Comparison of Mobile WiMAX and HSDPA KOLBRUN JOHANNA RUNARSDOTTIR Master of Science Thesis Stockholm, Sweden 2008 Comparison of Mobile WiMAX and HSDPA KOLBRUN JOHANNA RUNARSDOTTIR Master of Science Thesis performed at the Radio Communication Systems Group, KTH. June 2008 Examiner: Professor Ben Slimane KTH School of Information and Communications Technology (ICT) Department of Communication Systems (CoS) CoS/RCS 2008-10 c Kolbrun Johanna Runarsdottir, June 2008 Tryck: Universitetsservice AB Abstract HSDPA and Mobile WiMAX are high speed mobile technologies with different backgrounds. HSDPA is a data enhancement for a voice-centric 3GPP system while WiMAX is data-centric broadband technology that has an added feature of mobility. To investigate how the technical advantages and limitations of those technologies affect their feasibility as investment choices, first a general overview and comparison is given. Thereafter the deployment cost and operating cost over 10 years is estimated for the case of a new market entrant. This is done for the specific case of the city of Reykjavík. Radio planning assumptions and link budget calculations are made. This affects the network dimensioning, which in turn determines the infrastructure cost. As Reykjavík is a city with few inhabitants but a large area, coverage had more affect on the infrastructure cost than the capacity. Another factor that made the networks less sensitive to capacity limitations was that the system model did not consider interference. The 60% longer radius of HSDPA gave it an advantage in economic feasibility while 70 % higher throughput for Mobile WiMAX did not give any economic advantage under the considered assumptions. If the cell radius of the two technologies is set to be of the same length the outcome of the business cases is very similar. Variations in cell radius has a big impact on the NPV and IRR of both business cases while changing the average channel throughput has a little impact. 3 1. Introduction 2. Overview and comparison of mobile WiMAX and HSDPA 2.1.1 Background 2.1.3.Other features of Mobile WiMAX 2.2.1 Background 3.1.4 Link Budget 4. Techno economic analysis 4 5 List of tables Table 1. MCS levels and DL PHY data rates in Mobile WiMAX [5] ............................. 17 Table 2. Some features of W-CDMA ............................................................................... 25 Table 3. HSDPA terminal capability categories [15] ...................................................... 30 Table 4. Peak data rates with number of codes fixed at 15 [15],[10] .............................. 30 Table 5. Differences between WCDMA (R´99) and HSDPA ......................................... 32 Table 6. Comparison of some of the features of Mobile WiMAX and HSDPA ............. 36 Table 7. Model Parameters. ............................................................................................. 43 Table 8. Uplink link budget ............................................................................................. 45 Table 9. Downlink link budget ........................................................................................ 46 Table 10. Common system parameters for HSDPA and Mobile WiMAX...................... 49 Table 11. System specific parameters.............................................................................. 49 Table 12. SINR for different MCS levels, IEEE 802.16e [24], [31] ............................... 51 Table 13. HSDPA Rx sensitivity for different MCS levels. [27] .................................... 52 Table 14. Rx sensitivity for mobile WiMAX (5 and 10 MHz). [24], [38], [31]............. 52 Table 15. Cell radius, for HSDPA and Mobile WiMAX (5 MHz and 10 MHz)............. 53 Table 16. Distribution of modulation and coding schemes, HSDPA. ............................. 54 Table 17. Distribution of modulation and coding schemes, mobile WiMAX 10 MHz... 55 Table 18. Distribution of modulation and coding schemes, mobile WiMAX, 5 MHz.... 56 Table 19. WiMAX, peak data rates for different modulation and coding schemes.......... 56 Table 20. HSDPA peak data rates for different modulation and coding schemes........... 56 Table 21. A prediction of the mobile broadband market ................................................. 59 Table 22. Installation and site cost................................................................................... 60 Table 23. ARPU (Average Revenue/User) and data rate for 3 subscription types. [4] .. 60 Table 24. Cell radius and average throughput ................................................................. 61 Table 25. Nr. of BSs needed to fulfill coverage and capacity requirements. ................... 62 Table 26.Required investment capital, NPV and IRR for Mobile WiMAX 10 MHz ...... 65 Table 27.Required investment capital, NPV and IRR for Mobile WiMAX 5 MHz ........ 65 Table 28.Required investment capital, NPV and IRR for HSDPA .................................. 65 Table 29.Comparison of WiMAX with 5 MHz bandwidth and 10 MHz bandwidth ....... 67 Table 30. Comparison of HSDPA and mobile WiMAX (5 MHz BW)............................ 68 6 7 List of figures Figure 1. The layers of the OSI model, and the Data Link sublayers, LLC and MAC. .. 15 Figure 2. Network Architecture of Mobile WiMAX ........................................................ 20 Figure 3. UMTS network architecture ............................................................................. 33 Figure 4. Distribution of MCS, HSDPA ........................................................................... 55 Figure 5. Distribution of MSC for WiMAX 10 MHz....................................................... 55 Figure 6. Distribution of MCS for mobile WiMAX 5 MHz............................................ 56 Figure 7. Example of a cash balance curve...................................................................... 64 Figure 8. Cash balance curve for mobile WiMAX 10 MHz............................................ 66 Figure 9. Cash balance curve for mobile WiMAX 5 MHz.............................................. 66 Figure 10. Cash balance curve for HSDPA ..................................................................... 66 Figure 11.Sensitivity on cell radius. ................................................................................. 69 Figure 12. Sensitivity on channel throughput .................................................................. 70 Figure 13. Sensitivity on penetration levels..................................................................... 71 8 9 List of abbreviations AMC – Adaptive Modulation and Coding ARPU – Average Revenue Per User BS (BST) – Base Station BW – Bandwidth CAPEX – Capital Expenditures CPE – Customer Premises Equipment dB – Decibel dBi – dB isotropic dBm – dB milliwatt DL – Down Link DVB– Digital Video Broadcasting EIRP – Effective Isotropic Radiated Power FDD – Frequency Division Duplex H-ARQ – Hybrid Automatic Repeat Request HSDPA – High-Speed Downlink Packet Access IEEE – Institute of Electrical and Electronics Engineers IP– Internet Protocol IRR – Internal Rate of Return LOS – Line of Sight MAPL – Maximum Allowable Path Loss MCS – Modulation and Coding Scheme MMS – Multimedia Messaging Service MS – Mobile Station NLOS – None Line of Sight NPV – Net Present Value OFDM – Orthogonal Frequency Division Multiplexing OFDMA – Orthogonal Frequency Division Multiple Access OPEX – Operative Expenditures PDA – Personal Digital Assistant QAM – Quadrature Amplitude Modulation QoS – Quality of Service QPSK – Quadrature Phase Shift Keying SNR – Signal to Noise Ratio TDD– Time Division Duplex UL – Up Link UMTS – Universal Mobile Telecommunications System VoIP – Voice over Internet Protocol WiMAX – Worldwide Interoperability for Microwave Access 10 11 1. Introduction 1.1. Síminn and the Icelandic telecom market The work in this thesis has been requested by the Icelandic Telecom company, Síminn 1 . Síminn has carried out studies on both fixed and mobile WiMAX in collaboration with other European operators within Eurescom (The European Institute for collaborative research and strategic studies) and has been deploying fixed WiMAX in summer cottage areas in an experimental purpose. Recently Síminn started deployment of an UMTS network with HSDPA included in Iceland. The Icelandic market is in many ways interesting. It is small, but developed and there is a high purchasing power. The consumers are open to new products and technologies. For an example, Iceland has the highest ADSL penetration in the world and a year after Síminn started its IPTV service 19 % of the Icelandic households were subscribing, which is a world record [1]. Still the Icelandic market is difficult because of the country’s small population and the relatively big size of the island. One of the main purposes of a special law, Act8/2005 2 on Third Generation Mobile Telephony, is to guarantee that the rural areas will not be left out when it comes to 3G. This act applies to the allocation of frequency ranges 1900-1980 MHz, 2010-2025 MHz and 2110-2170 MHz [2] Two thirds of the Icelandic population lives in the Reykjavik area, which makes it by far the best area from market perspective. No frequency band has yet been assigned for TDD in Iceland but it is likely that the Icelandic Post and Telecom Agency will follow other European countries and the CEPT 3 recommendations. Hence it is likely that the frequency band 2500-2690 MHz will be allocated partly for TDD and there will be a change for operators to deploy WiMAX there. 1.2. Problem Definition The objective of this thesis is to provide a techno-economic comparison of two wireless broadband technologies, Mobile WiMAX and HSDPA. The comparison is divided into two parts. The first part is a general overview and comparison of the two technology standards. In the second part the deployment cost and operating cost over 10 years in terms of CAPEX, OPEX and estimated revenues is estimated, for the case of a new market entrant. The investigations will be performed for the city of Reykjavík. 1 1 Síminn is the largest telecommunication operator in Iceland. Its range of services includes fixed and mobile subscriptions, Broadband/DSL, Live TV and Centrex network solutions. Síminn has about 80% market share in fixed line telephony, 65% in the mobile telephony and 56% market share as an ISP. Síminn operates a GSM network reaching 98% of the population and since September 2007 an UMTS network that covers Reykjavik, Keflavik and Akureyri. [1] 2 The law states how frequencies should be assigned and what conditions the operators have to fulfill to get the frequency license. The country is divided into four areas and the service has to reach 60% of each of them. The law also states how fast the operator has to obtain this coverage. A better coverage gives discount. The duration of the right to use the frequencies is 15 years per each block. 3 The European Conference of Postal and Telecommunications Administrations. 12 1.3. Approach Methodology This thesis is partly based on studies on Mobile WiMAX carried out by Síminn in collaboration with a few other European incumbent operators within Eurescom. The Eurescom project aimed to clarify the role and usefulness of mobile WiMAX by performing a set of business case analysis. To perform the analysis first radio planning assumptions and link budget calculations were done, as this affects the network dimensioning, which in turn determines the infrastructure cost. [3], [4] In this thesis we will use the similar methodology but only one business case will be considered, the one for a new entrant and the analysis will be performed for two technologies, Mobile WiMAX and HSDPA. In the first part of the thesis the subject will be introduced by giving an overview of the two wireless broadband standards, this is covered in chapter 2. In this overview the technical characteristics of both standards will be studied and a comparison between them given. Other aspects like market position and future development will be briefly discussed. The second part, the techno-economic analysis, is covered in chapters 3, 4 and 5. In chapter 3, a framework for the comparison is established by introducing system models and establishing radio-planning assumptions. Then these assumptions are used to do the network dimensioning that will give us two important parameters for the technoeconomic analysis, the cell radius and average throughput per channel. In chapter 4 a framework for the business analysis is established and in chapter 5 the results are presented and further analyzed in a sensitivity analysis. In chapter 6 the conclusions are presented and discussed. Tool The OPEX and CAPEX calculations will be carried out by adapting a spreadsheet-based model created by Telenor and used in the Eurescom project that this study is partly based on. The tool was created to calculate the economical feasibility of a Mobile WiMAX Business Case. This tool has been chosen for several reasons. First of all it fits the purpose of this thesis well as it provides the output parameters we are looking for and takes into account all aspects that we are interested in. It is also simple to alter to fit our needs. WiMAX 2.1.1 Background WiMAX stands for Worldwide Interoperability for Microwave Access and is a wireless broadband communication system defined in the IEEE 802.16 industry standards. The WiMAX Forum is an organization formed to promote conformance and interoperability of the IEEE 802.16 standards. The IEEE 802.16-2004 (802.16.d) standard, also known as the fixed WiMAX standard, was approved by IEEE in June 2004. It provides fixed, point-to-multi point broadband wireless access service and is primarily used as a wireless metropolitan area network (WMAN). In 2005 IEEE approved the mobile WiMAX amendment, IEEE 802.16e-2005 (802.16.e). In the 802.16e-2005 amendment the most important addition to the previous standards is mobility with specifications of mobility management, extensible authentication protocol (EAP) and handover. Mobile WiMAX is based upon Orthogonal Frequency Division Multiple Access (OFDMA). Some of the features of Mobile WiMAX are: • High Data Rates: Can theoretically support peak download data rates up to 63 Mb/s per sector and peak upload data rates up to 28 Mb/s per sector in a 10 MHz channel, when MIMO 2*2 is used and DL:UL ratio is 1:0 and 0:1 respectively. • Quality of Service (QoS): QoS measures for WiMAX include service availability, data throughput, delay, jitter, and error rate. • Scalability: Mobile WiMAX has been designed to be able to work in different channel bandwidths from 1.25 to 20 MHz. • Mobility: Optimized handover schemes with latencies less than 50 milliseconds and flexible key management schemes are supported. Mobile WiMAX has been tested for speeds up to 120 km/h. [5] These features make the technology suitable for a range of applications. Among the applications that have been suggested for mobile WiMAX is to provide high speed data services and other Internet services like VoIP for fixed, nomadic and mobile terminals, to provide a wireless alternative to cable or DSL for last mile broadband access and to connect hotspots with each other and the rest of the Internet. The only application we will consider is the high speed Internet service to fixed, nomadic and mobile users. 