Datasets:

eswardivi

/

books

like 0

THE EXPERT'S VOICE® IN NETWORKING 4G: Deployment Strategies and Operational Implications MANAGING CRITICAL DECISIONS IN DEPLOYMENT OF 4G/LTE NETWORKS AND THEIR EFFECTS ON NETWORK OPERATIONS AND BUSINESS Trichy Venkataraman Krishnamurthy and Rajaneesh Shetty Apress For your convenience Apress has placed some of the front matter material after the index. Please use the Bookmarks and Contents at a Glance links to access them. friendsof Apress Contents at a Glance About the Authors Introduction Chapter 1: Network Planning Chapter 2: Self-Organizing Networks in LTE Deployment Chapter 3: Deployment Challenges in Evolving 4G Chapter 4: Network Roadmaps Chapter 5: Network Roadmap Evolution Chapter 6: A Process for Network Roadmap Evolution Index Introduction This book evaluates a range of design and deployment strategies for LTE network business development, and presents a process for planning and evolving network roadmaps. Among those who will find this book useful are new field engineers who have been entrusted with the arduous tasks of deploying 4G networks. The initial chapter in the book endeavors to arm you with enough information to understand what you are doing, and why. The book also demonstrates how self-organizing networks (SON) can help improve the deployment process and help reduce the round trip time in optimizing and tuning your network. Subsequent chapters cover roadmap development and how it improves your ability to plan, build, and deploy more successful networks. From a broader perspective, this book is for all people involved or entrusted with the maintenance of 4G networks, including architects, product managers, and program managers. Senior management executives will also find the book valuable, as it give them the requisite knowledge to better ensure that relevant stakeholders are involved in the process of roadmap management and evolve strategies to ensure that their 4G networks remain operational, meaningful, and successful. We cover potential roadblocks to successful deployments, and how to avoi

d or overcome them. We also delve into roadmap management, with suggestions on how to keep them relevant using reliability engineering, organizational culture, and evolution concepts. How This Book Is Structured Chapter 1, "Network Planning," covers the nuts and bolts of deployment, and gives a speedy tour of the whole process. Chapter 2, "Self-Organizing Networks and LTE Deployment," gives a general overview of SON concepts, and helps explain how SON attempts to solve various deployment issues. Chapter 3, "Deployment Challenges in Evolving 4G," introduces readers to the challenges of LTE deployment, and highlights trends in user and traffic profiles, as well as newer trends like the Internet of Things, which need to be accounted for by LTE networks. Chapter 4, "Network Roadmaps," introduces roadmap concepts for networks and provides further coverage of factors that can affect stakeholders. Chapter 5, "Network Roadmap Evolution," focuses on how network roadmaps have to evolve and adapt to changes in technology, markets, deployments, and traffic patterns. Chapter 6, "A Process for Network Roadmaps Evolution," presents a detailed set of processes for network roadmap management and evolution. Prerequisites For the deployment-related sections of the book, readers are expected to have knowledge of LTE and radio basics, and have some practical idea of what is to be accomplished. For sections covering network roadmap management and evolution, readers should have a basic understanding of network product development. CHAPTER 1 Network Planning Network planning, especially for a cellular network, can be an extremely complex as well as time-consuming procedure. There are many steps and parameters that should be considered to ensure a well-planned radio network. This chapter gives a fast tour through the network planning and optimization of a Long Term Evolution (LTE) radio network. We also present a strong grounding through the various aspects of the LTE standard and features that you can use as a guide through the various

options for deployment. This will equip you with the knowledge to understand the choices you can make when selecting a system and need to shortlist solutions. In some cases, you may already understand the options based on decisions you have already made for network options using some other method. We start by going through some basic concepts and steps that should be followed for deployment of any radio technology. After covering these basics, we deal with the different aspects of LTE features in terms of the deployment impact in the dimensions of coverage, capacity, and performance. We then cover some advanced features intended to make LTE deployments easier. We also cover the various services offered and what types of implications these hold for the solution being deployed. We revisit the generic topics of deployment with LTE radio frequency (RF)-specific deployment inputs and discuss issues that can arise during that process. Finally, we end the chapter with inputs on the performance matrix and how the different aspects of LTE-evolved node B (eNodeB) performance can be tested. The main goal in network planning is to ensure that the planned area is covered completely. Every cellular network needs cell-site planning to ensure coverage requirements, to maximize capacity requirements, and to avoid interference. The cell-planning process consists of many different tasks, which together make it possible to achieve a well-working network. The major activities involved in the cell-planning process are represented in Figure 1-1. Broadly, the radio network planning and optimization activity can be subclassified into the following phases: Dimensioning phase Planning and implementation phase Optimization phase CHAPTER 1 NETWORK PLANNING Site acquisition Network Dimensioning Coverage planning Performance Requirement Analysis and gathering and Continuous Strategy for Network Coverage, Optimization Capacity and Quality per Parameter service planning Capacity planning Dimensioning Phase Optimization Phase Detailed Planning an

d Implementation Phase Figure 1-1. Radio network planning phases Dimensioning Phase The dimensioning phase will mainly involve information and requirement gathering from the customer from which the initial objectives for the radio network planning can be set. Some of the key inputs that are considered or required to be performed in the dimensioning are outlined in the sections that follow. Configuration for the Site As part of the configuration details, it is important to understand whether the site will be configured for a multiple- input and multiple-output (MIMO) or single-input and single-output (SISO) system. If the system is MIMO, then the transmission mode needs configuration. Also, as a part of site configuration, it is important to understand how many cells will be installed or eNodeB (i.e., sector configuration) for each site. User and Traffic Volume Estimation As a part of dimensioning, it is important to estimate the user volume and the traffic volume for each site; for example, the number of users in an urban site will be very high compared with the volume for a rural site. Similarly, the traffic volume will be higher in an area that has small offices set up in comparison with a highway deployment. The user and traffic volume estimation directly impacts the cell size that can be supported for a particular area and the capacity requirement. It also is useful for parameter settings like physical random access channel (PRACH) configuration settings, scheduler settings, and so forth. Apart from the traffic volume, it is also important to understand the traffic type that will dominate the cell SO the dimensioning can be done accordingly by calculating the net bit rate for the traffic type, 4G voice over Internet protocol traffic (VOIP), streaming traffic, hypertext transfer protocol (HTTP) traffic, and so forth. CHAPTER 1 NETWORK PLANNING Coverage and Capacity Estimation The customer should be able to provide the information on the area that is planned for service and also the quality of service offered f

or each user terminal (UE) within the service area. With this input from the customer, the cell coverage and capacity estimates are performed. Radio link budgeting is performed to understand the cell size that can be achieved with the input given from which the number of sites or cells required to plan the network area can be determined. Interface Requirement The interface requirements mainly deal with the S1 (interface between the mobility management entity [MME] and eNodeB) and X2 interface (interface between two eNodeBs) dimensioning. Based on the number of sites required (derived from the link budget activity) and the operator's allocated budget, the interfaces for each eNodeB will be dimensioned. Budget Information Budget information is very important because the number of resources (hardware) can be derived from the this, and in cases of limited budgets, the capacity or coverage planning will need to be accomplished with limited resources for a given area. Figure 1-2 presents a flow chart of the budget planning process. LTE Network Dimensioning Cell edge criteria Transmit power Frequency and Link Budget bandwidth Antenna configuration calculation Maximum permissible Path Loss UE propagation model Environment specific inputs Coverage estimation Throughput per sector Forecasted number of Maximum cell range for each site UE's per cell. Traffic distribution etc Capacity Planning Total number of cell / sites required. Figure 1-2. Network dimensioning based on budget CHAPTER 1 NETWORK PLANNING Planning and Implementation Phase the dimensioning phase, the equipment requirements are determined based on the number of cells or sites needed to provide a network to the complete area. During the planning and implementation phase, the exact location of where these eNodeBs should be placed is determined. Site selection activity is performed to accomplish the planning done in the dimensioning phase. Upon determination of the sites where the eNodeBs will need to be placed, the network planning tools (i.e., Mentum, atoll) ca

n be used to reconfirm that the capacity and the coverage planning that was performed in the dimensioning phase has been accomplished. During the planning phase, backhaul planning must also be done. In cases where the site is a colocated site, the backhaul planning should be carried out for both colocates as well as the new site. Parameter planning and setting is a major part of this phase. Some of the parameters that will impact the coverage and capacity planning are: Uplink/downlink (UL/DL) frequency Bandwidth of operation Transmission mode Transmission power Quality of service (QoS) parameters Population distribution and density Outdoor environment type (urban, rural, small office, residential, etc.) Maps and clutter details for the area Fading model type (Extended Vehicular A model [EVA], Extended Typical Urban model [ETU], Extended Pedestrian A model [EPA], etc.) Predicted traffic type and its distribution These factors will be discussed in detail later as they all impact the capacity and coverage planning. As part of the planning process, signal-to-interference-plus-noise ratio (SINR) VS. throughput mapping is performed for different propagation models (i.e., EVA, ETU, EPA, etc.) and for different transmission modes (spatial multiplexing, transmit diversity, etc.). Cell edge definition would depend on the SINR mapping. Optimization Phase Once the planning and implementation are complete, it is very common practice to run drive tests for the planned sites. Drive tests verify the predictions made by the planning tools, and the results from the drive tests are compared against the results from the simulations. Fine tuning is performed after the drive tests to ensure that the deviation in the results between simulation and drive tests is minimal. A part of the drive test, parameters like reference signal receive power (RSRP), reference signal received quality (RSRQ), or SINR the UL and DL throughputs at different points of the cell are noted. The results are then compared with the SNR predictions made by the pl

anning tool and deviations are noted and tuned wherever required. CHAPTER 1 NETWORK PLANNING Coverage Planning Coverage planning targets for the complete service area are tested to ensure there are no coverage holes (i.e., the UE never experiences a no-service condition within the entire service area). Coverage plans, however, do not take consideration any quality of service that the user experiences within a cell or site. The end aim is to provide the count of the resources or eNodeBs and cells that are required for the complete service area. Some of the most important aspects that need to be considered as a part of the coverage planning are: The eNodeB transmitting power and the type of cell that is being planned. The eNodeB transmitting power is the key for any coverage planning, and the transmitting power will vary based on the cell size. For example, a macro cell will have a transmission power of 10 watts per port (40 watts per cell in cases of MIMO cells). The DL coverage cell radius should be derived based on the transmission power of the antenna added with the gains (antenna gain, diversity gain, etc.) with the assumption of path loss (receiver loss, propagation loss, etc.). Cell radius calculation will be covered in detail in the link budget calculation section. The eNodeB receiver sensitivity. In the uplink, in order to calculate the cell radius, one of the most important parameters that the operator relies on is the receiver sensitivity of the eNodeB. The eNodeB receiver sensitivity is a deciding factor for the maximum allowed path loss between the UE and the eNodeB in the uplink direction, beyond which the eNodeB cannot differentiate accurately between signal and noise. Better receiver sensitivity of the eNodeB will directly result in a larger cell radius (coverage radius) in the uplink. The 3GPP 36.141 defines the test for deriving the reference sensitivity of a receiver. The specification also requires that a receiver sensitivity of less than 100.8 decibel milliwatts (dbm) is acceptable. However, ma

ny vendors have a receiver sensitivity value of around -102 dbm or better. UE receiver sensitivity and transmission power. Similar to the eNodeB receiver sensitivity, UE receiver sensitivity is an important factor in determining the DL cell radius for coverage planning. Typically for a macro cell, the UL cell radius will be a limiting factor in comparison with the DL cell radius simply because of the difference in the transmission powers. In LTE category 2 UE and onward, the maximum uplink transmit power is 23 db. Terrain. Terrain is an important consideration for any site planning and will impact the absorption or attenuation capability of a site. For example, a site with irregular heights will not have linear loss and is subjected to shadow areas or reflection, whereas a site with fairly regular height will have a more predictable linear loss. Similarly, the indoor to outdoor ratio of a site also makes a difference when it comes to cell radius calculation (i.e., the penetration losses for an indoor user is higher compared with that for an outdoor user); therefore, planning an urban cell will be subject to more losses due to a higher percentage of indoor to outdoor users in comparison with a rural cell. Improving Coverage for a Given Service Area Some common practices to improve the coverage for a given service area are: Receiver selection. Selecting an eNodeB with a better receiver sensitivity will help to improve the coverage for a service area. Implementing receiver diversity. In UL highers, the chances of correctly decoding the received signal from the UE improve the coverage. Beamforming. For uneven heights, beamforming can be a very handy feature to compensate for any coverage hole. CHAPTER 1 NETWORK PLANNING Improving the antenna gain. This is particularly useful for smaller cells, wherein the DL cell radius is limited in comparison with the UL cell radius. Adding more sites. In case none of these techniques can be used to compensate for a coverage hole, the last option would be to add a new site. Capacit

y Planning The capacity of an eNodeB indicates the maximum number of users that can be served by the eNodeB with a desired quality of service or the maximum cell throughput that can be achieved for a particular site at a given time. Increasing the capacity would mean increasing the number of users that can be accommodated by a cell or eNodeB, which in turn means that the number of eNodeBs or cells required to accommodate a volume of users inside a given area would be lessened, thereby reducing the cost of deployment for an operator. Capacity planning, like coverage planning, also aims at providing an estimate on the number of resources or eNodeBs required for a given service area. However, in capacity planning, the quality of service that is provided to the users within the service area is the key factor. Typically, the resource calculation from capacity planning for a given service area is higher in comparison with the resource calculations made by coverage planning. Capacity planning is initially done by using a simulation tool (e.g., Opnet, Radiodim tool, etc.), which takes in various parameters and plots an SINR graph for a UE at different distances from the transmitter. The simulations are performed to at least derive these results: Average throughput for a close-range user Average throughput for a midrange user Average throughput for a far-range user Number of UEs that can be placed inside the cell with a throughput for each UE above the acceptable levels. With these results and an estimate on the total number of users within the service area, the total number of cells that would be required can be calculated. Later, during the drive test phase, some of these tests are repeated and the SINR plotting is performed at the actual site and matched with the simulated results for accuracy. Improve Capacity for a Particular Service Area Some of the common practices to improve the capacity for a given service area are: Adding more cells. Adding more cells to the service area would mean that the number of UEs that ne

