectrum, harmonisation and power limits rather than cost of spectrum. In the market, competition from ADSL is impacting the residential business case significantly, resulting in low numbers of users per Base Station. This is resulting in an increased focus on selling symmetric services to higher value SMEs and high-end home/soho users, and serving those residential users who can only get a restricted ADSL service or who can t get ADSL at all. There is a consensus that the current WISP business case is marginal. According to interviewees, the main factors which could mitigate this situation are: A significant reduction in network infrastructure cost (lower number of base sites needed and lower equipment costs) achieved via harmonisation with the US and higher power limits Strong growth in demand for symmetric services some interviewees suggested that WBA has an advantage in rural areas due to the lower distance from the exchange limit of SDSL compared to ADSL. This could become important in the medium term if demand for symmetric services grows significantly (for example for video messaging, operating servers or uploading large files) The evolution of a compelling WISP triple play of broadband + VoIP + nomadic capability. Some interviewees thought that broadband you can take with you alongside VoIP could become a key differentiator for WISPs. Competition between WBA operators was felt to be unlikely, as the business case is not thought to be strong enough to attract multiple sub-regional WISP competitors, even in prime target areas such as Science Parks. C13400 013 1510 Final Report v1-0.doc Page 98

7.8 Benefits modelling conclusion The results indicate that allowing higher power has a significant impact on the viability of rural WBA in Great Britain. For example the capex cost of covering areas unserved by DSL with a WBA network can be reduced by GBP50 million to around GBP90 million for the highest power scenarios. This would make more cost-effective broadband available to around 108,000 UK households and 68,000 UK businesses. More significantly for WISPs, the annual operating costs for running such a network fall substantially with power increases such as from 100mW to 1W at 2.4GHz or from 2W to 25W at 5.8GHz. These cost reductions would allow WISPs to serve a greater proportion of these households and businesses economically. For example, around 45% to 85% more residential customers and 55% to 115% more business customers could be served economically at 2.4GHz 1W and 5.8GHz 25W than under the current regulations. This amounts to around 40,000 additional households and around 80,000 additional businesses under the highest power increase scenarios considered. The benefits of higher power are measured by calculating consumer surplus, which arises from making cost-effective WBA available to consumers who are currently only able to access satellite broadband. In the absence of other considerations, higher power to 1W at 2.4GHz has been estimated to increase consumer surplus by 180M and to 10W by 443M. At 5.8GHz higher power to 25W has been estimated to increase consumer surplus by 238M. However, a modest power increase to 4W only increases consumer surplus by 85M These benefits must be considered in the context of the additional interference caused by higher power. This is considered in section 8. C13400 013 1510 Final Report v1-0.doc Page 99