14 2.1.2 Technology The IEEE 802 standards are a family of standards dealing with local area networks (LANs) and metropolitan area networks (MANs). The standards mainly focus on the lowest 2 layers of the Reference Model for Open Systems Interconnection (OSI)[6], the Data Link layer and the Physical (PHY) layer. Furthermore, IEEE 802 splits the Data Link Layer into two sub-layers, the Logical Link Control (LLC) sub layer and the Media Access Control (MAC) sub layer. In broadband wireless networks, most of the things happen in the MAC layer and the PHY (Air Interface) layer. Figure 1. The layers of the OSI model, and the Data Link sublayers, LLC and MAC. The Physical layer At the first level of the seven level OSI model there is the Physical layer, the most basic network layer. Its main functions are providing an interface to physical transmission media, modulation, coding, bit synchronization, flow control and circuit mode multiplexing. [7] Mobile WiMAX (802.16e-2005) uses a physical layer mode called scalable Orthogonal Frequency Division Multiple Access (sOFDMA PHY) mode. OFDM Orthogonal Frequency Division Multiplexing is a multiplexing technique that divides the input data stream into a large number of parallel substreams of reduced data rate and each substream is modulated and transmitted on a separate orthogonal subcarrier. Each subcarrier is modulated with a conventional modulation scheme. The reduced data rate causes the symbol duration to increase and hence improves the robustness of OFDM to delay spread. Because of the subcarriers orthogonality there will be no cross talk between them, even though their spectra is very close or overlapping. By allowing the subcarriers to overlap, the spectrum required is reduced and the bandwidth efficiency is increased. OFDM signals are generated with efficient Inverse Fast Fourier Transform (IFFT), which enables a large number of subcarriers (up to 2048) with low complexity. In the time 15 domain the resources of an OFDM system are available by means of OFDM symbols and in the frequency domain by means of subcarriers. For allocation to individual users the time and frequency resources can be organized into subchannels. [5] OFDMA Orthogonal Frequency Division Multiple Access is a multiple-access/multiplexing scheme that divides the channel in such a way that many users can share it. An individual subcarrier or a group of subcarriers are assigned to different users and thus the users can all use the same channel simultaneously.[8] In other words OFDMA is a multi-user version of OFDM where subsets of subcarriers can be assigned to different users. Different number of subcarriers can be assigned to different users according to channel conditions and the type of service the user is requesting. This is comparable to CDMA and how spreading codes can be used to assign different users with different data rates. Scalable OFDMA (S-OFDMA) is the modulation/multiple-access method used in Mobile WiMAX to address the problem of different channel sizes in different countries. With the scalability IEEE 802.16-2005 can support channel sizes ranging from 1.25 MHz to 20 MHz. This is done by adjusting the FFT size based on channel size or bandwidth while keeping the subcarrier frequency spacing fixed at 10.94 kHz. Since the resource unit subcarrier bandwidth and symbol duration is fixed, the impact to higher layers is minimal when scaling the bandwidth. [5] To do the sub-channelization there are two types of subcarrier permutations: 1. Diversity permutation that draws subcarriers pseudo-randomly to form a subchannel. This provides frequency diversity and inter-cell interference averaging. 2. Contiguous permutations, which groups a block of contiguous subcarriers to form a subchannel. The diversity permutations include DL FUSC (Fully Used SubCarrier), DL PUSC (Partially Used SubCarrier) and UL PUSC. The contiguous permutations include DL AMC (Adaptive Modulation and coding) and UL AMC. AMC permutation enables multi-user diversity by choosing the sub-channel with the best frequency response. Diversity permutation is best suited for mobile applications, while contiguous permutation works best for fixed, portable and low-mobility situations. Time Division Duplex (TDD) frame structure The 802.16-2005 PHY does support TDD and Full and Half-Duplex FDD operation but TDD is the preferred duplexing mode and FDD will only be considered to address specific market opportunities. When using TDD only one channel is required for both downlink and uplink as opposed to the two separate frequency channels of FDD. The advantages of TDD are that it enables adjustment of the downlink/uplink ratio to efficiently support asymmetric downlink/uplink traffic instead of fixed and equal DL and UL bandwidths. The ratio between the uplink and downlink determines how time is split between them, for example 16 the ratio 2:1 means that the UL gets two times more time for transmission than the uplink. TDD also supports channel reciprocity, which is important when using MIMO and other closed loop advanced antenna technologies. Transceiver designs are also simpler for TDD implementations than FDD. On the other hand the discontinuous transmission and reception of TDD reduces the average power of the system. [5], [9] Modulation and Coding In Mobile WiMAX support for QPSK, 16 QAM and 64 QAM is mandatory in the DL but in the UL 64 QAM is optional. Both Convolutional Code and Convolutional Turbo Code are supported. The coding and modulation schemes supported in Mobile WiMAX can be seen in table 1. To show how it is possible to vary the data rates by choosing different MCS levels, the different PHY DL data rates for different MCS levels are also shown in Table 1 (for PUSC mode and 10 MHz channel). Modulation Code rate DL PHY data rate (10 MHz) QPSK 1/2 (CTC) 6,34 Mb/s QPSK 3/4 (CTC) 9,50 Mb/s 16 QAM 1/2 (CTC) 12,67 Mb/s 16 QAM 3/4 (CTC) 19,01 Mb/s 64 QAM 1/2 (CTC) 19,01 Mb/s 64 QAM 2/3 (CTC) 25,34 Mb/s 64 QAM 3/4 (CTC) 28, 51 Mb/s 64 QAM 5/6 (CTC) 31,68 Mb/s Table 1. MCS levels and DL PHY data rates in Mobile WiMAX [5] Other features of the PHY layer The term Adaptive Modulation and Coding (AMC) is used to denote the matching of the modulation, coding and other signal and protocol parameters to the conditions on the radio link. Hybrid Auto Repeat Request (HARQ) is a feature that enables fast response to packet errors and asynchronous operation, with a variable delay between retransmissions. Fast Channel Feedback (CQICH): A Channel Quality Indicator (CQI) channel is used to provide channel-state information from the user terminals to the base station scheduler. Relevant channel-state information that is fed back by the CQI channel includes: Physical CINR, effective CINR, MIMO mode selection and frequency selective subchannel selection. 17 All combined HARQ, CQICH and AMC provide robust link adaptation in mobile environments. The MAC (Media Access Control) Layer: At the second level of the OSI model there is the Data Link layer that can be divided into two sub-layers, MAC and LCC. The MAC sub-layer serves as an interface between the physical layer and the LCC sub-layer. It provides channel access control mechanisms and addressing mechanisms (MAC address) that make communication between terminals and/or networks possible. Also it provides the protocol and control mechanisms that are required for several stations to share the same physical medium. The MAC layer of the IEEE 802.16-2005 is quite complex and advanced. The main features are the following: Quality of Service (QoS): Is a term that refers to control mechanisms that can provide different priority to different users or data flows, or guarantee a certain level of performance in accordance with requests from the application program. For certain types of network traffic a defined QoS is required. For example: • Voice over IP (VOIP): Requires maximum latency tolerance, jitter tolerance and maximum sustained rate. • Streaming multimedia: Minimum reserved rate, maximum sustained rate, latency tolerance and traffic priority. • File Transfer Protocol (FTP): Maximum sustained rate, minimum reserved rate and traffic priority. • Data transfer, web browsing: Maximum sustained rate, traffic priority. In the Mobile WiMAX MAC layer, QoS is provided via service flows. First the base station (BS) and the user-terminal establish a unidirectional logical link between the peer MACs. The QoS parameters associated with the specific kind of data to be transmitted define the transmission ordering and scheduling on the air interface. For whatever service is in use the parameters can be dynamically managed through Media Access Control (MAC) messages. This QoS mechanism applies to both DL and UL for a better QoS in both directions. [5] 18 MAC Scheduling Service Is a scheduling service that is designed to deliver broadband data services in an efficient way over time-varying broadband wireless channel. These data services include voice, data and video. To achieve this the scheduling service uses the following properties: • • • • • Fast Data Scheduler: To efficiently allocate available resources in response to bursty data traffic and time-varying channel conditions. Located at each BS for faster response. In order for the scheduler to correctly determine the packet transmission ordering, the data packets are associated to service flows with defined QoS parameters. Because of the CQICH channel feedback the scheduler can choose the best coding and modulation for each allocation. Scheduling for both DL and UL Dynamic resource allocation: Frequency-time resource allocation on a perframe basis is supported in both DL and UL. As the resource allocation is delivered in Media Access Protocol (MAP) messages in the beginning of each frame it can be changed on a frame-by-frame basis in response to traffic and channel conditions. The amount of resource can range from one slot to entire frame. QoS Oriented: Data transport is handled on a connection-by-connection basis. For each connection there is a single data service with a set of QoS parameters that define its behavior. Frequency Selective Scheduling: Provides different services for different types of subchannels. For PUSC subchannels, where the subcarriers are pseudorandomly distributed across the bandwidth, the subchannels are of similar quality; QoS with fine granularity and flexible time-frequency resource scheduling is supported. For AMC subchannels, where the subchannels may have different attenuation, mobile users can be allocated to their strongest subchannels. Power Management To enable power-efficient MS operation Mobile WiMAX supports Sleep Mode and Idle Mode. In Sleep Mode the MS conducts pre-negotiated periods of absence from the Serving Base Station air interface to minimize power usage and usage of the Serving Base Station air interface resources. In Idle Mode the MS is periodically available for DL broadcast traffic messaging without registration at a specific BS. 19 Handoff The IEEE 802.16e-2005 standard supports three types of handoff: Hard Handoff (HHO): Is the mandatory mode and has been improved with the goal of keeping Layer 2 delays to less than 50 milliseconds. Fast Base Station Switching (FBSS): Is an optional mode. The set of BS that support FBSS is called an Active Set and one of the BS is defined as the Anchor BS. When operating in this mode the MS only communicates with the Anchor BS. At every time the MS chooses the BS with the strongest signal to be the Anchor BS. Macro Diversity Hand-over (MDHO). Is an optional mode. The BS’s that support MDHO are part of an Active Set maintained by the MS and BS. One of the BS’s in the Active Set is defined as the Anchor BS. As opposed to FBSS in MDHO the Ms communicates with all BSs in the Active set, but the regular mode of operation is the particular case when there is only one BS in the Active Set. In the DL diversity combining is performed at the MS as two or more BSs provide synchronized transmission of MS downlink data. In the UL selection diversity is performed as the transmission from a MS is received by multiple