ed to be accommodated by a single cell will be reduced, therefore, the quality of service for each UE can be achieved. More sectors for a site. This again would mean adding more cells to the planned area; however, this activity involves sectorization for specific sites that provide service to a larger number of users with higher traffic. MIMO implementation. MIMO features enable capacity within a service area, and spatial multiplexing ensures that the user's throughput (in good channel conditions) is improved. The transmit diversity feature ensures the same for UEs in poor channel conditions. Also, there are advanced MIMO features like beamforming that target improvement of UE throughput, thereby enhancing the capacity of a particular cell. Increasing bandwidth. Another method to increase the capacity of a cell is to increase the bandwidth of the frequency. This method is very expensive and not very practical. CHAPTER 1 NETWORK PLANNING Radio Link Budget for LTE Radio link budgeting is where the maximum permissible path loss is calculated for a planned site. Budgeting is done in both UL as well as DL directions, and the cell radius is calculated for either capacity or coverage in both the directions and the minimal cell radius is decided upon. The link budget calculation depends on various parameters on the transmitter end or the receiver end, which contribute to the effective path loss calculation as presented in the equation: PL = Tx Power + Tx Gain + Rx Gain - Tx Loss - Rx Loss, where PL is the total path loss of the signal in decibels, Tx Power is the transmission power in decibel milliWatts, Tx Gain is the transmitter gain (antenna gain) in decibels, Rx Gain is the receiver gain (antenna gain) in decibels, Tx Loss is the transmitter loss in decibels, and Rx Loss is the receiver losses in decibels. Figure 1-3 diagrams this process. Signal transmit Transmitter power at the port antenna gain 40dbm Transmitted data Transmission loss + Propagation losses Cable and combiner losses Antenna gain at receiver end Rece

ived signal Rx losses ( cable (Noise + Interference) + combiner) Received data LTE Link Design with various Losses and Gains Figure 1-3. Process of gains and losses in transmission path Transmission Power Transmission power is the key to any link budget calculation. The higher the transmission power, the higher the permissible path loss and the greater the cell radius. Depending on the cell size, the transmission powers are of different levels, for example, a macro cell transmits at 10 to 20 watts per port, whereas for a pico cell, the power would be in the range of 2 watts. CHAPTER 1 NETWORK PLANNING Radio link budgeting is performed separately for the UL and DL as the transmission power of the signal will be of different power levels (i.e., the maximum UE transmit power is around 23 db, which is used for radio link budget calculation or acceptable path loss calculation for the UL). Features like MIMO increase the transmission power of the antenna and therefore increase the coverage and capacity of the cell. Antenna Gains Antenna gains, especially on the transmitter side (eNodeB antenna gain), are the most significant gain contributors for a link budget calculation. The reason for the gain is because of the directional behavior of the antenna (i.e., the power emitted or received by the antennas is focused in one particular direction). For a macro site, typically the antenna gain is in the order of around 18 dbm and the receiver gain on the UE side is in the order of 0 or 1 dbm. If there is no external antenna for the UE, then the gain is 0. Diversity Gain Diversity on the receiver side is useful when decoding the original signal, especially at the cell edge where the path loss is higher. The diversity capability at the receiver end helps in reducing the required energy per information to noise power spectral density (Eb/No) ratio at the receiver side. Typically, the diversity gain amounts up to 3 db both on the UE side as well as the eNodeB receiver side. Cable and Connector Losses Typically, the cable and conne

ctor losses can amount to between 2 to 3 db, depending on the quality of the cables and connectors used. Propagation Loss Propagation loss accounts for the largest variable in the link budget calculation. The propagation loss depends on a number of factors such as carrier frequency, UE distance from the transmitter, terrain and clutter, antenna height and tilt, among others. Path loss calculation is purely theoretical, and there are various propagation models that can be used to determine the path loss and, in turn, a cell radius for a particular site. Some of the popular propagation models are the Okumura-Hata model, free space model, irregular terrain model, Du Path loss model, and diffracting screens model. To calculate the path loss for a dense urban site using the Okumura-Hata model, the following formula is used: L=A+Blog10fc-13.82log10hb-a(hm)+[C-6.55log10hb].logd = - - where L is the Path loss in decibels, hB is the height of the antenna (eNodeB antenna) in meters, hm is the height of the UE antenna in meters, f is the carrier frequency in megahertz, d is the distance between the UE and eNodeB in kilometers, and A, B, and C are constants. CHAPTER 1 NETWORK PLANNING Table 1-1 is a sample RF link budget with various losses and gains on the transmitting and receiving sides. Table 1-1. Link Budget Parameters for the Transmitting and Receiving Entities UPLINK DOWNLINK TRANSMITTING ENTITY eNodeB Tx RF Output Power 23dBm 40dBM Body Loss Combiner Loss Feeder Loss 1.5dB Connector Loss Tx Antenna Gain 17.5dB 20dBm 54dBM RECEIVING ENTITY eNodeB Rx Sensitivity -104dBm -102dBm Rx Antenna Gain 17.5dB Diversity Gain Connector Loss Feeder Loss 1.5dB Interference Degradation Margin Body Loss Duplexer Loss Rx power -118dBm -96dBm Fade Margin Required Isotropic Power -114dBm -92dBm Maximum Permissible Path Loss 134dB 146dB With these parameters, a cell dimensioning is performed for 384 Kbps of data, assuming the Okumura-Hata propagation model in a dense urban area. The cell sizing is calculated as shown in Table 1-2. CHAPTE

R 1 NETWORK PLANNING Table 1-2. Cell Range Calculation for 384 Kbps Data Rate Using the Okumara-Hata Path Loss Model Allowed Propagation Loss = 146 dB in DL and 134 db in UL Carrier frequency 2300 MHz BS antenna height UE antenna height 1.5 M Parameter A value Parameter B value Parameter C value UE antenna gain function -0.00092 Pathloss exponent 3.574349 Pathloss constant 137.3351 Db Downlink range 1.496663 Km Uplink range 1.191201 Km Cell range 1.191201 Kms Site hexagon coverage area 2.911 sq kms LTE Band The Evolved Universal Terrestrial Radio Access (E-UTRA) band for frequency division duplex (FDD) and time division duplex (TDD) modes is provided in Table 1-3 as derived from 3GPP spec 36.104. It can be seen from the table that the range at which the LTE cell can operate is quite huge. Logically, every operator, if given a choice, would want to deploy their network with the E-UTRA band, which operates at a very low frequency, because the losses associated with lower frequencies are much less in comparison with higher frequency losses. This would have an impact on the cell size and, in turn, the coverage planning for an operator. CHAPTER 1 NETWORK PLANNING Table 1-3. Operating Bands for 3GPP TS 36.104 E-UTRA Operating Uplink (UL) operating band Downlink (DL) operating band Duplex Mode BS receive UE transmit BS transmit UE receive FUL_low-FUL_high FDL Iow - FDL_high 1920 MHz - 1980 MHz 2110 MHz - 2170 MHz 1850 MHz - 1910 MHz 1930 MHz - 1990 MHz 1710 MHz - 1785 MHz 1805 MHz - 1880 MHz 1710 MHz - 1755 MHz 2110 MHz - 2155 MHz 824 MHz - 849 MHz 869 MHz - 894 MHz 830 MHz - 840 MHz 875 MHz - 885 MHz 2500 MHz - 2570 MHz 2620 MHz - 2690 MHz 880 MHz - 915 MHz 925 MHz - 960 MHz 1749.9 MHz - 1784.9 MHz 1844.9 MHz - 1879.9 MHz 1710 MHz - 1770 MHz 2110 MHz - 2170 MHz 1427.9 MHz - 1447.9 MHz 1475.9 MHz - 1495.9 MHz 699 MHz - 716 MHz 729 MHz - 746 MHz 777 MHz - 787 MHz 746 MHz - 756 MHz 788 MHz - 798 MHz 758 MHz - 768 MHz Reserved Reserved Reserved Reserved 704 MHz - 716 MHz 734 MHz - 746 MHz 815 MHz - 830 MHz 860 MHz - 875 MH

z 830 MHz - 845 MHz 875 MHz - 890 MHz 832 MHz - 862 MHz 791 MHz - 821 MHz 1447.9 MHz - 1462.9 MHz 1495.9 MHz - 1510.9 MHz 2000 MHz - 2020 MHz 2180 MHz - 2200 MHz 1626.5 MHz - 1660.5 MHz 1525 MHz - 1559 MHz 1850 MHz - 1915 MHz 1930 MHz - 1995 MHz 1900 MHz - 1920 MHz 1900 MHz - 1920 MHz 2010 MHz - 2025 MHz 2010 MHz - 2025 MHz 1850 MHz - 1910 MHz 1850 MHz - 1910 MHz 1930 MHz - 1990 MHz 1930 MHz - 1990 MHz (continued) CHAPTER 1 NETWORK PLANNING Table 1-3. (continued) E-UTRA Operating Uplink (UL) operating band Downlink (DL) operating band Duplex Mode BS receive UE transmit BS transmit UE receive FULLIOW - FUL_high F DL low - FDL_high 1910 MHz - 1930 MHz 1910 MHz - 1930 MHz 2570 MHz - 2620 MHz 2570 MHz - 2620 MHz 1880 MHz - 1920 MHz 1880 MHz - 1920 MHz 2300 MHz - 2400 MHz 2300 MHz - 2400 MHz 2496 MHz - 2690 MHz 2496 MHz - 2690 MHz 3400 MHz - 3600 MHz 3400 MHz - 3600 MHz 3600 MHz - 3800 MHz 3600 MHz - 3800 MHz Note 1: Band 6 is not applicable. The bands are regulated in terms of the allowed operating bandwidth. This is driven largely by the amount of available spectrum in each of the bands. Band allocation is mainly based on the availability of the spectrum for LTE deployment. Also, the UEs will need to support these bands to be able to latch on to the network and, depending on the area of selling, the UEs are enabled for a particular set of LTE bands. For example, for North America, the bands that are reserved for deployment of LTE are bands 2, 4, 5, 7, 8, 10, 12, 13, 14, 17, 18, and 19. For China, the reserved bands are 1, 3, 34, 39, and 40. Bandwidth Options In LTE, as shown in Figure 1-4, there are multiple bandwidth options ranging from 1.4 to 20 MHz. In cases of carrier aggregation, multiple 20 MHz carriers are aggregated to obtain a higher bandwidth. Bandwidth FFT size 1.4 MHz Narrow spectrum refarming 3.0 MHz 5 MHz 10 MHz 20 MHz High data rates Figure 1-4. Multiple bandwidth options CHAPTER 1 NETWORK PLANNING Why is such a range of bandwidth required? The main target of operators using a lower bandwidth of 1.4

or 3.0 MHz is to perform spectral refarming, wherein the operator can maximize the global system for mobile communications (GSM) spectrum by refarming to LTE. Operators can refarm using a much narrower spectrum than before and deliver GSM and wideband code division multiple access (WCDMA) with less spectrum and also lower total cost of ownership. Moreover, they can deliver a vastly improved user experience and potentially attract more customers to increase revenues. Larger bandwidths of 10 MHz, 20 MHz, or more are used to provide higher data rates in a network. TDD VS. FDD This section compares TDD and FDD, and we will stick to the differences for these modes purely from an operational and implementation (deployment) point of view. However, the two modes of LTE have many more differences when compared from an architectural, designing, and testing point of view. A few differences that are seen between TDD and FDD are: In LTE TDD mode, there is no concept of a paired spectrum. This also means that in any given instance, the eNodeB or the UE will be involved only in transmission or reception of the data, but never both. For a TDD setup, the hardware design is much simpler, because at any given time, there is either transmission or reception happening, but not both. In other words, there is no need for a duplexer to have an isolated UL/DL path on the receiver/transmitter implementation for an LTE TDD device (UE/eNodeB). This also makes the equipment a little less expensive. Because there is no difference in frequency for UL and DL for LTE TDD, the channel estimation or path loss calculation in both directions is similar. This eases the link budget planning activity. This also means that the channel estimation can be more robust in LTE TDD under load conditions and fast-fading conditions wherein the eNodeB need not always rely on UE reported channel feedback for corrective actions. A large guard period is required for the eNodeB to switch from DL transmission to UL transmission. This results in a drop in efficiency an

d throughput for LTE TDD cell in comparison with an LTE FDD cell. In LTE TDD, it is possible for 3GPP to allow different configurations that have a different mix of UL and DL subframes. Based on the traffic volume, the TDD configuration mode can be selected (i.e., for sites where higher usage of UL data is predicted, TDD config 0 can be configured whereas for sites where higher DL data are predicted, TDD mode 2 or 5 can be used). TDD config mode not only depends on the amount of UL and DL data but it also depends on the nature of traffic that is being used and the block error rate (BLER) history of the site. For example, if for a site the majority of the data are display sensitive, then a faster switching time (5 ms) is required, whereas if throughput and spectral efficiency are the criteria, then a switching time of 10 ms will be good. Similarly, for an area that is subject to very few retransmissions and higher DL data, TDD mode 5 would be ideal, whereas if the error rate is higher, then TDD mode 0, 1, or 2 would be preferred. This section will explain what MIMO is and the different transmission modes and advantages of each mode. The basic intent of this section is to explain the implication of MIMO on radio network planning and how or which MIMO settings can help for different deployment types. CHAPTER 1 NETWORK PLANNING Transmit Diversity Mode In transmit diversity mode, each antenna transmits the same stream of data. At the receiver end, because there are multiple streams being received by multiple receivers, and the probability of reconstruction of the data is much higher, thereby improving the signal to noise ratio. The transmit diversity mode of MIMO, if implemented in an area, will improve the coverage area by around a 3-db margin. The transmit diversity mode is useful for cells that are planned for rural areas where the cell size is typically large and users are typically spread across the cell. Closed Loop Spatial Multiplexing Closed loop spatial multiplexing is useful when the user's throughput or cel