BSs. Network Architecture The end-to-end network architecture is not part of the IEEE 802.16e-2005 standard. As IEEE usually only defines the PHY and MAC layers. However operators and vendors have formed working groups to define standard network reference models. In Figure 2 a basic view of the all-IP Mobile WiMAX network can be seen. Figure 2. Network Architecture of Mobile WiMAX The three basic domains are: 20 User Terminals : Mobile, portable and fixed devices with WiMAX support. This could be for example cellular phones, MIDs (Mobile Internet Devices), UMPCs (Ultra Mobile PCs), laptops or Desktops. Access Service Network (ASN) : Provides a way to connect user terminals using OFDMA air interface to an IP backbone with session continuity (a session is not drop when user moves between wireless environment). An ASN consists of BSs and Access Gateways. The set of network functions in ASN include: Network discovery and selection of the preferred CSN/NSP (Network Service Provider) Network entry with IEEE 802.16e-2005 based layer 2 connectivity and AAA proxy. Relay function for IP Connectivity. Radio Resource Management. Multicast and Broadcast Control. Foreign agent functionality for inter-ASN mobility Paging and Location Management. Data forwarding. Service flow authorization. Quality of Service. Admission Control and Policing. Connectivity Service Network (CSN) : is defined as a set of network functions that provide IP connectivity services to the mobile subscribers. A CSN consists of network elements such as routers, AAA(Authentication, Authorization, Accounting) proxy/servers, user databases and Interworking gateway devices. A CSN is deployed by a NSP. AAA or Home Agent residing in CSN allocates the IP address. AAA also performs authentication, authorization, and accounting. Communication is through a RADIUS protocol. The policy server residing in the CSN is responsible to store the policy and QoS info of each subscriber, which is communicated to ASN during service flow creation. CSN is also responsible to access other IP networks. 2.1.3.Other features of Mobile WiMAX Smart Antenna Technologies Using OFDMA makes it possible for Mobile WiMAX to use smart antenna operations in rather simple way as they can be performed on vector-flat subcarriers and no complex equalizer are required to compensate for frequency selective fading. The MIMO technologies can be divided into the following three categories: • Spatial Multiplexing: To enhance the peak data transmission rate and overall throughput by using multiple streams with multiple antennas. • Space-Time Coding: A transmission diversity method for enhancing transmission robustness. • Beamforming: The received signal gain is improved and interference reduced with better coverage, higher capacity and reduced outage probability. 21 By using Adaptive MIMO Switching (AMS) between multiple MIMO modes, the MIMO system can use the transmission techniques one at a time or all together as is best at each time. In that way the spectral efficiency can be maximized with no reduction in coverage area. Fractional Frequency Reuse To maximize spectral efficiency Mobile WiMAX support the frequency reuse of one. In other words every cell/sector uses the same frequency. Normally, this can lead to heavy cochannel interference (CCI) and reduction in connection quality, especially for users at the cell edge. However, since in Mobile WiMAX each user operates on a subchannel that only occupies a small fraction of the bandwidth, configuring subchannel usage can significantly reduce the interference. Multicast and Broadcast Service (MBS) MBS is a technology that can be used to broadcast (or multicast) multimedia , similar to DVB-H, MediaFLO and DMB. MBS uses a Single Frequency Network (SFN) to obtain high data rate and coverage. It supports flexible allocation of radio resources, has low MS power consumption and low channel switching time. 22 2. properties of their spreading codes. 24 Scrambling In addition to spreading, part of the process in the transmitter is the scrambling operation. This is needed to separate terminals or base stations from each other. Scrambling is used on top of spreading by using scrambling codes. It makes the signals from different sources separable but doesn’t affect the bandwidth or the symbol rate as the chip rate is already achieved in spreading by the spreading codes. Main features in W-CDMA systems (R’99) In Table 2 some of the main features of W-CDMA systems are listed. Those features and others are better explained below. Multiple access method DS-CDMA Duplexing method Frequency division duplex/(time division duplex) Chip rate 3.84 Mcps Frame length (TTI) 10 ms Multirate concept Variable spreading factor and multicode Multiuser detection, smart antennas Supported by the standard, optional in the implementation Modulation QPSK Physical layer spreading factors 4 - 256 Table 2. Some features of W-CDMA 25 Multiple Access Method: DS-CDMA Multiple Access schemes are used to allow multiple users to access the radio resource in a way that maximizes the use of resource and minimizes the interference between users. In CDMA each user uses a different spreading waveform, which means that two or more different users can use the same carrier frequency and still transmit simultaneously. There doesn’t need to be any difference in time or frequency between the different users, only different waveforms. (See explanation of W-CDMA above). Frequency Division Duplexing (FDD) The basic mode of duplex is frequency division duplex, that is one 5 MHz channel is used for uplink and another 5 MHz channel is used for the downlink. TDD is also supported but FDD is by far the most common mode in UMTS WCDMA systems. Frequency duplex means that the radio transmitter and receiver operate at different frequencies and can thus send and receive at the same time without interference. FDD is more efficient than TDD in the case of symmetric traffic. In this case TDD tends to waste bandwidth during switch over from transmit to receive, has greater inherent latency, and may require more complex, more power-hungry circuitry. On the other hand TDD allows for assymetric traffic and has some other advantages (see TDD in WiMAX chapter). Chip rate Chip rate is the number of bits per second used in the spreading signal. A different spreading signal is added to the data signal to code each transmission uniquely. The number of chips (bits) in the spreading signal is significantly greater than the data bits. Chip rate is measured in "megachips per second" (Mcps), which is millions of chips per second. In WCDMA the chip rate is 3.84 Mcps, which leads to a carrier bandwidth of about 5 MHz. Frame length Frame length is also known as Transmission Time Interval (TTI). Before data is transmitted it is divided into blocks and then the bits in the block are encoded and interleaved (arranged in a non-contiguous way). This is done to make the data less sensitive to fading and interference on the radio link. The length of time required to transmit one such block determines the TTI. When the data is then received at the other end it is de-interleaved and decoded. But first all the bits of the specific block must be received. When all the bits of the block have been decoded and de-interleaved the receiver can estimate the bit error rate (BER). So as the shortest decodable transmission is one TTI the shortest period over which BER can be estimated is also one TTI. In networks with link adaptation (e.g. HSDPA and Mobile WiMAX) information about the estimated BER are used to adapt to the condition on the link. The shortest possible interval between reports of the BER are one TTI. So a very short TTI helps the network to adapt quickly to changing conditions in the radio link. On the other hand a longer TTI increases the efficiency of error-correction and compression techniques and the effect of 26 interleaving. In WCDMA (R’99) the shortest TTI is 10 milliseconds but 20ms, 40ms and 80ms are also supported. Variable user data rates W-CDMA supports highly variable user data rates, that is bandwidth can be obtained on demand. The user data rate is kept constant during each 10 ms frame. However, the data capacity among the users can change from frame to frame. This fast radio capacity allocation will typically be controlled by the network to achieve optimum throughput for packet data services. One of the things used to support variable user rates is the Orthogonal Variable Spreading Factor (OVSF). Different spreading factor means different code length. Different messages with different spreading factors (different code length) are combined and the orthogonality between them is kept. As the chip rate remains the same for all codes the short ones will be transmitted at a higher information rate than longer ones. Another thing used to support variable user rates is multicode transmission. Then each user can have data sent over more than one channel (that is each user gets to use more than one spreading or channelization codes). Channelization Codes Codes used to separate the different channels that may be present on a certain frequency in a CDMA system. To support a variety of data rates and spreading factors, it is necessary to chose channelization codes that are orthogonal no matter what their length are. WCDMA uses Orthogonal Variable Spreading Factor (OVSF) codes. Depending on the number of physical channels used and their data rate, the number of available channelization codes varies. The number of voice channels per carrier can be approximated to 250. (256 available orthogonal codes, minus some control channels). Softer and Soft handover During softer handover, a mobile station is in the overlapping cell coverage area of two adjacent sectors of a base station. The communications between MS and BS take place concurrently via two air interface channels, one for each sector separately. This requires the use of two separate codes in the downlink direction, so that the MS can distinguish the signals. The two signals are received by means of Rake processing; the fingers need to generate the respective code for each sector for the appropriate despreading operation. During soft handover, a mobile station is in the overlapping cell coverage area of two sectors belonging to different BSs. As in softer handover, the communications between MS and BS take place concurrently via two air interface channels from each BS separately. Both channels are received at the MS by maximal ratio combining Rake processing. 27 2.2.3 New features in HSDPA (release 5) HSDPA was designed to increase downlink packet data throughput of UMTS by means of fast physical layer (L1) retransmission and transmission combining as well as fast link adaptation and fast scheduling controlled by the BS. Several new channels are introduced in release 5. A new transport channel named HighSpeed Downlink Shared Channel (HS-DSCH) is the primary radio bearer. For the associated signaling a channel called high-speed shared control channel (HS-SCCH) has been added in the downlink and in the uplink the high-speed dedicated physical control channel (HS-DPCCH) that carries information about the link condition. High-Speed Downlink Shared Channel (HS-DSCH) HS-DSCH is the channel used to send packets on the downlink to the UEs and comes instead of the downlink shared channel (DSCH) used in R’99. Similar to the DSCH, the HS-DSCH resources can be shared between all users in a sector. The primary channel multiplexing occurs in the time domain, with a TTI of 2ms (or three slots.) In R’99 the shortest TTI is 10ms. This significant reduction of the TTI size will improve the link adaptation rate and the AMC efficiency as well as reducing the round trip delay between the BS and the UE. In DCH (R’99) it was possible to use variable spreading factor between 4 and 512, in HSDPA a fixed Spreading Factor of 16 is used for code multiplexing. Therefore, within each 2 ms TTI a maximum of 15 parallel codes allocated to the HS-DSCH can be used for code multiplexing. All 15 codes can be assigned to one user or split between several users in the sector, during each TTI. The number of parallel codes allocated to each user depends on cell loading, QoS requirements and the UE code capabilities (5, 10 or 15 codes). [11] The following list is a list of the main differences between HS-DSCH and DCH (R’99). Better explanations follow below. • • • • • • • Use of physical layer retransmission and retransmission combining (H-ARQ) Lack of fast power control, link adaptation and AMC used instead. Support of higher order modulation (16 QAM) Shorter TTI (2ms) Lack of soft handover Fixed spreading factor BS based Fast packet scheduling for controlling allocation of users. 28 Hybrid Automatic Repeat Request (H-ARQ): Automatic Repeat reQuest is an error control method for data transmission. If a receiving part of the system has received a correct data frame it sends an ACK (acknowledge character) message to the transmitting part of the system. If it receives an erroneous data frame it sends an NACK (negative acknowledge character) and the packet is discarded. In H-ARQ two methods are used to improve this scheme: Chase combining and incremental redundancy (IR). In chase combining the erroneous packet is not discarded but stored and later combined with the retransmitted packet to increase the probability of successful decoding. In IR the different transmissions are coded differently instead of the same code being repeated as in chase combining. [12] Adaptive Modulation and Coding (AMC): Denotes the matching of the modulation, coding and other parameters to the condition on the radio link. In HSDPA AMC is used to ensure that all users achieve the highest possible data rate regardless of its location in the cell. The modulation scheme and coding is changed for each user depending on signal quality and cell usage that has been reported as feedback from the UE. QPSK is the initial modulation scheme but in good ratio conditions 16QAM can be used to improve the data throughput significantly. The spreading factor cannot change but the coding rate can change between 1/4 and 3/4. The higher coding rate reduces the number of errors. [13] & [14] Fast Packet Scheduling: Allows the HS-DSCH channel to take advantage of favorable channel conditions to make best use of available radio conditions. Each EU periodically reports on the signal quality to Node B (Base Stations). That information is then used to decide which users will be sent data on the next 2ms frame and how much data can be sent to each user. Both Convolutional Coding and Turbo coding are supported but previously only CC has been supported. 2ms TTI: The shorter time interval enables higher speed transmission in the physical layer, so that the system will be more reactive to changing link conditions and can reallocate capacity to users quicker. Handover: HSDPA does not use soft handover. This is because the AMC, H-ARQ and fast packet scheduling are techniques that require a constant one-to-one connection between the HSDPA mobile terminal and the BS. Thus hard handover, in which the destination BS is selected each time the cell changes, is needed. Since the only traffic supported by HSDPA is delay-tolerant data traffic soft handover is also not as necessary as when dealing with voice traffic. 29 A higher modulation scheme, 16QAM, has also been added to the already existing QPSK modulation used for the R’99 channels. With also a lower encoding redundancy this gives an increased instantaneous peak data rate. The modulation to be used at each time is adapted to the link conditions. The code rate is also adapted to the condition. HS-DSCH uses Turbo coding and supports the following code rates: 1/4, 1/2, 5/8 and 3/4. In addition to the modulation scheme and code rate HSDSCH can also choose different number of multi-codes to provide different data rates to different users. Table 3 shows the UE categories defined in release 5. Category Max nr. of parallel codes per HS-DSCH Minimum interTTI interval Transport channel bits per TTI Achievable m aximum data rate (Mb/s) 1,2 5 3 7298 1,2 3,4 5 2 7298 1,8 5,6 5 1 7298 3,6 7,8 10 1 14411 7,2 9 15 1 20251 10,2 10 15 1 27952 14,4 11 5 2 3630 0,9 12 5 1 3630 1,8 Table 3. HSDPA terminal capability categories [15] Changing the coding rate and the modulation scheme can largely vary the terminal data rate. Table 4 shows the maximum data rates when the number of codes is kept constant at 15 codes but the coding rate and the modulation scheme are varied. Modulation Effective code rate Max throughput (Mb/s) QPSK 1/4 (turbo) 1,8 QPSK 1/2 (turbo) 3,6 QPSK 3/4 (turbo) 5,3 16 QAM 1/2 (turbo) 7,2 16 QAM 3/4 (turbo) 10,2 16 QAM 4/4 (turbo) 14,4 Table 4. Peak data rates with number of codes fixed at 15 [15],[10] Other channels In addition to the HS-DSCH channel, three other physical channels are new in HSDPA. 30 High Speed-Shared Control Channel (HS-SCCH): Informs the user that data will be sent on the HS-DSCH 2 slots ahead. HS-SCCH has a fixed rate of 60 kbps (spreading factor 128), which allow 40 bits per slot to be carried. Part of the information carried on HS-SCCH like which codes to de-spread in the HSDSCH, needs to be available before beginning to use the HS-DSCH. Thus the HS-SCCH is divided into two parts. The first part contains the most urgent information about the modulation and coding. This allows terminals to support only 5 or 10 codes even though the code space allocation is 15 codes. The second part carries information that is not as urgent. Which ARQ process is being transmitted and if the transmission is new or related to an older transmission. When HSDPA is operated using the time multiplexing principle there can only be one HS-SCCH channel configured. This means that only one user can receive data at a time. When code multiplexing is being used, more than one HS-SCCH channels are needed. A single terminal cannot use more than 4 HS-SCCH channels. The channel coding is one-third convolutional coding. Uplink High Speed-Dedicated Physical Control Channel (HS-DPCCH): Carries acknowledgement information and current channel quality indicator (CQI) of the user from the terminal to the base station. The base station then uses this value to enable the link adaptation and physical layer retransmissions. The H-ARQ feedback informs the base station whether the packet was decoded correctly or not and the CQI informs the base station scheduler about what data rates the terminal expects to be able to receive at that moment. The HS-DPCCH has a fixed spreading factor of 256 and a 2ms (three slots) structure. The fist slot is used for the H-ARQ information and the remaining to for the CQI information. An associated Dedicated Physical Channel (DPCH), a parallel code channel, provides the signaling. 31 High Speed-Physical Dedicated Shared Channel (HS-PDSCH): Is mapped to the HS-DSCH transport channel that carries actual user data. The table below shows some of the differences between WCDMA (R’99) and HSDPA (R’5): WCDMA (R’99) HSDPA (R´5) Modulation Scheme QPSK QPSK/16QAM Multiplexing Scheme (Downlink) DS-CDM DS-CDM/ TDM Duplex Method FDD FDD Channel Bandwidth 5 MHz 5 MHz Frame Size 10 ms (20, 40 and 80 ms) 2 ms Coding Cc Cc, Turbo DL Peak Data Rate 384kbps-2Mb/s 14 Mb/s Frequency Reuse Factor 1 1 Table 5. Differences between WCDMA (R´99) and HSDPA 32 End to End Network Architecture The UMTS architecture: An UMTS network can be divided into three interacting domains: • • • Core Network (CN):Provides switching, routing and transit for user traffic and contains the databases and networking management functions. UMTS Terrestrial Radio Access Network (UTRAN): Provides the air interface access method for UE. User Equipment (UE): The handset that the user needs to connect to the system. This can be for example a cellular phone or a laptop. Figure 3. UMTS network architecture A mobile radio network generally includes a set of base stations and base station controllers. In UMTS this network is called the UMTS terrestrial radio access network (UTRAN), a base station is called a Node B and a base station controller is called a radio network controller (RNC). The UTRAN communicates with mobile stations (UEs) via a Uu interface and with a core network (CN) via an Iu interface. The RNCs are connected to each other via a lur interface and to the BS (Node B) via a lub interface. 33 2.3 Comparison of WiMAX and HSDPA 2.3.1 Technical comparison The most fundamental difference between the two technologies is based on that HSDPA evolved from 3G CDMA standards and networks that provide mobile voice services while WiMAX has evolved from system concepts designed for fixed wireless access. WiMAX is optimized for broadband data services where as mobility was an additional feature added to the standard. Similarities : Despite their different background there are several technical features that the two technologies have in common. Those include: • Adaptive Modulation and Coding (AMC) • Hybrid ARQ • Fast Scheduling • Hard Handoff All those things have already been mentioned in the overview chapters. Both standards support AMC. Both technologies have similar approach on HARQ and support the chase combining (cc) method. The advantage of using HARQ CC is that since the retransmissions are identical, it is easy to extend to different transmission techniques such as MIMO. In order to enable fast response to traffic and channel condition variations both standards have included Fast Scheduling. However there is a little difference in that WiMAX applies the fast scheduling on a per frame basis and broadcasts the downlink/uplink scheduling in the MAP messages at the beginning of each frame. While HSDPA supports dedicated scheduling, where only a subset of UEs with pending data is selected to transmit over a given time interval with selected rates restricted. The user devices periodically transmit an indication of the downlink signal quality and the base station uses that information to decide which users will be sent data on the next 2 ms frame and how much data should be sent to each user. The network bandwidth allocated to HSDPA users is determined by the network and is “semi-static”. It can be modified while the network is operating but not on a per frame basis. Neither WiMAX nor HSDPA supports the traditional mobile network soft handover. Instead HSDPA uses “Network initiated Hard Handover” and WiMAX “Network Optimized Hard Handover”. Both methods are more bandwidth efficient than the soft handover where more than one BS in a mobile active set of base stations send simultaneously in order to minimize the handover delay. [16,page 27-30] 34 Differences HSDPA is based on a traditional mobile telephony while Mobile WiMAX is based on fixed wireless broadband technology. The Mobile WiMAX physical layer is based on an OFDMA while HSDPA is a CDMA based system. OFDMA has some advantages over CDMA and it is likely that future systems will be based on OFDMA technology. For exampleLTE, the next generation of UMTS systems, will be based on OFDM. [17], [15] Some of the features of OFDMA that make it well suited for high speed wireless networks are listed below: Tolerance to Multipath and Self-Interference : Since the subchannels maintain their orthogonality in a multipath channel the number of multipath components does not limit the performance of the system as long as the multipath is within the cyclic prefix window. Scalable Channel Bandwidth : Orthogonal Uplink Multiple Access: In OFDMA systems users are allocated different portions of the channel where as in CDMA each user transmits over the entire channel. This means that in OFDMA there is no multiple access interference (MAI) between multiple users. In CDMA orthogonal spreading codes are used to avoid MAI but due to the uplink synchronization issues, asynchronous CDMA is used in the uplink in most practical CDMA systems and there will be interference and reduced spectral efficiency. Frequency Selective Scheduling: As only a portion of the channel is occupied by the WiMAX signals frequency selective scheduling can be used to choose sub channels with the best condition at each time and hence improve QoS. Advanced Antenna Technology: For smart antenna technologies the processing complexity scales with the channel bandwidth. Since in CDMA the signals occupy the entire bandwidth this becomes quite a problem when used in broadband wireless channels and limits the options of using advanced Antenna Technology. OFDMA on the other hand is well suited for these technologies. From the beginning IEEE 802.16 was designed as an all-IP technology. It is based on an advanced all-IP core network, which includes interworking with IP Multimedia Subsystems (IMS) and with 2G, 3G and IP-based technologies. HSDPA supports GERAN (GSM EDGE Radio Access Network) and UTRAN (UMTS Terrestrial Radio Access Network) based access networks. GERAN and UTRAN can carry many traffic types from real-time Circuit Switched to IP based Packet Switched. This in turn means a high degree of compexity compared to an all-IP architecture. In the 3GPP LTE (Long Term Evolution) an All-IP Network is proposed as the future of UMTS. 35 Other differences Duplex method : Mobile WiMAX will most commonly use TDD while HSDPA generally uses FDD. FDD is more efficient than TDD in the case of symmetric traffic but TDD allow for assymetric traffic and as the downlink traffic is usually much heavier than the uplink traffic, assymetric traffic can be very practical. TDD requires system-wide frame synchronization to counter interference issues and the discontinous transmissions reduce the average power. On the other hand TDD assures channel reciprocity and thus better supports link adaptation, MIMO and other advanced antenna technologies. Modulation : Mobile WiMAX supports QPSK, 16 QAM and 64 QAM modulation but HSDPA only supports QPSK and 16 QAM (although in release 7 64 QAM has been added). In other words Mobile WiMAX supports higer order modulation than HSDPA. With 16 QAM 4 bits can be carried per symbol instead of 2 bits per symbol with QPSK. With 64 QAM 6 bits can be carried per symbol and thus higher data rates can be reached. Table 6, gives an overview of some of the features of Mobile WiMAX and HSDPA. HSDPA Mobile WiMAX Base Standard WCDMA IEEE 802.16e-2005 Duplex Method FDD TDD Downlink CDM-TDM OFDMA Uplink Multiple Access CDMA OFDMA Channel Bandwidth 5 MHz Scalable: 5,7,8.75, 10 MHz Frame size DL=2 ms UL=2/10ms 5 ms TDD Modulation DL QPSK, 16QAM QPSK, 16QAM, 64QAM Modulation UL BPSK, QPSK QPSK, 16QAM Coding CC, Turbo CC, Turbo DL Peak Data Rate 14,4 Mb/s 46 Mb/s DL/UL =3:1 (BW=10MHz and 2*2 MIMO) H-ARQ Fast 6-channel Asynchronous CC Multi Channel Asynchronous CC Scheduling Fast scheduling in DL Fast Scheduling in DL and UL Handoff Network initiated Hard Handoff Network optimized Hard Handoff Table 6. Comparison of some of the features of Mobile WiMAX and HSDPA 36 2.3.2 Market situation and opportunities Many operators around the world have invested in R’99 UMTS networks. For them HSDPA offers a significant service upgrade and an opportunity to accelerate the Return of Investment. HSDPA networks are already widely deployed and handsets have been on the market since 2006.