l capacity needs to be improved in general. Because the closed loop spatial multiplexing mode of MIMO works on the feedback mechanism (Precoding Matrix Index [PMI] feedback) provided by the UE to the network, it is important that the UE is not fast moving and is either stationary or very slow moving for best results. The urban small office model and dense urban model are two main deployment types where the cell can be configured for transmission mode 4 (TM4) (closed loop spatial multiplexing). Open Loop Spatial Multiplexing Open loop spatial multiplexing, like closed loop spatial multiplexing, is also targeted to improve the cell or sector throughput for a particular area of deployment. However, open loop spatial multiplexing does not rely on the PMI reporting from the UE but works on a predefined set of precoding selection for spatial multiplexing. More often, the cyclic delay diversity (CDD) technique is used for open loop spatial multiplexing. In CDD, the transmitting unit adds cyclic time shifts and creates multipath transmission. The eNodeB tries to ensure that the transmission happens on the resource blocks for which the UE has reported better channel quality indicator (CQI) value. By doing this, the UE is able to receive the original stream as well as the delayed stream of data, and the delay on the transmit side ensures that there is no signal cancellation on the receiver side. This is more useful for areas where the UE moves at a higher speed or the channel conditions change faster, for example, TM3 (open loop spatial multiplexing) can be set for cells that are modeled for highway deployment, which has many fast moving users. Beamforming Beamforming is a MIMO technique wherein the eNodeB transmitter tries to improve the quality of the signal that is received by specific users. This can be done by adjusting the tilt and power of the transmitter in the direction of the UE. Implementing beamforming can be very complicated, wherein the UE positioning has to be determined and the UE specific reference signals

have to be configured for a cell. Also, the antenna calibration and maintaining the timing between the antennas will be quite challenging. In practice, beamforming is useful for places where the cell geography is such that some users are in shadow areas and the only way to provide them with sufficient coverage is by beamforming. UE Capabilities Apart from the other factors discussed previously that can impact the LTE radio network planning and optimization, UE capability can also significantly influence the process of cell planning. The cell throughput will depend on the average UE category within the area (i.e., if a site has higher distribution of category 4+ UEs, then the spectral efficiency for that cell will be higher). Tables 1-4 and 1-5 are derived from 3GPP spec 36.306 and give an idea of the supported throughput for each category of UE in DL and UL. CHAPTER 1 NETWORK PLANNING Table 1-4. UE Category vs. Downlink Throughput Support UE Category Maximum Maximum number of bits Total number Maximum number number of of a DL-SCH transport block of soft of supported layers DL-SCH transport received within a TTI channel bits for spatial block bits received multiplexing in DL within a TTI Category 1 10296 10296 250368 Category 2 51024 51024 1237248 Category 3 102048 75376 1237248 Category 4 150752 75376 1827072 Category 5 299552 149776 3667200 Category 6 301504 149776 (4 layers) 75376 (2 layers) 3654144 2 or 4 Category 7 301504 149776 (4 layers) 75376 (2 layers) 3654144 2 or 4 Category 8 2998560 299856 35982720 Table 1-5. UE Category vs. Uplink Throughput Support UE Category Maximum number of UL-SCH Maximum number of bits Support for transport block bits transmitted of an UL-SCH transport block 64QAM in UL within a TTI transmitted within a TTI Category 1 Category 2 25456 25456 Category 3 51024 51024 Category 4 51024 51024 Category 5 75376 75376 Category 6 51024 51024 Category 7 102048 51024 Category 8 1497760 149776 The UL MIMO capability of the UE will also impact the link budget calculation in the UL direction,

as will the 4x4 MIMO support in the DL. However, in order for the planning to consider these inputs, it is important that a larger percentage of UEs in the area of deployment support these features. CHAPTER 1 NETWORK PLANNING Cell Sizes: Femto VS. Micro VS. Macro Table 1-6 provides a brief estimate of the cell types, their usage, range, and the transmission power level that is typically used. Table 1-6. Different Cell Types and Use Cases Cell Type Cell Range Transmission Power and Other Characteristics Macro cell 1 km to 100 kms Transmission power = 10-20 watt port. Usually an outdoor deployment (e.g., rural deployment, dense urban deployment, etc.) Micro cell 0.5 km to 1 km Typical transmission power = 4 watt port. Usually an outdoor deployment (e.g., small office deployment, stadium deployment, etc.) Femto cell ~500 m Typical transmission power = <2 watt port. Ideal for indoor deployment. LTE Performance Testing Performance testing for LTE eNodeBs can broadly be classified into four test areas: Key performance indicator (KPI) verification Traffic model-based testing Overload testing Long-duration testing To perform these tests, you must simulate traffic and generate load conditions with multiple UEs and simulate failures. Also, it is not possible to test all these cases as part of drive tests or field tests because some of these scenarios are not easy to re-create. There are many tools available in the market that specifically target performance and load testing, such as Azimuth, TM500 (Aeroflex), ERCOM, and JDSU. These tools are able to simulate multiple UEs performing different actions at the same time, and it is also possible to distribute the UEs at different distances from the cell center and simulate different fading models for these UEs (e.g., EPA, EVA, ETU, etc.). It would be ideal to perform these testings with a performance test tool and then verify a subset of these tests again as part of a drive test or field test and match the results, SO there will not be much difference between the lab results an

d the field results. CHAPTER 1 NETWORK PLANNING Key Performance Indicator Verification The KPIs are very important aspects for any network element because they determine the need and the nature of optimization that will be required. The 3GPP has standardized the areas for KPI validation in TS 32.425 as: Accessibility of KPI testing Retainability of KPI testing Integrity of KPI testing Availability of KPI testing Mobility of KPI testing Accessibility of KPI Testing Accessibility KPIs mainly determine how easy it is for the user to obtain service within specified tolerances and other given conditions. The radio resource control (RRC) establishment success rate is a common KPI in this category. Other examples include paging congestion rate, RRC reestablishment success rate, RRC reconfiguration success rate, initial E-UTRAN radio access bearer (E-RAB) setup success rate, additional E-RAB setup success rate, among others. In order to test accessibility KPI cases (i.e., the RRC establishment success rate), these considerations are required: The environment should consist of multiple UEs attempting RRC connections to move from the RRC_IDLE to the RRC_CONNECTED state. The UEs should attempt RRC connection setup at a higher rate of around 10 to 20 RRC connection setups per sector per second. In order to maintain a constant number of connected users per sector, it is also required to ensure a steady rate of users moving from the RRC_CONNECTED to the RRC_IDLE state per sector. At the end of the granularity period (which is known), the RRC establishment success rate KPI is derived using the following equation: RRCConnect ion Success RRCS SR = RRCConnectionAttempt The tests can be repeated for different rates of RRC connection setups per second and for different load conditions (e.g., 30% load, 50% load, 70% load, 90% load, etc.). Retainability of KPI Testing Retainability KPIs target evaluation of how easy it is for a user to retain an established service within specified tolerances and other given conditions. Examples for r

etainability KPIs are RRC abnormal release rate, E-RAB abnormal release rate, E-RAB release success rate, UE context release success rate, and average E-RAB number per active user. CHAPTER 1 NETWORK PLANNING In order to test retainability KPI cases, for instance, the UE context release success rate, these considerations will be required: Multiple UEs needs to initiate the attach procedure followed by a constant UL/DL data transfer procedure. For a fraction of these RRC_CONNECTED UEs, eNodeB should trigger the UE context release procedure due to user inactivity. For another fraction of the RRC_CONNECTED UEs, MME should initiate UE context release for various reasons (i.e., successful handover completion, handover cancellation completion, release of old UE-associated S1 connection, etc.). In order to maintain a steady number of attached users or sectors, new attached procedures must be maintained within the cell at the same rate at which UE context release requests or completions are maintained. Integrity of KPI Testing Examples for integrity KPIs are UL peak user throughput, DL peak user throughput, DL Internet provider (IP) latency, DL transport BLER, UL transport BLER, roundtrip time (RTT) latency, RTT packet loss (ping), among others. In order to test integrity KPI cases, for instance, DL peak user throughput, these considerations will be required: A single UE should perform the attachment and should have at least one digital radio broadcasting (DRB) for non-guaranteed bit rate (GBR) data and one DRB for GBR data in the DL as well as UL direction for throughput tests. The aggregate maximum bit rate (AMBR) and the GBR values of the UE RABs should be sufficiently high (equal to or more than the cell throughput) and the application server that is connected to the evolved packet core (EPC) should be able to pump data for these RABs with a steady flow, wherein there is sufficient data scheduled for both of these RABs. The UE reported CQI should be maintained very high (around 15) to measure the peak throughput in DL

for the user under ideal conditions. UE can be moved to the cell center, cell edge, and SO forth, and throughputs can be measured accordingly for each of these conditions. The steps can be repeated for different propagation models (pedestrian fading, vehicular speed, etc.). Availability of KPI Testing For testing availability of the eNodeB, various tests can be run on the eNodeB continuously and the average downtime of eNodeB should be noted using the following equation: Total testing time - eNodeB down time = eNodeB available time. No special testing is specified to verify the availability KPI, instead the eNodeB downtime should be noted while performing all other performance testing and the KPI value should be derived. CHAPTER 1 NETWORK PLANNING Mobility of KPI Testing Mobility KPI testing targets to verify the system performance during various handovers. Examples for mobility KPIs are intra-eNodeB handover success rate, intrafrequency handover success rate, interfrequency handover success rate, X2 handover success rate, S1 handover success rate, among others. In most deployments, the handover is triggered based on the A3 event reported by the UE incase of intra- or interfrequency handover and B1 or B2 event reported by UE in case of interradio technology transfer (RAT) handover. Events A3 and B1 are most often used to refer to a condition where the neighbor cell signal strength measurement is offset better than the serving cell signal strength. In order to test mobility KPI cases, for instance, the intra-eNodeB handover success rate, these considerations will be required: Multiple UEs for multiple sectors are required to perform UE's attach procedure followed by a constant UL/DL data transfer procedure for each of these UEs. UE mobility should be enabled with different speeds so that the event A3/B1 is triggered for the UEs and handovers to the neighboring sectors are initiated. For a given sector under test, it is required to maintain a steady rate of outgoing handovers and an equal rate of incoming handover

and observe the success ratio over a period of time. A3 and B1 parameters should be set to a different combination and consistency in KPI and should be observed. KPI should be monitored to be within an acceptable range. Tests can be repeated with inter-eNodeB over S1, X2, and interfrequency as well. The KPIs should be monitored for different load conditions of the cell or eNodeB and will be repeated for different traffic profiles as well as different UE channel condition. Table 1-7 provides a list of all KPIs for each category with targets for lab tests for most of these KPIs. Please note that these targets are assumptions and are based on some customer inputs and not necessarily a benchmark for pass/ fail criteria for any of these tests. For integrity KPI cases, the target values will be different for TDD and FDD modes, and there can also be quite a bit of difference in lab test results VS. field results for these KPIs. CHAPTER 1 NETWORK PLANNING Table 1-7. KPIs for Each Category KPI Category Lab Test Target Accessibility KPIs Attach success rate Detach success rate RRC establishment success rate RRC reconfiguration success rate Initial E-RAB setup success rate Additional E-RAB setup success rate E-RAB setup success rate E-RAB modify success rate E-RAB blocking rate UE context establishment success rate UE context modification success rate S1-signal connection establishment success rate Initial E-RAB accessibility Additional E-RAB accessibility Security mode success rate Attach rate (attaches/second) Retainability KPIs RRC abnormal release rate E-RAB abnormal release rate E-RAB release success rate UE context release success rate Average E-RAB number per active user Integrity KPIs Single UE downlink IP peak throughput Single UE uplink IP peak throughput Single UE downlink IP average throughput Single UE uplink IP average throughput Overall downlink IP peak throughput Overall uplink IP peak throughput Overall downlink IP average throughput Overall uplink IP average throughput Single UE cell edge DL IP peak throu

ghput Single UE cell edge UL IP peak throughput Single UE cell edge DL IP average throughput Single UE cell edge UL IP average throughput (continued) CHAPTER 1 NETWORK PLANNING Table 1-7. (continued) KPI Category Lab Test Target Overall cell edge DL IP peak throughput Overall cell edge UL IP peak throughput Overall cell edge DL IP average throughput Overall cell edge UL IP average throughput End-end latency eNodeB latency State transition latency: Idle to Active State transition latency: Sleep to Active Paging latency Downlink transport BLER Uplink transport BLER Availability KPIs Cell availability 5-7 weeks Mobility KPIs Success rate of intra-eNodeB outgoing handovers Success rate of S1 inter-eNodeB outgoing handovers Success rate of X2 inter-eNodeB outgoing handovers Overall success rate of inter-eNodeB outgoing handovers Preparation ratio of inter-eNodeB outgoing handovers Success rates of outgoing handovers per cause Outgoing handover failure rate Success rate of intrafrequency outgoing handovers Success rate of interfrequency outgoing handovers with gap-assisted measurements Success rate of interfrequency outgoing handovers with non-gap-assisted 95% measurements Success rate of outgoing handovers with DRX Success rate of outgoing handovers without discontinuous reception (DRX) Success rate of E-RAB establishment for incoming handovers Outgoing handover cancellation rate Inter-RAT mobility CHAPTER 1 NETWORK PLANNING Note 1: Details for each of these KPIs can be obtained by 3GPP spec TS32.425. Note 2: The Lab test target for the integrity KPI can vary depending on the UE category used and the System configuration (2x2 MIMO, 4x4 MIMO etc) for e.g the peak DL throughput for a 4x2 MIMO FDD system should be greater than 140Mbps Traffic Model Testing As a part of traffic model-based performance testing, testing for different traffic models for different durations and KPIs should be observed for any variation or drop. Some of the common traffic models used or simulated are: Dense urban traffic model Urban small offi

at 3 km/hour 20% of the speed for this traffic model. users are stationary (continued) CHAPTER 1 NETWORK PLANNING Table 1-8. (continued) Parameter Values Comment Average number of sessions/ Based on data usage, traffic mix distribution from UE/busy hours (BH) Sandvine, and application characteristics such as web page size, video duration. Low mobility users consume 50% more and high mobility users consume 50% less than the medium mobility users. Number of E-RAB addition/ Based on the number of voice calls during the BH UE/BH that would require one dedicated bearer setup. Low mobility users also make more calls than higher mobility users. Number of E-RAB deletion/ Same assumption as above to remove the dedicated UE/BH bearer. Average session duration (sec) 300 sec Based on the traffic mix and session duration per service type (e.g., streaming, browsing), assuming 25% longer session for low mobility user compared with medium mobility users. The difference could be viewed as low mobility users having a different traffic mix, which is heavier on video streaming. A similar assumption is made for the high mobility user in relation to the medium mobility user. Number of attaches/minute Number of detaches/minute Data bandwidth (BW) 8MB/user (including all Based on 1GB monthly consumption, 30 days per consumption the RABs) month, 5 BH per day, and 80% BW consumed during the BH. Low mobility users consume 50% more and high mobility users consume 50% less than the medium mobility users. Number of tracking area Based on a periodic TAU of 1 hour or more updates (TAU) considering that there will be 15 UEs in the network and the simulation will be for a period of 30 minutes, we can assume 75 TAUs for this traffic model. Number of RRC Based on 1% Radio Link Failure (RLF) probability for reestablishments medium mobility user and only connected users. Data generation Full buffer For simplicity we can assume full buffer transmission for all the RABs Indoor to outdoor ratio There are different urban residential and urban small offic

e and urban shopping mall models where the indoor to outdoor ratio is higher; in this model we assume that the traffic is mainly outdoor. DL node B Transmitter- MIMO is assumed for this traffic model. Receiver (Tx-Rx) scheme (continued) CHAPTER 1 NETWORK PLANNING Table 1-8. (continued) Parameter Values Comment Simulation time 30 minutes RSRP quality distribution Ratio of 30:30:40 The RSRP quality of distribution can be such that 30% of the users are experiencing excellent quality of signal, 30% of users are experiencing good quality of signal, and 40% of users are experiencing poor quality of signal. The reason for higher poor quality of signal is because the cell size for urban dense simulation will be smaller and many of the users will be toward the cell edge because they will be initiating a handover. Number of incoming This can further be divided into the type of handover handovers (S1/X2 handover). Number of outgoing handovers This can further be divided into the type of handover (S1/X2 handover). Number of data sessions/ subscriber Urban Small Office Model The main difference between the urban small office model and dense urban model will be the user distribution and the mobility of the users. In a small office model, the majority of the users will be stationary, and at the cell center, there will be a lesser number of handovers during busy hours. The usage of traffic will be higher, but the number of users will be lower for this model compared with that of the dense urban model. During the last 5 minutes of the simulation, you will need to simulate an inverse situation to that of the first 25 minutes wherein many cell center users will move toward the cell edge and handover to other cells and the traffic distribution will inverse from cell centric to cell edge and outward mobility. Table 1-9 presents a list of parameters and the values for the urban small office traffic model. CHAPTER 1 NETWORK PLANNING Table 1-9. Parameters for the Urban Small Office Traffic Model Parameter Values Comment Number of UEs Fo

r urban small office simulation, the number of UEs at any given time will be moderate and the system load for this kind of a setup is assumed to be around 85%. User distribution Concentrated at cell The user distribution will be dense in the cell center and center and scattered and scattered or uneven toward the cell edge. However, toward very low density toward the last 5 minutes of simulation, the user distribution the cell edge. will be opposite wherein the cell center users who were stationary earlier now become mobile and move toward the cell edge and handover to the neighboring cell. Also the throughput for the cell will drop during the last 5 minutes of the simulation. Terminal speed 80% of the users are stationary 10% of the users are moving at EPA (3 km/hr speed). 10% of the users are moving at around 30 kmph (vehicular speed). Average number of The assumption is that average duration of a session and sessions/UE/busy hour the average number of sessions/users are both higher in a small office model. Number of E-RAB Based on the number of voice calls during the BH that addition/UE/BH would require one dedicated bearer setup. Low mobility users also make more calls than higher mobility users. Number of E-RAB Same assumption as above to remove the dedicated bearer. deletion/UE/BH Average session duration 600 sec The assumption is that the average duration of a session (sec) and the average number of sessions/users are both higher in a small office model. Number of attaches/ minute Number of detaches/ minute Data bandwidth 15 MB/user (including all Considering that the number of users who are stationary consumption the RABs) is around 80% and there is not much inward/outward mobility for the first 25 minutes of simulation, the average data consumption of a user will be higher. (continued) CHAPTER 1 NETWORK PLANNING Table 1-9. (continued) Parameter Values Comment Number of TAUs Based on a periodic TAU of 1 hour or more considering that there will be 15 UEs in the network and the simulation will be for a perio

d of 30 minutes, we can assume 75 TAUs for this traffic model. Number of RRC Based on 1% radio link failure (RLF) probability for reestablishments medium mobility user and only connected users would experience RLF. Data generation Full buffer For simplicity, we can assume full buffer transmission for all the RABs Indoor to outdoor ratio Considering that the area is that of an urban small office, we can assume that there are high numbers of indoor users compared with outdoor users. DL node B Transmitter- MIMO is assumed for this traffic model. Receiver (Tx-Rx) scheme Simulation time 30 minutes RSRP quality distribution Ratio of 70:20:10 The RSRP quality of distribution can be such that 70% of the users are experiencing excellent quality of signal, 20% of users are experiencing good quality of signal, and 10% of users are experiencing poor quality of signal. The reason for higher good quality of signal being most of the users will be stationary for this model and at cell center (if rightly planned). Number of incoming This can further be divided into the type of handover handovers (S1/X2 handover). Number of outgoing This can further be divided into the type of handover handovers (S1/X2 handover). Toward the last 5 minutes of simulation, the number will be higher as many of the stationary users will be mobile and moving outward. Number of data sessions/ subscriber Urban Residential Area Model The urban residential model will be more or less similar to the small office model wherein most of the users stationary and the volume of traffic used by users will be on the higher side. However, the main differences between the small office model and urban residential model will be: Users will be more uniformly distributed in the residential model and will not be concentrated at some areas and scattered over the rest of area. The change in traffic conditions during nonpeak hours will not be as drastic as in small office case; the number of handovers and mobility of users will not be very high. CHAPTER 1 NETWORK PLANNING Towa

rd the last 10 minutes of the simulation period, a simulation will be triggered wherein the number of users will increase by around 30% and these users will be moving at a vehicular speed (30 0 kmph). The number of outgoing handovers will increase by around 30% during the first half of this period (5 minutes), and the number of incoming handovers will increase by 30% during the second half of this simulation (5 minutes). Table 1-10 presents a list of parameters and the values for the urban residential traffic model. Table 1-10. Parameters for the Urban Residential Traffic Model Parameter Values Comment Number of UEs 50 for the first 20 minutes and 65 For urban residential simulation, the during the last 10 minutes number of UEs at any given time will be moderate and the system load for this kind of a setup is assumed to be around 70%. During the last 10 minutes, we assume that there will be 30% more users involved in outward mobility for the first 5 minutes and inward mobility toward the last 5 minutes, and the load in the network will vary accordingly. User distribution Fairly uniform The user distribution will be uniform for a residential traffic model. However, in the last 10 minutes of simulation, the user distribution will involve 30% of users moving from cell center to cell edge in the first 5 minutes of the simulation and 30% of users moving from cell edge to cell center toward the last 5 minutes. Terminal speed 70% of the users are stationary 10% of the users are moving at EPA (3 km/hr speed). 20% of the users are moving at around 30 kmph (vehicular speed). Average number of sessions/ The assumption is that the average UE/busy hour (BH) duration of a session in a residential area will be higher and the average number of sessions/user will be lower. Number of E-RAB addition/ Based on the number of voice calls UE/BH during the BH, which would require one dedicated bearer setup. Low mobility users also make more calls than higher mobility users. Number of E-RAB deletion Same assumption as above to remove the

UE/BH dedicated bearer. (continued) CHAPTER 1 NETWORK PLANNING Table 1-10. (continued) Parameter Values Comment Average session duration (sec) 600 sec The assumption is that the average duration of a session in a residential area will be higher and the average number of sessions/user will be lower. Number of attaches/minute Number of detaches/minute Data bandwidth consumption 12MB/user (including all the RABs) Considering that the percentage of users who are stationary is around 70% and there is not much inward/outward mobility for the first 20 minutes of simulation, the average data consumption of a user will be higher. Number of TAUs Based on a periodic TAU of 1 hour or more considering that there will be 15 UEs in the network and the simulation will be for a period of 30 minutes, we can assume 75 TAUs for this traffic model. Number of RRC Based on 1% RLF probability for medium reestablishments mobility user and only connected users would experience RLF. Data generation Full buffer For simplicity we can assume full buffer transmission for all the RABs Indoor to outdoor ratio 03:01 Considering residential area, the majority if users will be indoors. DL node B Transmitter- MIMO is assumed for this traffic model. Receiver (Tx-Rx) scheme Simulation time 30 minutes RSRP quality distribution Ratio of 50:30:20 The RSRP quality of distribution can be such that 50% of the users are experiencing excellent quality of signal, 30% of users are experiencing good quality of signal, and 20% of users are experiencing poor quality of signal. The reason for higher good quality of signal being most of the users will be stationary for this model and at cell center (if rightly planned). (continued) CHAPTER 1 NETWORK PLANNING Table 1-10. (continued) Parameter Values Comment Number of incoming This can further be divided into the type of handovers handover (S1/X2 handover). Toward the last 10 minutes of simulation, the number will be higher as many of the stationary users will be mobile and moving outward or inward. Number of outgoin

g handovers 20 This can further be divided into the type of handover (S1/X2 handover). Toward the last 10 minutes of simulation, the number will be higher as many of the stationary users will be mobile and moving outward or inward. Number of data sessions/ subscriber Highway Model Simulation of a highway traffic model will require these considerations: The cell size should be considerably large. The average speed of a user will be high (around 70 to 100 kmph). The number of users will be lesser and the mobility of the users will be very high, with around 90% of the users involved in inward as well as outward mobility. It is possible that the cyclic prefix for the cells modeled around highway are of extended types as the cells are normally of larger size. User distribution is fairly uniform as the movement will be a particular direction on the highway. Table 1-11 presents a list of parameters and the values for the highway traffic model. CHAPTER 1 NETWORK PLANNING Table 1-11. Parameters for the Highway Traffic Model Parameter Values Comment Number of UEs For a highway model, the average number of UEs at a given time should be approximately 20 and the total throughput usage should be around 40% to 50%. User distribution Fairly uniform The user distribution for a highway model should be fairly uniform as the users will be moving along a specific path. Terminal speed 80% of the users are fast moving at a speed between 70 to 100 km per hour. 10% of the users are stationary 10% of users are slow moving at a speed of 3 kmph. Average number of session/ The assumption is that the average UE/busy hour (BH) duration of a session in highway area will be lower and the average number of sessions per user will be higher because of mobility and higher RLF. Number of E-RAB addition/ Based on the number of voice calls UE/BH during the BH, which would require one dedicated bearer setup. Low mobility users also make more calls than higher mobility users. Number of E-RAB Deletion/ Same assumption as above to remove UE/BH the dedicate

d bearer. Average session duration (sec) 180 sec The assumption is that the average duration of a session in highway area will be lower and the average number of sessions per user will be higher because of mobility and higher RLF. Number of attaches/minute Higher number of attaches/detaches due to the mobility of users. Number of detaches/minute Higher number of attaches/detaches due to the mobility of users. bandwidth consumption 4MB/user (including all the RABs) Considering that the users are on high mobility, the channel conditions will not allow for higher data rate for these users and hence the data bandwidth consumption will be lower. (continued) CHAPTER 1 NETWORK PLANNING Table 1-11. (continued) Parameter Values Comment Number of TAUs Based on a periodic TAU of 1 hour or more considering that there will be 15 UEs in the network and the simulation will be for a period of 30 minutes, we can assume 75 TAUs for this traffic model. Number of RRC Based on 1% RLF probability for reestablishments medium mobility user and only connected users would experience Data generation Full buffer For simplicity we can assume full buffer transmission for all the RABs Indoor to outdoor ratio Considering the traffic model is that of a highway, there must be a negligible number of indoor users in comparison with outdoor users. DL node B Transmitter- MIMO is assumed for this traffic Receiver (Tx-Rx) scheme model. Simulation time 30 minutes RSRP quality distribution Ratio of 20:20:60 The RSRP quality distribution will largely depend on the mobility of the users in this model; considering that 20% of users are stationary, we can assume around 20% of users to be in excellent RSRP conditions. Further assumption here is that at any given time there will be 20% of users in a good radio condition zone assuming that there are another 20% of users who are under slow- moving conditions. Remaining 60% of users will be under poor conditions assuming that they are moving fast. Number of incoming This can further be divided into the handovers

type of handover (S1/X2 handover). Number of outgoing This can further be divided into the handovers type of handover (S1/X2 handover). Number of data sessions/ subscriber CHAPTER 1 NETWORK PLANNING Rural Large Cell Model Simulation of a rural large cell traffic model will require the following considerations: The cell size should be considerably large with extended cyclic prefix due to the large cell (if possible). Number of users will be less and the mobility of the users will not be high. User distribution is uneven, with more users concentrated in a few places within the cell and no users in many other parts. Very low density of users and a higher number of outdoor users compared with indoor users, and because the cells are larger in size, the mobility ratio is low. Table 1-12 presents a list of parameters and the values for the rural large cell model. Table 1-12. Parameter for the Rural Large Cell Model Parameter Values Comment Number of UEs For a rural model, the average number of UEs at a given time should be approximately 40, and the total throughput usage should be around 40% to 50%. User distribution Unevenly distributed with users The user distribution for a rural model concentrated in a few places should be random with higher number of and no users in other places. users in some areas and no users or low users in some other areas. Terminal speed 80% of the users are pedestrian model moving at 3 kmph speed. 10% of the users are stationary 10% of users are fast moving at a speed of around 70 kmph. Average number of sessions/UE/ The assumption is that the average duration busy hour (BH) of a session in rural area will be lower and the average number of sessions/user will also be lower. Number of E-RAB addition/UE/ Based on the number of voice calls during the BH that would require one dedicated bearer setup. Low mobility users also make more calls than higher mobility users. Number of E-RAB deletion/UE/BH 2 Same assumption as above to remove the dedicated bearer. Average session duration (sec) 180 sec Th

e assumption is that the average duration of a session in rural area will be lower and the average number of sessions/user will also be lower. (continued) CHAPTER 1 NETWORK PLANNING Table 1-12. (continued) Parameter Values Comment Number of attaches/minute Number of detaches/minute Data bandwidth consumption 4MB/user (including all the Considering that the cell size is very high, the RABs) throughput consumption per user should be lower. Number of TAUs Based on a periodic TAU of 1 hour or more, considering that there will be 15 UEs in the network and the simulation will be for a period of 30 minutes, we can assume 75 TAUs for this traffic model. Number of RRC reestablishments Based on 1% RLF probability for medium mobility user and only connected users would experience RLF. Data generation Full buffer For simplicity we can assume full buffer transmission for all the RABs. Indoor to outdoor ratio Considering that the area is that of a rural large cell, we can assume that there are high numbers of outdoor users compared with indoor users. DL node B Transmitter-Receiver MIMO is assumed for this traffic model. (Tx-Rx) scheme Simulation time 30 minutes RSRP quality distribution Ratio of 50:30:20 Because the cell is larger in size, most of the users should be in the excellent to good reception area compared with the cell edge region. Also because the traffic profile is that of a rural area, the number of obstructions in the path that can result in shadowing or fading are lower in number. Number of incoming handovers This can further be divided into the type of handover (S1/X2 handover). Number of outgoing handovers This can further be divided into the type of handover (S1/X2 handover) Number of data sessions/ subscriber For all the traffic models used for simulation, the user data distribution was assumed to be those presented in Table 1-13. CHAPTER 1 NETWORK PLANNING Table 1-13. User Data Distribution by Traffic Model Type of Traffic Total Traffic (%) Streaming Browsing Social Virtual private network (VPN) Marketplace