[18] For Mobile WiMAX it is necessary to build new networks, and the manufacturing of handsets has been quite complicated and required a totally new set of chips and platforms. The fact that Intel, a giant on the semiconductor and microprocessor market, is a leader in promoting WiMAX and has included its WiFi/WiMAX solutions in it’s 5th generation Centrino platform for labtops, Montevina, and its next generation platform for MID’s and UMPC’s could help the promotion of WiMAX. [19] Current and future deployment of mobile WiMAX Deployment of mobile WiMAX is yet in early stages. However many companies have plans of deployment and some are already on the market. In South Korea a technology called WiBro has been developed by the Korean telecommunication industry. WiBro is a wireless broadband Internet based on the IEEE 802.16-2005 standard and is interoperable with WiMAX. In 2005 Samsung introduced the world’s first Wibro phone, M8000, and Wibro PDA. The services offered in the M8000 include broadcasting, home networking, video-conferencing and video on demand. [20] , In the US, Sprint Nextel announced in 2006 that they had chosen WiMAX as its next generation 4G technology platform. Working with Intel, Motorola and Samsung, Sprint Nextel are planning on a nationwide network and mobile WiMAX products that will support advanced wireless broadband services for computing, portable multimedia, interactive and other consumer electronic devices.[17] Motorola has been developing Mobile WiMAX chipsets to use in its WiMAX devices in cooperation with Texas Instruments and the first chipsets are scheduled to support commercial Motorola WiMAX devices in 2008. [21],[22] Also Nokia is planning on introducing its first Mobile WiMAX handset in 2008, based on a Intel/Nokia chips. [14] LG is another manufacturer that is involved in WiBro/WiMAX. Current and future deployment of HSDPA In November 21, 2007 GSA, the Global mobile Suppliers Association, confirmed that 154 HSDPA networks had entered commercial service in 71 countries. [23] Also according to the GSA, 403 HSDPA devices are announced. Those include 203 phones/consumer devices, 39 wireless routers, 61 laptops and 100 devices for laptop connectivity (USB modems etc). The number of HSDPA networks, devices and subscribers is constantly growing. For example WCDMA/HSDPA was responsible for 75% of the mobile subscription growth in Western Europe in Q3 2007. [23] 37 The next upgrade of the UMTS protocol (release 6) includes High Speed Uplink Packet Access (HSUPA) that provides improved up-link performance of up to 5.76 Mb/s theoretically. The standard is already established and a small number of networks have already been launched. A new 3GPP technology, LTE will use scalable OFDM air interface, MIMO, 64 QAM and have an all-IP network. Performance studies A few studies have been carried out to compare the performances of Mobile WiMAX and HSDPA. The WiMAX Forum published the study:Mobile WiMAX-Part II: A Comparative Analysis in May 2006. [5] In this study Mobile WiMAX is compared to HSPA and EVDO (Rev A & Rev B). The simulations results show a clear advantage of Mobile WiMAX in both spectral efficiency and sector throughput. For Mobile WiMAX the net information throughput per channel was found to be 14.1 Mb/s in the downlink using 10 MHz bandwidth, TDD 3:1 and MIMO and 9.1 Mb/susing SIMO. For HSDPA this number was 3.9 Mb/s using 2 * 5 MHz FDD and single antenna Tx and dual antenna Rx (1x2, SIMO) with RAKE receivers. In the paper “The Performance Comparison between WiBro and HSDPA” a team of Network Engineers at the Network R&D Center of SK Telecom in Korea published the results of their performance comparison of WiBro and HSDPA. The two factors compared where coverage and capacity. The team concluded that the overall performance of WiBro was better than HSDPA. The biggest difference was that WiBro showed better robustness in multi-path fading channels. On the other hand the shorter TTI of HSDPA, 2ms for HSDPA and 5ms for WiBro, made HSDPA more robust to the dynamic channel variation. Another advantage of HSDPA, according to the Korean Team, is that it uses the well-established WCDMA network. [13] In may 2007 Ericsson published the paper “ HSPA, the undisputed choice for mobile broadband”. In the paper the performance of HSPA and Mobile WiMAX is said to be comparable. “Both technologies offer similar peak data rates, spectral efficiency and network complexity”. However, in this paper the coverage of HSPA is found to be better than the coverage of Mobile WiMAX. [9] 38 3.1 Network dimensioning and planning 3.1.1 Frequency band The frequency band in which a network is operated is an important factor of the network dimensioning. The use of radio frequency bands is regulated by governments. When regulators have decided how a certain frequency band should be used and regulated then frequency licenses are allocated to operators that fulfill the conditions, for example through auctions. Radio communication technologies are designed to be deployed in certain frequency bands. Cellular networks use the UHF band whose frequency is between 300 MHz and 3 GHz. The signal propagation characteristics are better in the lower frequency bands. Many radio communication standards can be designed for use in the same frequency band but only one can be deployed in a particular area in a certain frequency band. That´s why the regulations and allocations are needed. Therefore many standards define several different frequency bands that the networks can be operated in. For example 15 different FDD paired frequency bands and 6 different TDD frequency bands are defined in the UMTS standard. [24] Networks based on the IEEE 802.16e-2005 standard will operate on licensed bands between 2 GHz and 6 GHz. Release-1 Mobile WiMAX profiles cover 5, 7, 8.75, and 10 MHz channel bandwidths for licensed spectrum allocations in the 2.3 GHz, 2.5 GHz, 3.3 GHz and 3.5 GHz frequency bands. [5] So in which frequency band a network will be operated can depend on the standard, regulations in the selected geographical area and possibly the preferences of the operator. In Iceland a special law, Act8/2005 on Third Generation Mobile Telephony states how frequencies for UMTS networks should be assigned and what conditions the operators have to fulfill to get the frequency license. This act applies to the allocation of frequency ranges 1920-1980 MHz, 2010-2025 MHz and 2110-2170 MHz, that is band I according to 3GPP specifications. [2] No frequency band has yet been assigned for WiMAX in Iceland but it is likely that the Icelandic Post and Telecom Agency will follow other European countries, in particular the Nordic countries. The European Conference of Postal and Telecommunications Administrations (CEPT) has recommended that the frequency band 2500-2690 MHz should be allocated as 2 x 70 MHz for FDD and 50 MHz for TDD. Some regulators think that there is a bigger demand for TDD than the CEPT guidelines imply. Norway has decided to allocate 2 x 40 MHz for FDD and 100 MHz for TDD. The Swedish National Post and Telecom Agency has also announced that it will issue licenses in the 2500-2690 MHz frequency band through an auction planned for the first half of 2008. The licenses are service and technology neutral. That is, they can be used for mobile telephony or wireless broadband services. [25],[26] 39 In order to make the design of our system as realistic as possible we have chosen to use the 2.5 GHz band as the operating frequency band for mobile WiMAX and as the frequency ranges 1920-1980 MHz, 2010-2025 MHz and 2110-2170 MHz have already been allocated to UMTS in Iceland we have chosen the 2.0 GHz band as the frequency band for HSDPA. 3.1.2 Fading model In mobile systems the radio link conditions are constantly changing. Thus the signal strength varies as a function of time and deviates around a mean value. To take the uncertainty of the signal strength into account fading models are used. When electromagnetic waves meet obstructions on their way through the air, the wavefronts are scattered and the waves travel along multiple paths. The scattered portions of the signal arrive at different times at the receiver and have been affected in different ways. This problem is called fading and refers to time variations in the received signal power because of changes in the path or the medium that the signal has traveled through. Fading is divided into small-scale fading (also called multipath fading or fast fading) and large-scale fading (also called shadowing or slow fading). Small-scale fading happens because of small obstructions in the propagation path or small movements of the mobile station. Large-scale fading happens because of large movements or large obstructions in the propagation path. To analyze the effects of fading, mathematical channel models are used. Fading channel models are divided into categories depending on some properties. Small-scale fading channels can be divided into categories based on the channels multipath time delay spread (Tm ), Doppler spread (Bd ), coherence bandwidth (Bm ) or coherence time (Td ). These parameters are related in the following way: Bd = 1/ Td and B m =1/ T m By looking at the coherence bandwidth/delay spread we can divide small-scale fading channels into frequency selective fading and frequency nonselective fading channels. The coherence bandwidth is the range of frequencies over which the channel can be considered flat (i.e. the channel passes all spectral components with equal gain and linear phase. [8] Frequency selective fading : If the bandwidth of the input signal is larger than the coherence bandwidth of the channel, the different frequency components of the signal will be subject to different attenuation and phase shift. Hence the distortion is said to be frequency selective and if the distortion leads to sharp drop in amplitude in some part of the spectrum we have frequency selective fading. [8] Frequency nonselective (flat) fading : Happens when the signal bandwidth is smaller than the coherence bandwidth. An amplitude variation will be over the signals entire frequency spectrum. Thus the shape of the waveform does not change, only the received power level. [8]) 40 In broadband wireless systems, the symbol rate is usually high. If the bandwidth of the signal becomes greater than the coherence bandwidth we get frequency selective fading. One way to avoid this problem is to use sub-channels. This is the method used in OFDM and thus in the IEEE 802.16e standard. Then we get many narrowband flat fading subchannels instead of one broadband frequency selective fading channel. Shadow fading distribution is usually modeled as a lognormal distribution that describes the variation of the decibel value of the mean signal as a normal or Gaussian distribution. Multipath fading is usually modeled using Ricean or Rayleigh distributions. The models assume that the signal is sufficiently narrowband so that the fading is flat, i.e. not frequency-selective. OFDMA and CDMA systems both have quite effective but different methods to combat multipath fading. In OFDM systems, a wide frequency band is divided into multiple subcarriers, each of which is assumed to experience flat fading. In CDMA rake receivers are used that have several fingers that differ a little in time to tune into the different components of the multipath signal. Then each component is decoded independently and then they are combined to make use of path diversity. In UMTS R’99 the fast fade margin is defined as the headroom needed in the MS TX power to maintain adequate closed loop fast power control. HSDPA does not use closed loop fast power control and thus does not need this headroom. As both systems are capable of significantly reducing the multipath fading and even taking advantage of it, there is no need for a big fast fading margin. In the link budget only one fading margin is given that is supposed to include both fast and slow fading (a UE can not move both fast and slow at the same time but there can always be obstacle of different sizes). The value will be the same for both HSDPA and Mobile WiMAX. The value of the total margin is 9 dB . The same value of log-normal shadowing standard deviation will be used for both technologies: 8 dB. [5], [27] 41 3.1.3 Path loss model Path loss models are used to predict the path loss or power attenuation of a signal traveling from a transmitter to receiver. The simplest path loss model is the free space model, which describes the loss in signal strength when the signal travels through free space where there are no obstacles. The power loss is proportional to the square of the frequency of the signal and the square of the distance between the transmitter and receiver. Equation (1) describes the free space path loss: L= (4* *r/ )^ 2 = 32.4 + 20*log(r) + 20*log(f) (1) where: : Is the signal wavelength f: Carrier frequency r: The cell radius (distance between transmitter and receiver) The free path loss model cannot be used in most practical situations. Instead empirical models based on measurements and observations in real propagation environments are used. We will consider two different path loss model for our systems, both models predict mean path loss as a function of various parameters, such as the cell radius, antenna height and environment considered. Same values will be used for both the HSDPA and Mobile WiMAX systems everywhere possible. The environment, the antenna heights, the reference distance d 0 have the same values. The only parameters that will not have the same values are the carrier frequency, f c , and the frequency correction term (in Erceg). COST 231-Hata The first model considered is one of the most widely used path loss models, the COST 231-Hata model. This is a model that extends the Okumura-Hata model to cover a wider range of frequencies. The formula for this model is given by: L=46.3+33.9log(f c )-13.82log(hb )+Alog(r)-B + G Where: A = 44.9-66.5log(h b ), B =(1.1*log(f c )-0.7)*(hm )-(1.56*log(fc )-0.8) (for small to medium cities) G= 0 dB (for medium sized cities and suburban areas, would be 3 dB for metropolitan areas) f c : Carrier frequency h b : Height of BS antenna h m : Height of MS antenna r: The cell radius L: Sum of all gains and losses from the receiver to the transmitter. 