Others including voice over Internet protocol (VoIP) Total UE simulation will be triggered using a performance simulation tool and channel conditions along with UE speed. Distribution can also be done using either the performance simulation tool or a channel emulator tool. The eNodeB will be connected to the element management system (EMS) where the KPIs can be observed over the course of testing. Overload and Capacity Testing Overload and capacity testing can broadly be classified into two categories: Control plane overload and capacity testing. User plane overload and capacity testing. Control Plane Overload and Capacity Testing Control plane capacity and overload testing deal mainly with determining the signaling capacity of the eNodeB and estimating the signaling load. Control plane overload and capacity testing will involve tests like: Maximum number of RRC connected UEs that can be supported by a sector without compromising the KPIs. Maximum number of RRC connected UEs that can be supported by an eNodeB without compromising on the KPIs. Maximum number of E-RABs (default plus dedicated) that can be supported by a sector without compromising the KPIs. Maximum number of E-RABs (default plus dedicated) that can be supported by an eNodeB without compromising the KPIs. Number of simultaneous attaches procedures (number of attach requests per second) that can be supported by the sector without compromising the KPI requirements. Number of incoming handovers that can be supported by the sector without compromising the KPIs. CHAPTER 1 NETWORK PLANNING For control plane capacity testing, a test setup similar to test setup 2 is essential. For example, to test the maximum number of attached UEs per sector, the following steps will be required: For the sector running on a no-load condition, perform a steady rate of UE attaches in steps of 32 attaches, 64 attaches, 96 attaches, 128 attaches, 156 attaches, 200 attaches, and SO forth. Observe the success rate for KPIs and monitor the drop in success rate. Continue performi

ng UE attaches until the success rate drops below an acceptable KPI threshold. In order to ensure that the attached UEs are not disconnected from the sector because of inactivity, maintain a steady UL/DL data rate for each of these attached UEs. Note the number of attached UEs after which the attach success rate starts to drop below an acceptable limit. This would be used as the maximum number of attached UEs per sector. Try attaches with different rates (30 attaches/second, 50 attaches/second) and also under different channel fading models such as the EPA, EVA, and ETU models. User Plane Overload and Capacity Testing User plane overload and capacity testing will deal mainly with data throughput capacity of the eNodeB. User plane overload testing will involve tests like: Maximum throughput that can be supported by a sector for MIMO users under ideal radio conditions. Maximum throughput that can be supported by an eNodeB for MIMO users under ideal radio conditions. Number of users that can be scheduled by a cell for each transmission time interval (TTI). Tests to verify the GBR share/sector. Tests to verify the n-GBR share/sector. Apart from these tests, the following tests also need to be performed under overload and capacity testing category: Memory consumption tests for different loads (no UE attaches, 32 UE attaches, 64 UE attaches, etc.). Computer processing unit (CPU) load consumption tests. Configuration of thresholds 1, 2, 3, and 4. Tests to load the eNodeB or sector to exceed threshold 1 and verify the actions that are triggered by eNodeB/sector to bring the CPU load below the threshold. Tests to load the eNodeB/sector to exceed threshold 2 and verify the actions that are triggered by eNodeB/sector to bring the CPU load below the threshold. Tests to load the eNodeB/sector to exceed threshold 3 and verify the actions that are triggered by eNodeB/sector to bring the CPU load below the threshold. Tests to load the eNodeB/sector to exceed threshold 4 and verify the actions that are triggered by eNodeB/sector

to bring the CPU load below the threshold. CHAPTER 1 NETWORK PLANNING Long Duration Testing Long duration tests are mainly stability tests, wherein the eNodeB/sector will be tested with calls that last for 48 to 72 hours. Some of the tests that fall under this category are: Single UE with UL/DL non-GBR data without any mobility and no change in channel condition observed for 72 hours (full throughput test). Single UE with UL/DL GBR data without any mobility and no change in channel condition observed for 72 hours. Single UE with UL/DL non-GBR moving at 3 kmph speed in a circle for 48 hours. Multiple stationary UEs with data transmission or reception of similar QoS class identifier (QCI) observed for 72 hours. Stability tests for UE using TM4 for DL transmission. Stability tests for a mix of UEs working under different transmission modes and engaged in data. An application server integrated with Iperf will be required to simulate constant UL/DL data at a desired rate. For multiple UE simulations, simulation tools can be used, and for simulation of UE speed and fading to test stability at different modulations, channel emulators can be used between the UE and the sector under test. EMS should be connected to the eNodeB to observe and monitor the KPIs. CPU load and memory consumption should be monitored as well during the course of the test. Summary This chapter covered the various phases in radio network planning, the parameters that can impact the different phases of planning, the essence of capacity and coverage planning, the various modes of deployment, and verification tests and steps for a deployment that a specialist should perform. Most of these topics are technology independent (i.e., the steps and goals would remain similar irrespective of 2G, 3G, or 4G deployment), however, the targets, especially in terms of KPIs, will be different because the throughput KPI targets in LTE will be much larger in comparison with the targets in 3G. The following chapter will introduce you to the concept of a Self-organizin

g Network along with a detailed overview of its architecture, major aspects and features. CHAPTER 2 Self-Organizing Networks in LTE Deployment This chapter is an introduction to self-organizing networks (SON). We will start with a brief introduction to the current network, its practical limitations, and the advantages of SON in the current network. We will then proceed with a discussion of SON architecture wherein we will explain the different types of SON, such as centralized, distributed, and hybrid, along with their advantages and disadvantages. We will then briefly discuss the different phases of SON and what activities are performed in these phases, and finally we will detail some of the SON features specific to LTE like automatic eNodeB setup, automatic physical cell identification (PCID) allocation, automatic neighbor relation, random access channel (RACH) optimization, mobility robustness optimization (MRO), and intercell interference coordination (ICIC) and explain the concept, design, and implementation of these features. Introduction to Self-organizing Networks In today's network, the activity of installation, deployment, and maintenance of a radio access network involves very high costs, especially considering the number of nodes that require deployment and maintenance. Also because of the dynamics of radio and traffic conditions, it becomes a very tedious task to be able to change the network settings to cater to these dynamics. A self-organizing network is often used to categorize a cellular network for which the tasks of configuring, operating, and optimizing are largely automated. It mainly targets to reduce operational expenses, improve operational efficiency, and enhance and maintain a gratifying user experience, even under adverse conditions. SON features can markedly improve the user experience by optimizing the network automatically and rapidly mitigating outages as they occur. This is an extremely important characteristic for all network operators because time to operation and time to repair

are critical factors for an efficient and well-managed network. By embracing the SON procedures and algorithms, operators will be able to use these capabilities significantly to their advantage. This chapter discusses the various SON features, how they are able to overcome the current issues, and how they benefit the operators when deployed. SON Architecture With LTE maturing and an increase in the demand for capacity, it has become crucial for the optimal use of all network resources to achieve higher end-user experiences and better revenue from the available resources. For any self-optimizations to take effect, some optimization algorithms are needed to drive the self-optimizations. CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT Based on the implementation and deployment, SONs can be classified as: Centralized SON Distributed SON Hybrid SON Centralized SON Centralized SON, as the name suggests, consists of a centrally located SON framework. In other words, the optimization algorithms take place in the Operations, administration and management (OAM) system. One main advantage of this kind of solution is that the SON functionality will be centralized and will be at a higher level, thereby being more use-case driven. The flip side of this solution is that the decisions are not as fast as the distributed SON type. Figure 2-1 presents a flowchart for the centralized SON architecture. Central SON Optimization decisions are made centrally and passed Element on to individual nodes. Management System Itf-N Itf-N ENodeB N Itf-N Cell N+2 Cell N+1 X2-link ENodeB 1 ENodeB 2 Cell N Cell A Cell C Cell D Cell F Cell B Cell E Figure 2-1. Centralized SON architecture In centralized SON, all SON functions are located in the element management or network management level, making it easier to deploy. However, multivendor integration for the optimization of a solution can be quite challenging, as there is no standard driven optimization procedure, and vendors can have the centralized SON implementation suited to their network

element. A centralized SON solution is a preferred SON solution for optimization solutions that may impact the functioning of more than one network element. Optimization results can be stored in a configuration database and, upon completion, can be automatically distributed toward the impacted eNodeBs (network elements). However, centralized SON restricts the possibility of quicker optimizations and SON adaptation, as there will be lag time in the network elements reporting the problem to the element manager and the central SON server applying the optimization algorithms. CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT Many vendors have a common optimization for simultaneous operation across multiple radio technologies. This can be a huge advantage, wherein joint optimization of LTE and GSM or WCDMA will be possible in an efficient way for these vendors. A centralized solution is preferred for the following LTE optimization features: Physical cell identity management Neighbor cell relation optimization Interference reduction optimization Coverage and capacity optimization Handover optimization and load balancing optimization Radio optimizations, like RACH optimization Distributed SON In distributed SON, the optimization algorithms reside not at the network or element management system level, but at the network element level (i.e., eNodeB). The SON functionality and the optimization algorithm are not necessarily at a high level but can reside in many locations or network elements (e.g., eNodeB). This makes the deployment and optimization activities quite complex, especially when there is a requirement for multiple network elements to coordinate with one another as part of a solution. However, for simpler cases with two or fewer network elements being impacted, it is more effective to approach this optimization or solution with a distributed SON model. Because the distributed SON solution resides at a relatively low level, the optimization algorithms can be executed much faster when compared with that for the

centralized SON. Figure 2-2 presents a flowchart for the distributed SON. Central SON Element Management System Optimization decisions are made at each eNodeB and parameters are tuned Itf-N Distributed SON accordingly. ENodeB N Itf-N Itf-N Cell N+2 Cell N+1 Distributed SON Distributed SON X2-link ENodeB 1 ENodeB 2 Cell N Cell A Cell C Cell D Cell F Cell B Cell E Figure 2-2. Distributed SON example CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT A distributed SON solution is preferred for the following LTE optimization features: Mobility robustness optimization Scheduler optimization for spectral efficiency VS. cell edge user throughput The reason that distributed SON is preferred for these features is because these optimizations do not impact the other cells in the network and would only optimize the cell or eNodeB to improve the KPIs of the cell. Also, the adaptations can be much faster when these optimization parameters exist on the distributed level. Hybrid SON Hybrid SON is a combination of centralized SON and distributed SON, wherein a part of the optimization algorithms is centrally located and another part exists at the element or nodal level in eNodeB. In hybrid SON, normally at the distributed level, algorithms and optimization schemes that exist are targeted to provide a quicker solution, whereas a detailed, well-educated solution algorithm normally resides on the centralized level. This kind of hybrid solution provides more flexibility to the network over that of a purely centralized or purely distributed SON solution, thereby making it practical for different types of optimizations. Figure 2-3 presents the flowchart for the hybrid SON. Central SON Some parts of the optimization decisions are Element made centrally and passed Management on to individual nodes. System Some parts of the optimization Itf-N Distributed SON decisions are taken on the ENodeB N individual node level. ltf N Itf N Cell N+2 Cell N+1 Distributed SON Distributed SON X2-link ENodeB 1 ENodeB 2 Cell N Cell A Cell C Cell D Cell

F Cell B Cell E Figure 2-3. Hybrid SON example A hybrid SON solution is preferred for the following LTE optimization features: Enhanced intercell interference coordination PCID collision or confusion detection The SON deployment activities can be grouped into several phases, as described in the sections that follow. CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT Planning and Provisioning Phase The planning phase of the SON is the phase wherein the operator decides first on the area for deployment. Depending on the coverage and capacity requirements, network planning needs to be performed. A planning tool is used extensively to create a network configuration and plans for the location for each of the network elements within the planned area to meet the capacity and coverage targets. Parameter planning is also part of this phase wherein the parameter settings for the eNodeBs are planned beforehand. Typically the parameters that need to be planned in this stage are the transmit power, antenna tilt, handover-related settings, and Tier-1 neighbor list. The PCID and root sequence index planning are also performed in the planning phase. Another important aspect of the planning phase is the transport network configuration planning, which enables the eNodeB to establish a link with other network elements like MME, neighboring eNodeBs, among others. Commissioning and Operation Phase This phase involves tasks like plug-and-play commissioning, during which the eNodeB automatically detects the hardware inventory by performing self-tests and brings itself up in an automatic manner. Upon bringing itself up, the eNodeB is able to download the new or upgraded software from a central location. Another important activity that is performed in this phase is the establishment of links with peer entities like MME or neighboring eNodeBs. During this phase, the eNodeB also performs automatic configuration by downloading some of the parameters from the central SON entity. More details on these processes are provided later in the cha

pter. Optimization Phase The optimization phase is a continuously ongoing phase wherein the eNodeB as well as the central SON entity monitor the performance of the network element for various aspects like throughput, handover success rate, CPU utilization or load, and so on and performs optimization tasks to improve these KPIs. Some of the key optimization tasks that are performed in this phase are: Mobility optimization, wherein the handover parameters are fine tuned to improve the handover success rate. RACH optimization, wherein the RACH parameters such as PRACH Config Index, frequencyOffset and SO forth are fine tuned to improve the RACH success rate and thereby improve the accessibility KPI. Scheduler parameter optimization, wherein the spectral efficiency and user experienced throughput are improved. Load balancing, wherein the cell suffering from high load and high usage is identified in the network and corrective actions such as diverting the traffic to a relatively low-loaded cell and SO forth are taken to improve the overall situation. The following sections present more detail on some of the features activated in these different phases of deployment. CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT SON Features Broadly, SON features can be classified into three categories: Self-planning features Self-optimization features Self-healing features Each will be discussed in the following sections. Self-planning Features Self-planning features are mainly targeted to reduce the initial deployment and installation costs. Features like automatic cell planning, automatic eNodeB setup, PCID planning, and automatic neighbor relation planning are grouped under this category. Self-optimization Features Self-optimization features continuously monitor the network performance, identify the problem areas, and then apply an optimization algorithm to improve the KPIs and the network performance accordingly. Some of the features that fall under this category are MRO, ICIC, and RACH optimization. Self-healing Features S

elf-healing features identify if there are any network elements or components that are down due to failure in a network and apply techniques to compensate for the performance degradation. They also automatically and remotely bring up the affected component. Features like cell outage compensation and load balancing fall under this category. The following sections discuss in details some of the SON features from the different categories and explain how they are designed and implemented. Automatic e-NodeB Setup A major part of the capital expenditure (CAPEX) for an operator, apart from the equipment hardware and spectrum costs, is the deployment cost for the network. Deployment of an LTE radio network can be a very demanding and challenging activity and can involve subtasks such as: Network planning Hardware and software commissioning or implementation Integrating the network elements in the hybrid or multivendor environment Optimizing the network elements according to the field results Maintenance and support For a complex deployment scenario, these deployment activities can cost more than the hardware and software costs. Therefore, it is very important that the complexity and cost of deployment are reduced and optimized as much as possible. CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT When it comes to the deployment of a network element, the integration and configuration of a network element are the most time-consuming parts of the process, and in most cases these are performed manually. Automatic eNodeB setup is a SON feature that aims to reduce the manual intervention for an eNodeB setup and commission as much as possible. This would also mean that the skill set requirement for the person who physically installs eNodeB is minimal, thereby bringing down the cost of deployment. Also, because there is less human interfacing the commissioning or bring up for the eNodeB, automatic eNodeB setup is less prone to human error. Automatic eNodeB setup will mainly involve the network management system as well as the

network element configurations. Steps for implementing an automated eNodeB bring up will involve: An automated self-test and self-discovery. This will involve eNodeB identifying the backhaul and management links that are connected to it. Also, the base band unit and the remote radio head unit will need to be able to detect and identify each other in this step so they can communicate within it. IP assignment. IP planning is done as part of preplanning, and the IP assignment can be done using a local console or dynamic host configuration protocol (DHCP). Here, the IP address assignment is not only for the new node, but also for the other connections that are configured, such as IP addresses for the Mobility Management Entitiy (MME), Serving gateway (SGW), other neighboring eNodeBs, and so forth. Automatic software management. Once the secure tunnels are established with the EMS, the eNodeB will attempt to download the software from the central repository. Automatic software management involves centrally maintaining (at the element management system level) the software versions that are used in individual network elements (eNodeB in this case), keeping track of the latest available software, and scheduling the upgrade for individual eNodeBs. Typically, an eNodeB will have a base band unit (BBU) and remote radio head (RRH) as its two main components, and there would be separate software versions maintained for these two units. Management is required for each software component separately. Automatic software management will also require the element management system to support preplanning software loads for newly commissioned eNodeBs and provision a default image for these eNodeBs. Another important aspect of automatic software management is support for a "backup" and "rollback." In cases of unsuccessful upgrade events, there should be a possibility for a rollback to the previous version of the software. Automatic configuration. Automatic configuration refers to the process by which the eNodeB configuration parameter

Individual cell's radio parameters, such as antenna tilt, power, and so forth PCIDs for each cell Any preconfigurations pertaining to interference cancellations (like frequency restrictions) Root sequence index CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT Handover-related settings (frequency offset, cell individual offsets for each neighbor, hysteresis, etc.) Blacklist or whitelist cells Transmission mode (transmit diversity or spatial multiplexing) Automatic inventory management. Automatic inventory management enables the element management system to collect and maintain information about the components within the eNodeBs in an automated manner. Typically the inventory data will consist of: Serial number and other manufacturing information Identification of all field replaceable units Firmware and component inventories All software inventories Typically, the inventory information is uploaded to the network element in an automated manner and will be looked up during the these events: When an eNodeB is initially commissioned When a change is made to the eNodeB When an operator needs this information Automatic interface setup. This step would involve the eNodeB in performing S1 setup with the configured MME and X2 setups with its neighbors. Upon success of these setups, it informs the element management system about its operational state. PCID Allocation Every cell in the network must be assigned one of 504 physical-layer cell identity (PCID) values (0-503). PCID values can be reused as long as no conflicts exist. The PCID assignment function automatically assigns PCID values to enable a newly commissioned eNodeB. The PCID allocation SON functionality should be able to support several high level functions, including: Initial assignment of temporary PCID values Transition to permanent PCID values following SON convergence Collision-free PCID value allocation procedure by the SON (i.e., the direct neighbors should not be using the same PCID values) Confusion-free PCID value allocation procedure by the SON (i.

e., the PCID values of neighbors of direct neighbor cells must again not use the same PCID values) Avoidance of PCID group associations with PCID groups in use nearby Avoidance of PCID group affinity, preferring assigning three values that form a PCID group Avoidance of PCID sector uniqueness, preferring assigning three unique PCID sector values Avoidance of PCID sector alignment with antenna bearing, preferring PCID alignment with antenna direction PCID allocation in a SON feature should also support collision and confusion detection and resolution. CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT Automatic PCID Assignment Background The PCID is a fundamental input for the physical layer, which implies potential radio interference if PCID assignment is not done carefully. As mentioned earlier, every cell in the system must be assigned one of 504 PCID values; therefore, PCID values will need to be reused in large systems. The automatic PCID assignment feature of SON removes the planning (reuse pattern) and provisioning issues from the process. Common Ground Initial PCID assignment provides the following capabilities: Operator can manually assign PCID values or use a planning tool Operator can choose to overwrite PCID values at any time It is inevitable that some PCID conflicts will occur regardless of the initial PCID assignment solution (e.g., vendor boundaries, RF oddities). When PCID conflicts occur, they can be resolved with a distributed algorithm. To do conflict resolution, eNodeBs use their neighborhood data obtained via X2 for new PCID selection(s) and automatically resolve the conflict. PCID Collision PCID collision occurs when the eNodeB cell is using the same PCID value and frequency as another cell with a direct neighbor relationship. Figure 2-4 provides an example of PCID collision. ENodeB 2 X2-link ENodeB 1 PCI 3 PCI 3 PCI 5 PCI 4 Figure 2-4. Example of PCID collision This figure shows that the two neighboring eNodeBs, eNodeB1 and eNodeB2, both having cells with the same PCID (i.e., PCI3). This

can be detected by the cells involved in PCID collision regardless of them being neighbors (i.e., both ends of a one-way relationship can detect it). CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT If two eNodeBs exchange PCIDs and the neighbor lists data via the X2 link, then both cells are in a position to detect and resolve the PCID conflict. A vendor-specific algorithm can determine how the conflict should be resolved. Some of the challenges in PCID conflict resolution may arise when: An operator has chosen a fixed PCID value for a particular cell and SON algorithms are not allowed to override it. An operator has chosen to disable the conflict resolution feature for SON. There are no PCIDs available in the allocated range; this might result in a situation wherein no value would avoid PCID conflict. PCID Confusion When two or more cells on the same neighbor list are using the same PCID value and downlink frequency, it can result in PCID confusion. This condition can be detected at the cell that owns the neighbor list as well as at all of the neighbor cells that are using the same PCID value. Figure 2-5 shows an example of PCID confusion. ENodeB 3 X2-link ENodeB 2 X2-link PCI 3 PCI 8 ENodeB 1 PCI 7 PCI 3 PCI 6 PCI 5 PCI 4 Figure 2-5. Example of PCID confusion In this figure, eNodeB 1 sees that eNodeB 1 and eNodeB 3 have the same PCID (i.e., PCI 3). However, eNodeB 1 and eNodeB 3 do not have an X2 link and are not neighbors, so they cannot resolve the conflict. PCID confusion is detected by eNodeB 2, and it sends an eNodeB configuration update message with Neighbor Information IE (NI) filled appropriately to eNodeB 1 as well as eNodeB 3, which in turn would trigger conflict resolution based on its settings. PCID confusion detection can also be a learned algorithm wherein the eNodeB detects the PCID confusion based on the UE measurement report. For example, consider Figure 2-5 where both eNodeB 1 and eNodeB 3 have a cell with PCI 3. Assuming that the eNodeB 2 is unaware of the cell with PCI 3 within eNodeB 1

, if the UE measurement reports to the eNodeB 2 with PCI 3 as the only strongest cell, there is a good possibility that the handover will be initiated by eNodeB 2 toward the PCI 3, which is under eNodeB3. However, if the UE reports multiple strong cells in its reports (i.e., PCI 3 and PCI 2) the service cell can detect this as a possible PCID confusion case as the reported PCIDs according to the serving eNodeB belong to a completely different geographic location. The eNodeB can further detect the exact cells and location of the cells that are under confusion by requesting the UE to report the E-UTRAN Cell Global ID (E-CGI) for the cells reported. In such cases, corrective actions can be taken to detect and resolve the PCID confusion. The resolution can be centralized wherein the eNodeB 2 reports the confusion to the central SON, and the central SON then perfoms a resolution act. CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT Automatic Neighbor Relation In any mobile networks, including LTE, the mobility of the user equipment is usually guided by the network and based on the measurements that are reported by the UE for the neighboring cells. The UE is usually configured by the network to report the measurement for a set of neighboring cells based on various parameters such as: Geographic location of the neighboring cell Cell capabilities (e.g., does the neighboring cell support LTE technology? If yes, does it support LTE FDD or LTE TDD?) Neighboring cell relation with the serving cell (e.g., what type of link exists between the two cells-X2 link, S1 link, etc., and is the neighboring cell blacklisted, etc.) Priority of the neighbor (e.g., the operator might prefer an intra frequency cell over an inter frequency cell; in such cases, the neighboring configuration should be done accordingly) It becomes very important for a cell to have a combination of static preplanned or commissioned neighbor allocation as well as a dynamic adaptive neighbor relation update based on the UE measurement reports and changes in t

he network to maintain flexibility within the network. Broadly, the neighbor relation management can be classified into two main steps: commissioned neighbor cell configurations and automatic neighbor relation updates. Commissioned Neighbor Cell Configurations This step deals with offline planning of the cells and its neighbors with the help of network planning tools. The operator may have minimal needs to plan for the IP addresses and the base station IDs or cell IDs for all the configured adjacent neighboring base stations or cells. The neighbor configuration can be static and does not require any assistance from the UE or other network elements. During the startup of the base station, the commissioned parameters should be read, and X2 connections should be established to all the commissioned neighboring base stations. The remaining cell configuration information can be exchanged between the two base stations via the X2 link formed. Typically, a single X2 connection is established between two base stations regardless of the number of supported cells for each of these base stations. This means all cells, each of them assigned with a unique global cell ID, have the same X2 IP address. If a newly deployed eNodeB has all the commissioning data, these will include the configuration data it runs from the X2 setup procedure to each configured neighbor. When the connection could be established successfully (e.g., the listed neighbor is already installed and commissioned), all required neighbor information can be exchanged between the two eNodeBs. If a listed neighbor does not respond, it can be marked as not reachable. When a base station in operational mode receives an X2 setup request from another base station, it typically should respond to the request, send its own cell configuration data, store the received configuration information in its own neighbor cell list, and mark it as reachable. If the requesting base station for the X2 setup procedure is already known and neighbor configuration information is available

with the target base station, the target base station can still send its own cell configuration to the initiation base station. The received information can then be compared with the existing ones and updated in cases of identified modifications. Automatic Neighbor Relation Updates This aspect is mainly 3GPP driven and is based on UE's measurement reporting. This means the UE detects the neighbor's signal strength and then reports to the serving cell with the PCID of the new cell. The serving cell is then based on these measurements, and it will add the newly discovered cell to its neighbor list. CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT However, the relationship between neighbor cells needs to be known by the respective base stations and should be planned accordingly because wrong configurations can result in a high rate of handover failures and call drops. Automatic neighbor relation updates can be broadly classified as a four-step procedure: Neighbor cell detection X2 configuration discovery of the neighboring site The X2 connection setup Neighbor relation optimization Neighbor Cell Detection An LTE cell can be identified in two ways. First, it is identified with the help of PCID, which is used for most of the RRC signaling, as it requires fewer bits for transmission. LTE cells can be identified and classified as neighbors based on the PCID; however, because there are only 504 possible PCIDs, there is a risk of a duplicate PCID being used by two different cells that are close to each other. The radio network planning and optimization engineers should ensure that for a particular coverage area, there are no two cells with the same PCIDs to avoid any conflicts. Second, the LTE cell can be identified with the help of E-UTRAN cell global ID (E-CGI) broadcasted as part of the SystemInformationBlock1, E-CGI is unique in the whole network and allows an unambiguous identification of a cell. Figure 2-6 shows the relation between the PCID and the E-CGI. PCID of the cell is transmitted as a part of reference s

ymbol Physical Cell id E-UTRAN Cell Global ID E-CGI of the cell is transmitted as a part of SystemInformationBlock Type 1 Figure 2-6. Physical cell ID and E-UTRAN cell global ID During a UE attach, the eNodeB sends across the measurement configurations to the UE. When the UE moves away from the serving cell, it starts reporting measurements indicating the strongest cell or cells to the eNodeB for any handover-related actions. The measurement report consists of only the PCID of the cells that match the measurement criteria. In such events, it is very possible that the UE reports a PCID that is unknown to the source eNodeB. In this case, the base transceiver station BTS configures the UE to further report the E-CGI (E-UTRAN cell global ID) for the unknown PCID reported cell. Upon the UE reporting the E-CGI for the new target cell, the source eNodeB will try to derive the IP information for the newly discovered neighbor. Figure 2-7 depicts this process. CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT ENodeB 2- Target cell ENodeB 1- Serving cell Figure 2-7. Neighbor cell discovery X2 Configuration Discovery of the Neighboring Site The E-CGI reported by the UE will now be used by the source eNodeB to discover the neighboring eNodeB. The source eNodeB will exchange the transport network layer (TNL) configurations with the target eNodeB via the MME by means of an information transfer procedure. Considering that the MME has already established the S1AP connections with the source and targeted eNodeBs individually, the eNodeBs can now exchange their configurations by means of transparent containers, as shown in Figure 2-8. eNodeB2 eNode81 1. S1 setup procedure (ECGI of cells for the eNodeB are known by MME) 3.eNB CONFIGURATION TRANSFER to eNode81 requesting for TNL 2. UE reports discovered cell-to configuration Info for X2 connection attached cell in eNodé8 2 4 .MME CONFIGURATION TRANSFER requesting for eNode81 TNL configuration information 5 .eNB CONFIGURATION TRANSFER with TNL configuration information 6 .MME CONFI