42 This model is valid for medium to small size cities, for frequencies around 2 GHz and antenna heights between 30m-200m (BS) and 1m-10m (MS). [8, page 85-87] Erceg Another well known model is the Erceg model. The path loss model is derived from experimental data collected by AT&T Wireless Services across the United States in 95 existing macro cells at 1.9 GHz. The model is for suburban areas, and it distinguishes between three different terrain categories. The model applies to base station antenna heights from 10 to 80 m, and distances from 0.1 to 8 km. [28] The formula for this model is given by: L = Ld0 + 10*n*log(d/d 0 ) + Xf + X h + s (3) where Ld0 is the free space path loss at d 0 , d 0 = 100 meters, n is the path loss exponent, d is the distance in meters, Xf is the frequency correction term, Xh is the receive antenna height correction term, and s is the shadow fading component. The formula to calculate the path loss exponent n is: n = a - b* hb + c/hb (4) To calculate the frequency correction term we use: Xf = 6 * log (f c /2000) (5) Where a,b and c are constants that represent different terrain categories. The values of a, b and c for the different categories can be seen in Table 7. Model parameter Category A (Hilly, heavy trees) Category B Category C (Intermediate) (flat, few trees) a 4,6 4 3,6 b 0,0075 0,0065 0,005 c 12,6 17,1 20 Table 7. Model Parameters. To calculate the receive antenna height correction term we have: X X = -10.8* log (h m /2000) (for category A and B) (6) = -20* log (h m /2000) (for category C) (7) h h 43 In our calculations the following values will be used in the path loss models: Height of BS antenna, h b : 30 m Height of MS antenna, h m : 2 m Shadow fading component: 8 dB 3.1.4 Link Budget A link budget is where all gains and losses that a signal will experience while traveling between a transmitter and receiver are listed and summed up to find the received power when the transmitted power is given. Attenuation of the transmitted signal due to propagation, antenna gains, penetration losses and implementation losses are examples of losses and gains that are included in a link budget. Margins because of interference, fading and penetration loss (indoor coverage) are also taken into account. Later we will use the Link Budget to find the maximum allowable path loss (MAPL). That is how much loss a signal can experience but still be detected at the receiver. This parameter will than help us find the cell radius. 44 HSDPA Mobile WiMAX Mobile WiMAX (5 MHz) Transmission Power (dBm) 24 23 23 Tx antenna gain (dBi) 0 0 0 Tx loss (dB) 1 1 1 EIRP (dBm) 23 22 22 Penetration loss (dB) 15 15 15 Fading Margin (dB) 9 9 9 Interference Margin (dB) 3 3 3 Total Margin (dB) 27 27 27 Rx antenna gain (dB) 18 18 18 Subchannelization gain (dB) 0 12 15 Processing gain (dB) (17,8) 0 0 Rx loss (cable loss in BS) (dB) 3 3 3 Rx signal strength (dBm ) Table 8. Uplink link budget 45 11 (28,8) 22 25 (10 MHz) Parameter HSDPA Mobile WiMAX (5MHz) Mobile WiMAX (10MHz) Transmission Power (dBm) 43 41 41 Tx antenna gain (dBi) 18 18 18 Tx loss (dB) 3 3 3 EIRP (dBm) 58 56 56 Penetration loss (dB) 15 15 15 Fading Margin (dB) 9 9 9 Interference Margin (dB) 3 3 3 Total Margin 27 27 27 Subchannelization gain (dB) 0 0 0 Processing gain (dB) 12 0 0 Rx loss (dB) 1 1 1 Rx signal strength (dBm ) 42 28 28 Table 9. Downlink link budget • Transmission power: According to our references the output power for both base station and mobile station is slightly lower for Mobile WiMAX than for HSDPA. In the downlink the difference in the numbers we have assumed is 2 dB but in the uplink 1 dB. HSDPA’s advantage is possibly because the HSDPA equipment has been tested and developed longer and further than the WiMAX equipment. [5, table 10], [10, page 148], [29], [30] • Antenna gain : In our link budget we have assumed that both HSDPA and Mobile WiMAX will have antenna gains of 18 dBi. This assumptions were made after looking at for example the following references: [5, table 10], [10, page 130], [30] • Tx loss: Is the cable loss in the base station in the case of the downlink and the body loss at the terminal in case of the uplink. We have assumed same values for both technologies, 3 dB in the downlink and 1 dB in the uplink. • EIRP (the effective isotropic radiated power) : can be found by using the following equation: EIRP = Tx Power (dBm) + Antenna Gain (dBi) - Implementation loss (dB) And is the amount of power that would have to be emitted by an isotropic antenna to achieve the peak power density in the direction of maximum antenna gain. 46 • Penetration Loss: Is the loss the signal experiences when it travels through walls and is added to the link budget to allow for indoor coverage. This value, 15 dB is a typical assumption that applies to both technologies. It is assumed that 70 % of the users are indoors and thus experience penetration loss. • Fading Margin : We have chosen to use total fading margin of 9 dB for both technologies. This margin accounts for both multipath fading and shadowing. An explanation of fading margins was given in chapter 3.1.2. • Interference Margin : In WCDMA systems the interference margin is needed because of the loading of the cell, the load factor. The more load that is allowed in the cell the larger interference margin is needed. In IEEE 802.16e the interference margin is needed because of cochannel interference. Margins quoted for both technologies are typically around the standard 2-3dB. Hence we assume the interference margin to be equal for both technologies, 3 dB (corresponding to 50% load). [27] • Subchannelization gain : For 5 MHz bandwidth in OFDMA, we have 512 subcarriers organized into 17 different sub-channels. We can concentrate the available power into one sub-channel, therefore gaining a factor of 17, which is about 12 dB compared to single carrier systems [13]. For 10 MHz bandwidth there are 1024 subcarriers organized into 34 subchannels. When all available power is concentrated into one subchannel the gain is of factor 34 or 15 dB. This applies only to the uplink. (See subchannelization chapter) • Processing gain : As discussed in chapter 2.2.2. in CDMA (and all spread spectrum systems) systems the length of the spreading code determines the spreading factor or how much the signal can be spread. This in turn determines the Processing Gain or the ratio of the spread bandwidth to the unspread bandwidth. The HS-DSCH channel uses a fixed spreading factor 16. Which means that the processing gain is 12 dB. If we assume a non-real-time 384 kbps data service in the uplink we will have processing gain: 10*log (3840/384) = 10 dB. A 12,2 kbps data rate give processing gain: 10*log(3840/12,2) = 25 dB. A 64 kbps data rate gives processing gain 10* log(3840/64) = 17,8 dB • Rx loss : Is the cable loss in the base station in the case of the uplink. In the case of the downlink it is the loss experienced when the terminal (UE) is close to the user’s head. In UMTS this is called body loss. We have assumed same values for both technologies, 1 dB in DL in 3 dB in uplink. 47 3.1.5 Other system characteristics HSDPA A dedicated HSDPA network The HSDPA system is assumed to be HSDPA-only system. That is the channels are used for HSDPA traffic only and no R’99 traffic. This is usually not the case in UMTS systems but as we are comparing throughput and coverage of HSDPA and Mobile WiMAX which both only consider packet switched traffic the results will be more comparable if the HSDPA traffic gets all the resources of the network instead of having to share them with R’99 traffic. User Equipment Like we saw in Table 3, in chapter 2.2.3. there are 12 different UE categories for HSDPA. In our calculations we will assume that all UEs support 15 HS-PDSCH codes and 16-QAM modulation and can reach peak data rate of 10.2 Mb/s. Duplexing method and cell configuration The duplexing method used in the HSDPA network is FDD and thus we have paired frequency bands. Each channel has 5 MHz for downlink and 5 MHz for uplink, each cell has three sectors which each have their own channel. The total frequency allocation is thus 30 MHz. Mobile WiMAX Duplexing method and cell configuration For Mobile WiMAX, two types of systems will be considered. Both will use TDD as duplexing method and thus the same channel will be used for both uplink and downlink. The channel will be used for downlink transmissions 2/3 of the time and for uplink transmissions 1/3 of the time. The first system will have 5 MHz channel bandwidth and thus the total bandwidth of each cell will be 15 MHz (3 sectors per cell). The second system will have 10 MHz channel bandwidth and thus 30 MHz total bandwidth. In Table 10 a list of the common system parameters is given. Those values were chosen by comparing a few references and should be quite typical values for mobile radio networks. [5], [27], [25], [29], [30] 48 Com mon system parameters Channel bandwidth 5 MHz (and 10 MHz for WiMAX) Cell configuration 3 sector/cell BS Antenna height 30 m MS Antenna height 2m BS Antenna Gain 18 dBi MS Antenna Gain 0 dBi Area Coverage Probability 95% Standard Deviation (log-normal shadowing) 8 dB Penetration loss 15 dB FadingMargin 9 dB Interference Margin 3 dB (50% load) Path loss Model Cost 231 Hata / Erceg Table 10. Common system parameters for HSDPA and Mobile WiMAX. The system parameters that are different for the two technologies can be seen in Table 11. System Parameters HSDPA Mobile WiMAX Frequency 2000 MHz 2500 MHz Duplexing method FDD TDD, DL/UL =3:1 BS transmitted power 43 dB 41 dB MS transmitted power 24 dB 23 dB Modulation schemes 16QAM/QPSK 64QAM/16QAM/QPSK Subchannelization gain (UL) Does not apply 12 dB (5MHz) 15 dB (10MHz) Processing gain 12 dB DL and 17,8 dB UL Does not apply Nr. of HS-PDSCH codes 5,10 and 15 Does not apply Table 11. System specific parameters 49 Modulation scheme and coding rate: Higher order modulation schemes can improve data rate when the radio conditions are good. HSDPA supports QPSK and 16QAM. Mobile WiMAX supports QPSK, 16QAM, 64QAM in DL and QPSK, 16QAM in UL. Both technologies support both turbo coding and convolutional coding and both use adaptive modulation and coding which means that the modulation scheme and coding is changed on a per-user basis depending on signal quality and cell usage. 3.2 Network Dimensioning 3.2.1. Coverage planning To find the cell radius, r, the path loss model is rearranged and the equation is solved for r. Equation (7) shows how the COST-231 Hata model has been rearranged to solve for r: r = 10^((L+B-46.3-33.9log(f c )+13.82log(hb ))/A) (7) The only parameter that is still not known in equation (7) is L, the maximum allowed path loss. To find L we first find the minimum signal that the receivers can detect, the Rx sensitivity and the Rx signal strength. L = Rx signal strength - Rx sensitivity The Rx signal strength can be taken straight from the link budget calculations in chapter 3.1.4. In the chapter we have calculated the Rx signal strength for both the uplink and the downlink. It is the limiting link that we will have to choose for the coverage planning. Therefore it is the uplink Rx signal strength that we will use to calculate the cell radius. To find the Rx sensitivity we need to know the SINR values at the receivers. For Mobile WiMAX these values are defined in the IEEE 802.16 standards for different MCS as in Table 12 below. For the HSDPA network we have in the downlink, different data rates for different MCS levels but need to find the SINR values for each of those. To do that we use that simulation results have showed that the performance of the HS-DSCH channel is quite close to the Shannon limit of error-free data rate that can be transmitted with a specified bandwidth in the presence of noise and interference. We thus use the Shannon boundary shifted by 5 dB to find the SINR values for specific data rates. To do that we use the following equation: R=B*log2 (1+SINR/a) (8) where a=3,16 (5dB). The results can be seen in Table 12. 50 MCS WiMAX, SNR (dB) HSDPA QPSK 1/2 5 4,6 QPSK 3/4 8 7,04 16QAM 1/2 10,5 9,23 16QAM 3/4 14 12,7 64QAM 1/2 16 64QAM 3/4 20 Table 12. SINR for different MCS levels, IEEE 802.16e [24], [31] We will now find the receiver sensitivity for both technologies Rx sensitivity, HSDPA To calculate the Rx sensitivity we will use the following equation: Sin = KTB RF (dBm) + SINR(dB) - PG(dB) + NF (dB) (9) where: NF: Receiver noise figure, referenced to the antenna port. The assumed values are 4 dBfor the BS and 7dB for the MS. KTB : RF K = Boltzmann's constant = 1.381 10-23 W/Hz/K, T = 290K at room temperature and B = RF carrier bandwidth (Hz) RF So this will be -174dBm + 10*log(3.84MHz) = -108.1 dBm SINR: Signal to interference plus noise ratio, for downlink see Table 12 for the uplink the assumed value is 4.6 dB . P.G: Processing gain, 17,8 dB for uplink, 12 dB for downlink. references: [10],[32], [33], [34] For QPSK modulation and effective rate of 1/2 the SINR is 4,6 dB and by using equation (9) we get: Rx sensitivity, HSDPA = (-108,1dB) + 4,6 dB -17,8 dB + 4 dB = The HSDPA Rx sensitivity for other MCS can be seen in Table 13. 51 -117,3 dB Mobile WiMAX We use the following equation to find the Rx sensitivity: Rss = -114 +SNR RX -10log(R)+10log(F s *Nused /NFFT ) + NF(dB) (10) Where: -114 = -174 + 10*log(5Mz) NF: Receiver noise figure, referenced to the antenna port. The assumed values are 4 dBfor the BS and 7dB for the MS. R: The repetition factor (We will only consider the case of R=1) Fs: The sampling frequency in MHz (5.6 and 11.2) N Number of used carriers (421 and 841) used: N FFT size (512 and 1024) FFT: references: [35], [36], [37] For bandwidth of 5MHz, QPSK modulation and effective rate of 1/2 we get the following: Rss (QPSK 1/2) = (-114) +5dB+6,6 dB +4 dB = -98,4 dB The mobile WiMAX (5MHz and 10MHz) Rx sensitivity for the different MCS levels can be seen in Table 13 and in Table 14. QPSK 1/2 QPSK 3/4 16QAM 1/2 16QAM 3/4 -117,3 -114,9 -112,7 -109,2 Rx Sensitivity (dBm) Table 13. HSDPA Rx sensitivity for different MCS levels. [27] QPSK 1/2 Rx Sensitivity (dBm) QPSK 3/4 16QAM 1/2 16QAM 3/4 64QAM 1/2 64QAM 3/4 -98,4 -95,4 -92,9 -89,9 -87,4 -83,4 BW=5MHz Rx Sensitivity (dBm) -95,4 -92,4 -89,9 -86,4 -84,4 -80,4 BW=10MHz Table 14. Rx sensitivity for mobile WiMAX (5 and 10 MHz). [24], [38], [31]. Now that we know both the Rx signal strength and the Rx sensitivity we can calculate L, the maximum allowed path loss. Below we show how we calculate L for both technologies and outdoor only coverage: 52 HSDPA: MAPL = Limiting Rx signal strength - Rx sensitivity = 12 dB - (-117,3 dB) = 128,4 dB Mobile WiMAX (5 MHz): Mobile WiMAX (10 MHz): MAPL = 22 dB - (-98,4) = MAPL = 25 dB - (-95,4) = 120,4 dB 120,4 dB And now we can use equation (7) and the corresponding equation for Erceg to find the cell radius. The results are listed in Table 15 where the cell radius of both systems, for both Erceg and COST-231 Hata path loss models, and for both outdoor only coverage and indoor4 coverage can be found. For the indoor coverage it has been assumed that 70 % of the users are indoor and 30 % are outdoor. Technology and path loss model Cell radius, outdoor (0dB penetration loss) Cell radius, indoor (15 dB penetration loss) HSDPA, COST-231 Hata 1800m 700m HSDPA, Erceg (terrain B) 2300m 1000m WiMAX (5MHz), COST-231 Hata 900m 300m WiMAX (5MHz), Erceg (terrain B) 1300m 600m WiMAX (10MHz), COST-231 Hata 900m 300m WiMAX (10MHz), Erceg (terrain B) 1300m 600m Table 15. Cell radius, for HSDPA and Mobile WiMAX (5 MHz and 10 MHz). Coverage planning conclusions We have designed a system for fixed, nomadic and mobile users that can be situated outside or inside. The total difference between HSDPA and mobile WiMAX in the maximum allowable path loss in the uplink (which determines the cell radius) is 8 dB. This 8 dB difference and the lower carrier frequency, which has better propagation characteristics, results in a 400 m longer cell radius for the HSDPA system. The 8 dB difference in downlink path loss is because of HSDPA’s higher transmission power and because of the 17,8 dB processing gain. OFDMA offers no processing gain but instead there is a subchannelization gain of 12 dB in the case of a 5 MHz channel and 15 dB in the case of a 10 MHz channel. The processing gain is related to the data rate that the channel uses, a lower data rate gives a higher processing gain. Here we have assumed an uplink channel with data rate 64 kbps which results in 10*log(3840/64) = 17,8 dB processing gain. 64 kbps is not a very high data rate, if we had chosen for example data rate of 384 kbps the PG would have been 10 dB and then the two systems would have had approximately the same cell radius. 4 By indoor coverage it is meant that connectivity is at least provided within outer walls of buildings, wherefore a signal attenuation of approximately 15 dB is allowed for in the received signal strength. 53 3.2.2. Capacity planning When planning the capacity the downlink is the much more important link. Here we don’t consider the uplink at all, except that we use the maximum cell range found by using the uplink budget in chapter 3.2.1. It is the cell range when designing for indoor coverage that is used, that is 600m and 1000m for mobile WiMAX and HSDPA respectively. To calculate the average throughput we assume an isolated cell with evenly distributed users. We first calculate the fraction of the coverage area that will have acceptable power levels for each MCS. As a higher modulation and coding scheme demands a higher SNR and thus the maximum allowable path loss will be lower for higher MCS. To find the fraction of the cell where each MCS has acceptable power level we assume lognormal fading centered around the selected threshold level and that the cell is a circle with radius R, and y is the acceptable power level. Then P(Pr(R)>y) is the likelihood of coverage at the boundary. By integrating the probabilities over the cell we get the fraction of area where the power levels are acceptable. U = 1/2(1-erf(a) + exp((1-2ab)/b 2 ) [1-erf((1-ab)/b]) (11) where: a=(y-P,(R))/ st.dv.*sqrt(2) b = 10*path loss exp. *log(exp)/st.dev*sqrt(2) reference [39] To get the best average throughput the highest MCS is used everywhere within its range, then the next highest MCS will be used inside its range except where the highest MCS will be used and so on. This will give us the distribution of MCS levels which in turn will give us the average throughput when we have multiplied the date rate of each MCS with the portion of the cell where it is used. For HSDPA the distribution of MCS levels can be seen in Table 16 and Figure 4. QPSK, 1/2 QPSK, 3/4 16QAM, 1/2 16QAM, 3/4 Area coverage probability 99,78% 99,49% 98,97% 97,24% Distribution of MCS 0,29% 0,52% 1,74% 97,45% Table 16. Distribution of modulation and coding schemes, HSDPA. 54 Distribution of MCS, HSDPA QPSK, 1/2; 0.29% QPSK, ; 0.52% 16QAM, 1/4; 1.74% 16QAM, 3/4; 97.45% Figure 4. Distribution of MCS, HSDPA QPSK 1/2 QPSK 3/4 16QAM 1/2 16QAM 3/4 Area coverage probability 97,83% 95,34% 91,90% 84,42% 78,69% 64,58% Distribution of MCS 2,55% 3,52% 7,64% 5,86% 14,42% 66,01% 64QAM 1/2 64QAM 3/4 Table 17. Distribution of modulation and coding schemes, mobile WiMAX 10 MHz. Figure 5. Distribution of MSC for WiMAX 10 MHz. 55 QPSK QPSK 3/4 1/2 16QAM1/2 16QAM3/4 64QAM1/2 64QAM 3/4 Area coverage probability 97,83% 95,34% 91,90% 91,03% 86,90% 75,45% Distribution of MCS 2,55% 3,52% 0,89% 4,22% 11,70% 77,12% Table 18. Distribution of modulation and coding schemes, mobile WiMAX, 5 MHz. Distribution of MCS, mobile WiMAX 5 MHz QPSK, 1/2; 2.55% QPSK, ; 3.52% 16QAM, ; 0.89% 16QAM, 3/4; 4.22% 64QAM, ; 11.70% 64QAM, ; 77.12% Figure 6. Distribution of MCS for mobile WiMAX 5 MHz In Tables 19 and 20 the peak data rates for the different modulations and coding rates are given. QPSK 1/2 QPSK 3/4 16QAM 1/2 16QAM 3/4 64QAM 1/2 64QAM 3/4 3,57 5,36 7,14 10,71 14,09 16,07 Peak data rate (Mb/s)BW=5MHz Peak data rate (Mb/s) 7,14 10,71 14,28 21,42 21.42 32,14 BW=10MHz Table 19. WiMAX, peak data rates for different modulation and coding schemes. QPSK 1/2 QPSK 3/4 16QAM 1/2 16QAM 3/4 Peak data rate (Mb/s) 3,6 5,3 7,2 10,7 Table 20. HSDPA peak data rates for different modulation and coding schemes. 56 Average throughput: HSDPA: 10.7 * 0.974 + 7.2 *0.017 + 5.3*0.005 + 3.6 * 0.003 = WiMAX 5MHz: 16.07 *0.771 + 14.09*0.117 +... = 14.44 Mb/s WiMAX 10MHz: 32.14 *0.660 + 21.42 *(0.144+0.059) +14.28 *0.076+10.71*0.035+7.14*0.025 = 27.21 Mb/s 10.59 Mb/s Once the average throughput per sector and the cell range has been found, we can find the aggregate dowlink traffic that each base station can support. We have assumed three sector sites for both technologies and the base stations used are 3 sector base stations hence to find the throughput per site/base station the sector throughput is multiplied with 3. For Mobile WiMAX we first have to multiply the DL average throughput with 2/3 as the channel is used for DL 2/3 of the time and UL 1/3 of the time. Aggregate DL thoughput per site: HSDPA: 3 * 10.59 = 31.77 Mb/s Mobile WiMAX 5 Mhz: 14.44 Mb/s *2/3 * 3 = 28.88 Mb/s Mobile WiMAX 10 Mhz: 27.21 Mb/s *2/3 * 3 = 54.42 Mb/s As long as the simultaneous traffic demand per base station is less than the calculated total downlink capacity per base station the system is coverage limited. When the simultaneous traffic exceeds that capacity the system becomes capacity limited and new base stations need to be added. Capacity planning conclusions For HSDPA we see that 97.45% of the users get the highest MCS which gives the highest peak data rate, 10.7 Mb/s. This happens as there is a big difference in the downlink and uplink link budgets. The uplink maximum allowable path loss is 10.2 dB less than the downlink maximum allowable path loss. It is the uplink that is the limiting link and decides the cell radius. However as the downlink cell radius could be much bigger if we didn´t have to worry about the uplink we loose all users outside a small radiusthat would be included in the cell if the uplink limitation wouldn’t apply. Hence it turns out that only users that are so close to the base station that they get the best modulation and coding scheme are included in the cell. For the WiMAX cases the difference in uplink and downlink link budgets is not as big. Hence the distribution of modulation schemes (Table 17) is more even. But the radius is still relatively small and a high percentage of users (77% for 5 MHz and 66% for 10 MHz) get the highest bit rate. 57 4. Techno economic analysis 4.1 Tool To carry out the techno economic analysis we will use a spreadsheet based model that was created by Telenor 5 and used in the Eurescom project that this study is partly based on. The model was created originally to calculate the economical feasibility of a few different Mobile WiMAX Business Cases. This tool was chosen for several reasons. First of all it fits the purpose of this thesis perfectly as it gives us the outputs we are looking for and considers all input parameters that we are interested in. It is available for us to use and is straightforward and easy to alter to fit our needs. 4.2. Inputs to the techno-economic analysis In this chapter the input parameters of the techno economic analysis tool will be listed and the values that will be used in this paper given. When choosing those values the most important thing for the purpose of this thesis was to make it the same values for both technologies everywhere possible, as it is the comparison we are after. Therefore we chose to use mostly all the same input data as was used in the Eurescom study. The HSDPA specific information, like HSDPA equipment cost and frequency license fee was either provided by Síminn or obtained from official documents. The deployment scenario For the purpose of a simple comparison the scenario of a new entrantis the scenario chosen for both technologies. That is the operator has no previous customers and no previous network and will use mobile WiMAX/HSDPA as the only technology to compete with other operators. All customers are stand-alone customers, that is they don’t buy any other services from the operator. The following coverage rollout is assumed: First year of deployment: 30 % coverage Second year of deployment: 60 % coverage Third year of deployment: 100% coverage Population and area size This data is exactly the same for both mobile WiMAX and HSDPA. The geographic area is Reykjavík and its neighboring communities, with population of about 200.000 and area of 500 km2 . [40] 5The model was created by Arild Jacobsen, Borgar Olsen and Markku Lähteenoja at Telenor R&D 58 Number of Customers, penetration levels In Table 21 a prediction of how the mobile broadband market will develop in the next 10 years is shown. The prediction is both for the whole broadband market with all technologies included and for HSDPA and mobile WiMAX specifically. The numbers that show the number of users for HSDPA and mobile WiMAX (applies regardless of what bandwidth size has been chosen) is calculated for the city of Reykjavik with approximately 200.000 inhabitants. The prediction was made by the authors of the Eurescom study and was made for a European city with 1.600.000 inhabitants. To predict how the new entrant will succeed in getting new customers in his first 10 years it has been assumed that the saturation level of HSDPA is 50% of the total mobile broadband market and for mobile WiMAX a 40 % saturation level. WiFi will have the remaining 10 %. Years 0 1 2 3 4 5 6 7 8 9 10 Total market share of mobile broadband 10% 15% 23% 32% 42% 53% 65% 77% 86% 92% 96% HSDPA market share 0,5% 1% 2% 4% 8% 14% 23% 31% 39% 44% 47% Nr of HSDPA users 1000 2058 4187 8333 15903 28230 45000 62989 77974 88043 93871 WiMAX market share 0% 1% 2% 3% 6% 11% 18% 25% 31% 35% 38% Nr of WiMAX users 800 1664 3350 6667 12722 22584 36000 50391 62379 70435 75097 Table 21. A prediction of the mobile broadband market In our comparison we have however considered two cases of how the penetration levels will look like. In the first case we have chosen to use the same nr of users (or penetration levels) for both technologies. The reason for this is that it is quite hard to make a predictions of this kind and they will have large error margins. As this thesis does not focus on market aspects but rather technological differences this is the case used when comparing the business cases in chapter 5. The second case uses the prediction shown in Table 21 and the difference in NPV when using case 1 or case 2 can be seen in Figure 13 in chapter 5.3. 