GURATION TRANSFER with eNode81 TNL configuration information Figure 2-8. IP address resolution procedure CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT X2 Connection Setup with Neighbor Cell Configuration Updates When the transport network layer configuration is received, the source eNodeB can then initiate an X2 connection setup with the target eNodeB. As a part of the X2 setup procedure, the two eNodeBs will exchange the list of serving cells with the respective eNodeBs. With this information, the eNodeBs can update the existing neighbor list with the new set of cells. The procedure is illustrated in Figure 2-9. eNodeB2 eNodeB1 MME CONFIGURATION TRANSFER with eNodeB1s X2 TNL configuration information IP tunnel establishment between the 2 eNodeBs SCTP establishment X2 SETUP REQUEST (eNodeB-ID, served cells information Neighbor cell table update on X2 SETUP RESPONSE eNodeB1 eNodeB-ID, served cells information) Neighbor cell table update on eNodeB2 Figure 2-9. X2 setup and exchange of configurations between the two eNodeBs Neighbor Relation Optimization Broadly, there are two types of neighbor relations. First, there is a neighbor site relation, wherein there is a direct X2 connection existing between the two sites. In this case, the communication link between the two neighbors is known. Second, there is a neighbor cell relation, wherein the given two cells have a common or overlapping coverage area. UE always reposts a neighbor cell relation, but never a neighbor site relation, and only properly configured neighboring relations are relevant for handover performance figures. Figure 2-10 depicts these relations. CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT X2-AP connection exists ENodeB 1 ENodeB 2 Neighbor relation PCI 6 PCI 5 PCI 4 No neighbor relation Figure 2-10. Neighbor cell and site relations Referring to the example given in Figure 2-10: The eNodeB 1 has an X2 connection to the eNodeB 2 to neighbor site relation The eNodeB 2 parents three cells with PCIDs PCI 4, PCI 5, and PCI 6 A neighbor cell

relation exists only to the cell with PCID 6 When a new neighbor site relation is established, the configuration information of all parented cells is stored in the neighbor cell list. But only the identified and respectively measured neighbor cell is listed as a relation in the neighbor relation table (NRT) between one cell and a neighboring cell of this site. During daily operation, a cell relation could fail to work properly for handovers (i.e., handover performance counters show higher failure rates than average). In this case, optimization algorithms can blacklist a relation. Another possible optimization to the NRT is to mark the neighboring cells that are either under congestion or in temporary outage where handover is not allowed (shown by the field handover allowed in Figure 2-11). By doing this, the system information block (SIB) for the serving cell can be updated to avoid the cell information that is congested or blacklisted. The RRC connection setup and RRC connection reconfiguration messages sent by the serving cell to the UEs can be updated to remove measurement criteria for the cell. This information can also be used when UE reports measurement to the serving cell; if there are multiple reports from the UE, the serving cell can ignore the measurement report for the cell that is marked as handover not allowed and proceed with handover preparations for the other measured cells by the UE. CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT Neighbor Site and cell List for eNodeB 1 Neighbour E-CGI Ip-address Connection ECGI1 172.168.10.1 ECGI2 172.168.10.1 eNodeB2 ECG13 172.168.10.1 PCI18 ECG14 172.168.10.2 eNode83 PCI19 ECGI5 172.168.10.2 eNodeB4 PCI102 ECGI6 172.168.10.4 Neighbor relation table on eNodeB1 Handover SI No E-CGI X2 Link Allowed Priority ECGI1 ECG12 ECG13 ECGI4 ECGIS ECGI6 Figure 2-11. Neighbor site and cell list and neighbor relation table The NRT can be further optimized to improve the KPIs for a particular neighbor by adding an X2 link if there are too many handovers performed betwee

n the two cells or eNodeBs and there is no existing X2AP link between the two. Considering that the X2AP links that can be maintained by an eNodeB are often limited, there needs to be some mechanism for the eNodeB to decide on maintaining an X2 link with one eNodeB over another. One way to achieve this is for the eNodeB to maintain a priority tag against each neighbor in the NRT and update it at a predetermined interval based on the number of handovers performed between the two eNodeBs and the percentage of handover failure for the pair and the operator to which the neighboring eNodeB belongs. With this combination, the eNodeB should be able to use the X2 links only between high-priority neighbors and release the X2 link for the lower-priority neighbor. A sample NRT is shown in Figure 2-11 with some useful information for each of the neighbors. SON and Self-Optimization Motivation of Intercell Interference Coordination The LTE systems operate mainly with a reuse 1 factor, as this helps maximize the capacity and bandwidth usage. However, this also means that there is a very high scope of interference for the UEs that are at the cell edge. Depending on the measured signal strength from the neighboring cell at the cell edge, the interference can lead to a high performance or throughput loss for the cell edge users. In turn, this will lead to a drop in spectral efficiency for the entire network. Intercell interference coordination (ICIC) solutions try to bring balance between the cell throughput and cell edge users' data rate. CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT Principle of ICIC and Frequency Reuse Frequency reuse is one of the best means to reduce intercell interference. The frequency reuse coefficient can be 1, 3 or 7. When the reuse coefficient equals 1 (i.e., adjacent cells are using the same frequency resource), at the cell edge, interference would be serious. A larger reuse coefficient can reduce intercell interference further, but care should be taken when doing this, as it will directly resu

lt in reduced spectral efficiency and cell throughput. ICIC works on the principle wherein the frequency resource is divided into multiple frequency blocks (three blocks or seven blocks). Each cell will have a set of predefined physical resource blocks (PRBs) that are marked as cell center PRBs and another set of PRBs will be marked as cell edge PRBs. Depending on the user location, a UE is either marked as a cell center UE or a cell edge UE, and the PRBs that are marked for cell center usage are provided to the cell center UEs. The PRBs that are marked for cell edge usage are provided to the cell edge UEs by the eNodeB scheduler. Depending on the pattern of usage, care is taken to ensure that no two cells that are adjacent to each other will have the same set of PRBs marked as cell edge. Two UEs from neighboring cells (cell edge users) will not use the same set of PRBs, thereby reducing any interference to each other and improving the cell edge throughput. Figure 2-12 is an example of an ICIC deployment with a reuse coefficient pattern of 3. Cell edge Cell center Cell 1 Cell center Cell 2 Cell center Cell edge Cell 3 Reuse Pattern 3 Cell edge Figure 2-12. ICIC deployment with a reuse coefficient pattern of 3 ICIC may be of two types: static and dynamic. Static ICIC is a basic and simpler approach to implementing the ICIC feature. It is more suitable for cells that are predictable based on the type of traffic and the load distribution. In static ICIC, the parameters are configured during eNodeB commissioning itself. There are no reconfigurations and signaling with peers to negotiate on the resourcing pattern involved in static ICIC. Though it is a simpler approach, it is not necessarily efficient, as the planning does not always cover the different loading conditions and can result in performance degradation. Dynamic ICIC is a more complex approach based on the traffic conditions and cell loading. Dynamic ICIC is more suitable for cells where the cell loads a condition and the user distribution within the cell is

not always predictable. Reconfigurations are typically triggered as part of dynamic ICIC to bring a balance between the cell edge user throughput experience VS. spectral efficiency. As part of dynamic ICIC, there are resource negotiations between the two eNodeBs via the X2 link as well. Though it is not as simple to implement the dynamic ICIC, it is more efficient and practical, as the cell sizes and the user distribution pattern in any network are not uniform. CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT Some of the key aspects that should be considered when designing and implementing ICIC with fractional frequency reuse mechanism are: Implementing fractional frequency reuse. In the downlink, the fractional frequency reuse (FFR) scheme can be based on a preferred list of red, blue, greens (RBGs; color scheme). For each sector, its edge band consists of a subset of PRBs listed earlier in the list, and its center band consists of the remaining PRBs in the list. The common control channels of given sector are assigned edge PRBs to the maximum extent possible to avoid overlapping with edge PRBs of neighboring sectors. Further, it could be made flexible by allowing the operator to configure the number of PRBs that are reserved for cell edge and cell center. Cell edge UE identification. Considering the fact that the fractional frequency reuse design revolves around the PRB allocation for UEs ranged at different distances from the ENodeB, it becomes important that the UEs are classified as cell edge or cell center as accurately as possible for optimized results. There can be different mechanisms used by different vendors to indentify the UE as a cell edge user. One of the most popular methods to determine or classify the UE as a cell edge user is based on the UE reported RSRP or RSRQ. Typically, there would be a configurable threshold value that can be set per the operator's requirement. When the reported or average RSRQ or RSRP value from the UE goes below the threshold value, the UE can be classified as a ce

ll edge UE. Further, the bandwidth allocation for cell edge users can be either static or dynamic. In static resource allocation, the frequency band for the cell edge users and cell center users are statically configured. In dynamic resource allocation, for each subframe, the target allocation bandwidth is determined by resource usage in past subframes. When the system is less loaded, all users will be scheduled within the edge band only when the system is overloaded beyond a particular threshold. Edge users will be scheduled in the edge band, and center users will be scheduled in the center band. RACH Optimization In LTE, a random access procedure can be performed by user equipment under the following conditions: IDLE user equipment performs an initial access procedure to move to the CONNECTED state. User equipment performs a handover to a target cell. When user equipment is in a CONNECTED state but out of time, synchronization occurs with the network, and it receives downlink data. When a UE is in a CONNECTED state but out of time, synchronization occurs with the network, and it attempts to transmit uplink data. When a UE performs reestablishment after it detects a radio link failure. Need for RACH Optimization In a homogenous network wherein the neighboring cells are using the same center frequency for uplink transmission, PRACH planning becomes very important and can very well be performed using self-optimization techniques. User equipment that intends to move from an IDLE to a CONNECTED state to send or receive data will need to first perform a RANDOM ACCESS procedure. CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT The necessarily cell-specific RANDOM ACCESS procedural details are provided to the UE in systemInformationBlockType2( (SIB2). Each of these broadcasted SIB2 parameters can be fine tuned or optimized by the network in order to: Optimize the balance between the RANDOM ACCESS opportunities and UL data transfer. This is especially a concern in TDD modes of operation where there are not too many

uplink opportunities for data transmissions, as well as RANDOM ACCESS procedures. During busy hours, when there are many users already connected to the cell and many more trying to connect to the cell by means of RANDOM ACCESS procedures, it becomes very important for the operator to be able to select the right base station setting for RANDOM ACCESS opportunities in order to balance the distribution of the uplink subframes between RANDOM ACCESS opportunities and user data transfer opportunity. Reduce interference during a RANDOM ACCESS procedure. In a homogeneous network, where there are multiple cells deployed, there is a huge probability of uplink interference that can result in performance drops for the entire network. Care should be taken to ensure that the neighboring cells that are operating at the same frequency avoid using the same root sequence number. Also it is very important to ensure that the PRACH opportunities between the two neighboring cells are different and do not have RACH opportunities at the same time and frequency. Reduce call setup and handover delays. This is extremely important in a deployment where there are users who are traveling at a higher speed. The RACH optimization feature aims at automatically fine tuning the RACH parameters to enhance system performance. One of the targets listed in Table 2-1 should be used. Some of the performance targets that are configured by the operator as per the guidelines from 3GPP specs are outlined in the table. Table 2-1. Performance Targets from 3GPP Specifications Target Name Definition Legal Values Access probability (AP) The probability that the UE has for accessing the network Cumulative Distribution Function after a certain number of random access attempts. (CDF) of access attempts Access delay The probability of the delay that an UE can experience CDF of delays probability (ADP) while accessing the RACH. RachOptAccessDelay This defines the list of ADPs that is acceptable. ADPs are listed as a pair (a,b). Probability For a sampling period of ti

me, each entry in the ADP An a in the list represents list defines the maximum time delay before an UE probability in percentages and a b can access the random access channel with a specific represents the delay in ms. success percentage. If ADPx's a is larger than that of This target is suitable for RACH Optimization (RO) ADPy, then ADPx's b also has to be larger than the b of ADPy. (continued) CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT Table 2-1. (continued) Target Name Definition Legal Values RachOptAccess This defines the list of Access Probability (APn) that is APn's are listed as a pair (a, n). Probability acceptable. An a in the list represents Each instance of APn of the list is the probability that probability in percentages and the UE gets access on the RACH within n number of n represents the number of attempts over an unspecified sampling period. attempts. For a sampling period, each entry in the APn list If APx's a is larger than that of APy, defines the probability of an UE to be able to access the then APx's n has to be also larger RACH within n attempts. than n of APy. This target is suitable for RO. roSwitch Indicates if the RACH optimization feature is activated On, off or deactivated. The parameters presented in Table 2-2 are broadcasted by a base station under SIB2 that corresponds to the PRACH characteristics. Table 2-2. PRACH Configuration Elements ASN1START PRACH-ConfigSIB ::= SEOUENCE { rootSequenceIndex INTEGER (0..837), prach-ConfigInfo PRACH-CinfigInfo PRACH-Config ::= SEQUENCE { OPTIONAL Need ON rootSequenceIndex INTEGER (0..837), prach-ConfigInfo PRACH-CinfigInfo PRACH-ConfigInfo := SEQUENCE { prach-ConfigIndex INTEGER (0..63) highSpeedFlag BOOLEAN, zeroCorrelationZoneConfig INTEGER (0..15) prach-FreqOffset INTEGER (0..94) Prach-Configlndex The parameter Prach-ConfigIndex provides the information about random access opportunities that are available in a particular sector of the UE. There are 64 possible values that a Prach-ConfigIndex can assume, and each value provides info