59 Installation and site cost Most of the base stations will be on new sites, or 70 % and 30 % will be co-sited with other operators’ sites. Table 22 lists all the installation and site cost. Price in Euros Installation cost for co-siting 6,000 Installation cost for a new site 50,000 Annual site rental co-siting 3,000 Annual site rental new site 6,000 CPE installation cost (operator) 250 CPE installation fee (paid by subscriber) 250 Table 22. Installation and site cost. Subscription Types In the model two types of subscriptions are defined: Stand-alone subscription (new customers or customers taken from other providers). Add-on subscription. For a new entrant all subscriptions are stand-alone as there are no previous customers. Three classes of subscriptions (Gold, Silver, Bronze) are also defined with different tariffs (monthly flat rate) and different data rates. ARPU (Euro per month) for Data rate stand-alone customers Gold 50 2 Silver 40 1 Bronze 30 0,5 (Mb/s) Table 23. ARPU (Average Revenue/User) and data rate for 3 subscription types. [4] A constant distribution of Gold (20%), Silver (60%) and Bronze (20%) customers is assumed. 60 Network dimensioning For the network dimensioning the results from chapter 3 will be used. We use the cell radius found when using the Erceg model and designing for possible indoor usage (15 dB penetration loss). HSDPA Mobile WiMAX (5MHz) Cell radius 1000m 600m 600m Cell throughput 31.77 Mb/s 28.88 Mb/s 54.42 Mb/s Mobile WiMAX (10MHz) Table 24. Cell radius and average throughput Maximum number of simultaneous users (busy hour): 60% It is assumed that at busy hour up to 60% of the users can be active at the same time. Contention rate: 5 % (1:20) It is assume that the contention rate is 1:20 or 5 %. That is up to 20 users might have to share the system bandwidth at any given time and still get the data rates they have paid for (according to Table 23). When the number of base stations needed in the network is calculated two requirements have to be fulfilled. The whole service area needs to be covered (coverage) and the users have to get the data rates they have paid for. To fulfill the first requirement we simply divide the total area by the calculated cell area. In reality this would be too much of a simplification and the base stations would need to be placed with users density, geographic in mind. Nr of BSs need for coverage: Area/cell area = 500 km 2 / *(R cell )2 To fulfill the second requirement of capacity we first find what is the total bit rate required of the system by multiplying the data rate of each subscription type with the number of users of each subscription type: Nr of users*(0.2*2 Mb/s + 0.6*1 Mb/s + 0.2*0.5 Mb/s) = Nr of users * 1.1 Mb/s It is also assumed that the maximum number of simultaneous users (busy hour) is 60 % and the contention rate is 5%. Hence the required capacity is: Nr. of users*1.1 Mb/s *0.6 *0.5/BS throughput = Nr. of users * 0.33 Mb/s/BS throughp. In Table 25 the number of BSs needed to fulfill both the coverage and capacity requirements are shown for both HSDPA and WiMAX when it has been assumed that the total number of subscribers is 40.000 or 20% of the total population for both networks. 61 HSDPA WiMAX 5 MHz WiMAX 10 MHz Nr. of BSs needed for coverage 160 443 443 Nr. Of BSs needed for capacity 42 46 25 Number of BSs required 160 443 443 Table 25. Nr. of BSs needed to fulfill coverage and capacity requirements. Cost assumptions For WiMAX all equipment cost numbers is taken from the Eurescome study. In the case of HSDPA Síminn has provided an estimation of the equipment cost. The frequency acquisition costs for HSDPA are known official numbers and from those numbers we made an estimation of what these costs could be for mobile WiMAX. According to the cost assumptions we have used the mobile WiMAX base stations are almost twice as expensive as the HSDPA base stations. This is probably because the mobile WiMAX equipment is still in a developing phase while the HSDPA equipment is more mature and has already been sold in big quantities. On the other hand the core network cost is smaller for mobile WiMAX as it is much simpler than the quite complicated architecture of an UMTS networks. Backhaul : The traffic of a cellular network needs to be backhauled from its base stations to a central point, like a RNC, MSC or a central server. This can be done in a few different ways, for example by renting a leased line from the local incumbent, laying optical fiber or using a microwave link. Here the same backhaul solution is assumed for both technologies, a microwave link. The OPEX costs include manpower costs for operations/maintenance, customer support and sales/marketing. This ranges from 5 to 12 people. It also includes advertising costs and OPEX variable cost (per customer per year). Equipment cost WiMAX BS of 3 sectors with 2nd order diversity in all sectors: 47.000 HSDPA BS of 3 sectors, 1 carrier frequency: 25.000 Core network costs WiMAX, Radius server, routers, SIP server etc. 1.000.000 HSDPA: RNCs, MSC/VLR, SGS etc.. 4.400.000 Frequency acquisition costs WiMAX Frequency license acquisition cost (10 MHz): 2.000.000 WiMAX Frequency license acquisition cost (5 MHz): 1.000.000 Annual WiMAXfreq. license fee (10 MHz): 100.000 Annual WiMAXfreq. license fee (5 MHz): 50.000 HSDPA Frequency license acquisition cost: 2.000.000 62 Annual HSDPA frequency license fee: 100.000 Annual Microwave frequency license cost: Backhaul Microwave backhaul STM1 (155Mbit/s) 100.000 16.250 OPEX OPEX variable cost 50 Salary cost including overhead 100.000 Installation cost for a new site 50.000 Site rental for a new site 6000 References: [2], [4] 4.3 Outputs from the techno-economic analysis 4.3.1 NPV, IRR and cash balance curve Outputs from the techno-economic analysis The techno-economic analysis will give results in terms the net present value (NPV), the internal rate of return (IRR) and a graph of the accumulative cash flow (a cash balance curve) for an investment project, as well as the OPEX, CAPEX and revenues for a period of 10 years. Net present value The NPV is a measure of profitability and is defined as: cash flows minus the initial investment . [41] The present value of a project’s In other words it is the difference between the discounted value of the future incomes and the amount of the initial investment. A project with a positive NPV is a profitable project. Internal rate of return The IRR is the discount rate that equates the present value of the project’s cash flows with the initial investment, i.e. the interest rate that would make the NPV of a project zero. A project that has an IRR that is greater than the required rate of return should be accepted. When comparing projects the investment project with the highest IRR should be chosen [41, page 249] The discounting rate is set to 10% and the time period for calculating Net Present Value (NPV) is set to 8 years. Payback period and cash balance curve Payback period is the time it takes to recoup an investment’s cost. That is the number of years necessary for the accumulative cash flow to equal the investment cost. 63 From or model we will get the NPV, IRR and the cash balance curve for both the mobile WiMAX and the HSDPA business cases. The NPV is given in Euros and the IRR in percentages. An example of a cash balance curve is given below. Figure 7. Example of a cash balance curve The cash balance curve shows the accumulated cash flow for each year for some period of years. The accumulated cash flow for the first year is the difference between the total revenue and the TOTEX (CAPEX + OPEX). That is, what is left of the income when everything else has been paid. The accumulated cash flow the second year is that years cash flow plus the first years cash flow and so on. The cash balance curve gives a graphical overview of the profitability. The lowest point on the graph gives the total amount of funding needed for the project. The point where the curve crosses the x-axes gives the payback period in years. Capital expenditures (CAPEX) are the amount of money spent to acquire or upgrade the physical assets of a business. In the case of a mobile broadband operator this includes the base stations, the core network equipment, backhaul equipment, frequency licenses, CPE equipment cost and site installation cost. Operational expenditures (OPEX) are the ongoing costs of running a business. This includes advertising, maintenance of the network, sales, manpower etc. 64 5. Results 5.1. NPV, IRR and cash balance curves In this chapter the results from using the spreadsheet based techno-economic analysis tool described in chapter 4.1 are shown. The input parameters have all been listed in chapter 4.2. Results Discount rate 10% Required investment capital ( ) -52 047 260 Lifetime 7 year NPV ( ) IRR -28 532 068 -8% Lifetime 10 years -4 094 340 9% Table 26.Required investment capital, NPV and IRR for Mobile WiMAX 10 MHz Results Discount rate 10% Required investment capital ( ) -50 847 260 Lifetime 7 year NPV ( ) IRR -27 238 647 -8% Lifetime 10 years -2 737 112 9% Table 27.Required investment capital, NPV and IRR for Mobile WiMAX 5 MHz Results Discount rate 10% Required investment capital ( ) -20 784 629 NPV ( ) IRR Lifetime 7 year 3 605 235 14% Lifetime 10 years 29 884 828 27% Table 28.Required investment capital, NPV and IRR for HSDPA Tables 26 to 28 show that HSDPAs NPV after 10 years is 29 884 828 while both mobile WiMAX cases have negative NPV. Moible WiMAX that uses channels of 10 MHz bandwidth have NPV of -4 094 340 and the mobile WiMAX network that uses 5 MHz channels has NPV of -2 737 112 . 65 Figure 8. Cash balance curve for mobile WiMAX 10 MHz Figure 9. Cash balance curve for mobile WiMAX 5 MHz Figure 10. Cash balance curve for HSDPA 66 Figures 8, 9 and 10 show the accumulated cash flow of our business cases. Other results HSDPA Total revenue: 156 001 677 Total CAPEX: 36 279 123 Total OPEX: 38 081 959 Total CAPEX + OPEX: 74 361 082 WiMAX 5 MHz Total revenue: 156 001 677 Total CAPEX: 63 484 821 Total OPEX: 51 822 159 TOTEX (CAPEX + OPEX): 115 306 980 WiMAX 10 MHz Total revenue: 156 001 677 Total CAPEX: 64 484 821 Total OPEX: 52 372 159 Total CAPEX + OPEX: 116 856 980 As we have here assumed exactly the same number of users for all our cases and same ARPU all cases have the same total revenue. On the other hand HSDPA has lower CAPEX and OPEX, and thus has a better cash flow. 5.2 Comparison WiMAX 5 MHz vs WiMAX 10 MHz NPV IRR TOTEX Mobile WiMAX 5 MHz Mobile WiMAX 10 MHz -2 737 112 -4 094 340 9 % 9% 115 306 980 116 856 980 Table 29.Comparison of WiMAX with 5 MHz bandwidth and 10 MHz bandwidth As we see in Table 28 the business case where 5 MHz channels are used for mobile WiMAX gives slightly better results in terms of NPV and IRR than the case where 10 MHz channels are used. But the difference is very small. The reason for this is that as the system will be coverage limited for the whole time period of 10 years the cell radius is much more important than the capacity. The channel bandwidth only affects the 67 capacity and not the coverage. The slightly better TOTEX of the 5 MHz system is because of lower frequency license costs. WiMAX vs HSDPA Mobile WiMAX 5 MHz HSDPA NPV -2 737 112 29 884 828 IRR 9 % 27% TOTEX 115 306 980 74 361 082 Table 30. Comparison of HSDPA and mobile WiMAX (5 MHz BW). In Table 29 a comparison of the two cases HSDPA and mobile WiMAX (BW 5 MHz) is given. We see that HSDPA has much better results and is thus by far the most feasible choice from an economical perspective. The much lower TOTEX is mainly because of the big difference in number of base stations needed to cover the whole area. In the case of mobile WiMAX the number of base stations needed for coverage is 443 base stations . For HSDPA the number is 160 base stations . This is an enormous difference, which results in an enormous difference in economical feasibility as well. 68 5.3 Sensitivity analysis In chapter 5.3.1 we look at how different parameters affect the outcome of our business cases. In chapter 5.3.2 we discuss how choosing the radio planning parameters and the system model differently would have affected the results. 5.3.1 Business case sensitivity The cell radius and cell throughput directly affect the economic calculations as they determine how many base stations are needed to serve the area and hence a large part of the whole infrastructure cost. Cell radius In Figure 11 it can be seen how the cell radius affects the NPVs (for 10 years lifetime) of our business cases. We use the channel throughput calculated in chapter 3 (see Table 24) , that is different channel throughput for each case, and keep it constant. Figure 11.Sensitivity on cell radius. We see in Figure 11 that the curves for HSDPA and both WiMAX cases are all very similiar. For the same cell radius the NPV of all the cases is close to being the same. The curves all pass zero on the x-axis close to 0.6 km. Which means that for the business cases to be feasible the cell radius needs to be larger than 0.6 km. A network that has cell radius of 1 km will be a much more attractive from a business perspective than a network with 0.6 km cell radius. But if the cell radius is extended from 1 km to 1.4 km or more the difference is not as drastic. 69 24. Figure 12. Sensitivity on channel throughput Figure 12 shows that the throughput has a minor effect on the NPV. Indeed, changes on the scale, from 2 Mb/s to 20 Mb/s as shown in the figure, do not have any impact at all on the WiMAX business cases. On the other hand if the channel throughput of the HSDPA network would be decreased down to 2 Mb/s from the calculated average channel throughput of 10.6 Mb/s the would NPV would be decreased by 64 %, from 29.884.828 Euros to 19.206.192 Euros. To increase the channel throughput to more than 10.6 Mb/s does not increase the NPV if nothing else is changed. The reason for that HSDPA NPV seems more sensitive to changes in channel throughput than the WiMAX NPV is that we have used the calculated cell radiuses (1km for HSDPA and 0.6 km for WiMAX ) This means that when the throughput is decreased the system with longer radius (fewer cells and more users per cell) becomes capacity limited earlier. 70 71 . .