rmation about the network the UE can use to attempt random access: Preamble format of the PRACH used Number of PRACH opportunities CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT System frames during which the PRACH opportunities are configured Subframes where the PRACH opportunities are configured by the network The preamble format and parameters are shown in Figure 2-13. Sequence Random access preamble format Random access preamble parameters Preamble format 3168T 24576T 21024Ts 24576T 6240-T 2.24576.T 21024T 2.24576.T 4096T Figure 2-13. Preamble format and paramters used for PRACH For an FDD deployment, the number of PRACH opportunities within a subframe cannot exceed 1. However, due to the shortage of uplink opportunities in the TDD network, there can be up to six PRACH opportunities. In order to reduce the call setup delays or handover delays, one of the parameters that can be optimized is the Prach-ConfigIndex, which can enable multiple PRACH opportunities for a UE wanting to establish connection. Mobility Robustness Optimization Mobility robustness optimization aims to improve system performance by optimizing the handover parameters. There are three popular reasons for handover failures for any LTE network: too early handover, too late handover, and handovers to a wrong cell. The mobility robustness optimization feature tries to optimize the cell offsets for two neighboring cells in order to reduce the failures that occur for these reasons. Optimization of these failures not only helps reduce the call drop ratio, but also significantly brings down the signaling load on a cell or eNodeB. Mobility robustness is part of the 3GPP's SON use cases (TR 36.902 V9.0.0). During deployment, the operator defines a default set of configuration for the mobility parameters based on the type of deployment (e.g., urban, dense urban, small office, rural). This set of parameters (default parameterization) not always optimal and might not yield the desired KPIs upon deployment. Mobility robustness features aim to target

these cells that are not optimally configured to continuously improve the mobility parameters by fine tuning them until a desired result is obtained, hence improving the overall mobility performance. Detection of poor mobility performance is based on the long-term evaluation of certain mobility counters and KPIs. The operator can further specify how mobility robustness should be fine tuned by being able to configure trigger conditions and exit criteria for these triggers. Let's have a look at the types of handover failures, how they can be detected, and what actions can be taken for each of these failures. CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT Late Handovers Late handovers often result in a radio link failure to the established connection even before a handover procedure is initiated on the source eNodeB side or the handover procedure is completed on the target eNodeB side. Often, in a late handover case, the UE tries to reestablish the radio links with the destination eNodeB upon radio link failure. A popular means to reduce or have a check on the late handover failures is to detect such RRC reestablishment messages sent across from the UE to the target eNodeB and send them across an RLF indicator to the original source eNodeB. There the source eNodeB will recognize this as a late handover and can appropriately adjust the cell individual offset (CIO) for that particular neighbor SO the percentage of late handovers is reduced for the cell pair. This mechanism is described in detail in Section 22.4.2 of TS 36.300-990. The RLF indicator message is described in Section 8.3.9 of TS 36.423-960. For cases where the source and destination cell pairs belong to the same eNodeB, there will not be any X2 messaging to pass on the RLF indicator. Instead, the MRO corrective action will be taken care of by the eNodeB internally. Figure 2-14 shows an example of how a late handover case is handled as part of the MRO solution. Cell belonging Cell belonging to eNodeB B to eNodeB A X2 AP X2 AP handover process initiat

ed or uninitiated Connection re-establishment Radio link failure detection RRCtoX2RLFMessage X2 AP: RLF Indicator X2APtoMRORLFMessage Too Late handover detection Figure 2-14. How a late handover is handled as part of the MRO solution CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT Early Handovers Early handovers often result in a radio link failure for a UE that has been handed over to a target cell after handover is complete. Often in an early handover case, the UE tries to reestablish the radio links with the previous source eNodeB upon radio link failure. A very popular means to reduce or check on early handover failures is to detect such RRC reestablishment messages sent across from the UE to the source eNodeB and then send across an RLF indicator to the target eNodeB where the radio link failure occurred. The target eNodeB then recognizes the UE as the one that was originally handed over from the same eNodeB and sends across a handover report message to the original source eNodeB, thereby indicating this as an early handover case. In the source eNodeB, upon detecting this case as an early handover case, appropriate adjustments to the CIO for that particular neighbor can be carried out so the percentage of late handovers is reduced for the cell pair. This mechanism is described in detail in Section 22.4.2 of TS 36.300-990. RLF indicator message and handover report message are described in Sections 8.3.9 and 8.3.10 of TS 36.423-960. For cases where the source as well as the destination cell pairs belong to the same eNodeB, there will not be any X2 messaging to pass on the RLF indicator or the handover report. Instead, the MRO corrective action will be taken care of by the eNodeB internally. Figure 2-15 shows an example of how a too early handover case is handled as a part of the MRO solution. Cell belonging Cell belonging to eNodeB B to eNodeB A X2 AP X2 AP handover completion Connection re-establ shment Radio link failure detection RRCtoX2RLFMessage X2 AP: RLF indicator X2 AP: hand ver report X2APtoMRORL

FMessage Too early handover detection Figure 2-15. Too early handover case is handled as a part of the MRO solution CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT Handover to Wrong Cell Handovers often result in a radio link failure for a UE immediately after being handed over to a target cell, followed by the UE trying to reestablish the radio link connection with a cell that was not involved in the handover procedure. For example, after a successful handover from the cell belonging to eNodeB A to a cell belonging to eNodeB B, RLF happens, and the UE attempts connection reestablishment in the cell belonging to eNodeB C (as shown in Figure 2-16). In this case, eNodeB C sends an RLF indicator message to eNB B, followed by eNodeB B sending across a handover report message to eNodeB A indicating handover to the wrong cell. This mechanism is described in Section 22.4.2 of TS 36.300-990. Cell belonging Cell belonging Cell belonging to eNB B to eNB C to eNB A X2 AP X2 AP X2 AP HO completion Connection re-establ ishment Radio link failure RRCtoX2RLFmessage detection RRCtoX2RLFMessage X2 AP RL Indicato X2 AP HC report X2APtoMRORLFMess HO to wrong cell Figure 2-16. How handover to the wrong cell is handled Note that the handover report message will be sent even if eNodeB B and eNodeB C are the same and RLF indication is internal to the eNodeB. Also, if handover fails during handover from a cell in eNodeB A and the UE attempts to reestablish the connection to a cell in eNodeB C, eNodeB C will then send an RLF indicator to eNodeB A. In handover to a wrong cell, eNodeB A should note both eNodeB C and eNodeB B as handovers to the wrong cell CT (correct target) and handover to the wrong cell WT (wrong target) and adjust the corresponding cell individual offsets accordingly to reduce the rate of handover to wrong cell failure. Figure 2-16 shows an example case of how handover to the wrong cell case is handled as a part of the MRO solution. CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT Load Balancing Clear load bala

ncing strategy helps the multitechnology operator to optimize the use of their investments. Also, having a clear criteria for serving different types of traffic in certain network layer helps to ensure the capacity and service performance for mobile network subscribers. For these tasks, a centralized approach over the entire network area helps in minimizing the effort and time spent on load balancing. Load balancing in LTE focuses mostly on air interface load balancing between different cells, between cells on the same of different frequency, and between different technology cells. Real-time load sharing is controlled with the handover triggering parameters and thresholds. The local optimization is based on real-time measurements. There is also centralized support for load balancing in order to follow and ensure the adequate capacity in each network layer. Typically, the network management system collects the actual network statistics of load, handover, and cell performance and defines different criteria for these statistics based on the type of cell. Typically, the central SON optimizer tool should be able to use the centralized information in other tools and analyze the load and handover performance against the actual parameters. The optimizer tool should be able to detect the cells continuously having high loads cells that are soon to reach the capacity limits and cells that experience the handover ping-pong effect and need optimization to improve the overall situation. SON and Self-healing Self-healing is one of the most important aspects of SON. There are various features that facilitate the self-healing aspect of SON, such as handover optimization, load balancing, cell outage detection and compensation, energy saving, among others. Each self-organization use case or feature involves fine tuning and optimization of a set of parameters in a network element. The main goal for this parameter tuning is to enable self-configuration, self-optimization, and self-healing for the individual network elements or for th

e complete network as an entity. The main aim of self-healing is to detect a failure or outage and perform repair actions to recover the system from that outage or failure. Let's look at an example case of cell outage detection with compensation to further explain the SON self-healing mechanism. Cell Outage Detection Cell outage is detected by means of various alarms and alarm correlations. The outage can be from a service, process, or the complete cell itself. The network element should be designed to raise alarms or report counters or KPIs corresponding to these outages and pass it on toward the element manager system (EMS) for further actions. At the central SON level, outage KPIs are continuously monitored. If KPIs are below threshold or if the alarms for the network element exist for an extended period of time, the cell is then marked on the outage list by the central SON. Cell Outage Compensation As part of outage compensation, the central SON will need to detect the best set of cells that can compensate for the cell outage. The compensating set for an outage cell can be predetermined by the central SON, wherein for each cell, an outage compensation is calculated prior to and upon outage detection, and these compensation algorithms are then executed. The downside of this method in outage compensation is that the compensation calculation is static and does not consider the current load condition of the compensating cells. Also in the case of multiple outages within the network, the predetermined cell outage compensation algorithm will be suboptimal. CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT Another way to compensate for the outage cell is to dynamically (upon outage) determine the best cells that should be used for compensation. Some of the considerations for candidate cell selection can be: Proximity of the cell to the outage cell. Often the higher priority neighbors are selected as compensating cells as the amount of parameter tuning that will be required for these tier-1 neighbors will be compa

ritively lesser. Current load condition of the candidate cell. As a result of cell outage compensation, it very likely that after the compensation procedure is performed, the candidate cells will have a larger coverage area with an increase in the number of users that are served by it. Also the incoming handover ratio can increase for the cell because of the compensation procedure. Care should be taken to ensure that the candidate cells are not already under high load condition with poor KPI performance. Type of the cell. Compensation algorithms will also consider the type of the candidate cell for any outage calculation (i.e., an indoor cell is not typically chosen as a candidate cell for compensation of an outdoor cell). Also, it is important that the candidate cell belongs to the same service provider so the impact to the revenue will be minimal. The advantage of dynamically determining the candidate cells for compensation is that the estimation is more realistic and the risks are less as compared with that of static determination of the candidate cells. However, the time taken for the cell outage compensation will be much higher when performed dynamically. As part of outage compensation, the parameters that are reconfigured on the candidate cells are typically cell mobility parameters, neighbor lists, cell transmit power, and antenna tilt. The tuning is performed in smaller steps and the KPIs are monitored for improvement. The action is stopped once the KPIs are in an acceptable range or the alarm for the outage is cleared. Figure 2-17 presents a flowchart representing the cell outage compensation procedure for central SON. Monitoring Alarms and KPI's for outage Cell Outage detection Select the Candidate List for outage compensation Perform compensation action for the candidate cells in smaller steps rollback evaluate If the compensation effect is causing improvement in KPI This step is performed only till the Outage detection is true but KPI's are improving. Figure 2-17. Cell outage compensation procedure fo

r central SON CHAPTER 2 SELF-ORGANIZING NETWORKS IN LTE DEPLOYMENT Benefit of Cell Outage Compensation There are three advantages to cell outage compensation procedures: Minimizes the damage and downtime until the faulty cell is repaired. Repair activity can be postponed or planned, as the impact will not be that high due to cell outage compensation. This will result in reduced maintenance costs (e.g., no permanent on-call service, smaller stocks of spare parts, combining of trips). Cell outage compensation is extremely useful and critical in places where the frequency of faults is higher. Summary This chapter provided a brief introduction to SON and its benefits and provided an overview of SON and its architecture, deployment aspects, and some of its major features. The following chapter talks in length about the deployment, business, and economic challenges that the fourth generation of mobile telecommunications technology is surrounded by. CHAPTER 3 Deployment Challenges in Evolving 4G Experienced network operators will often agree that every deployment of any technology is a unique experience. Veteran players in the telecom field often have stories to swap when it comes to relating their experiences with the deployment of solutions in the field. For development organizations, design wins might indicate revenues to be enjoyed, but successful deployments are the real crowns to be worn with pride and bragged about in competitive environments. So what it is about deployments that make them such celebrated events upon completion yet such daunting tasks to undertake? There are a variety of reasons for this. Many times, there tend to be problems inherent in the technology itself. The practical aspects of running a business in a highly competitive regulated environment like telecom lead to many other challenges. A critical part of network and service planning rests upon user profiles (i.e., how the users and their usage patterns are projected). The changing nature of real-world users and their usage patterns pose a c

hallenge. Applications and services that are used when designing a solution may become outdated by the time of deployment. Marginal services might gain some time in the lime light due to changes that cannot be foreseen, like the sudden popularity of some particular applications in the market or when new user equipment is introduced that seemingly increases traffic of a particular type. Being watchful for such applications is important, because usage profiles should definitely be taken into consideration when designing a solution. This chapter will cover some of the traditional challenges that are encountered in LTE deployments when taking into account the nature of business, technology, and user equipment. We will then shift our focus toward traffic and user profiles. We also highlight some futuristic trends that would necessitate certain requirements of 4G solutions in these areas to help prepare for these types situations in a more efficient manner. Technology-Related Challenges LTE as a technology has been adopted widely around the world. Yet the issues presented here pretty much remain active and need to be taken into account for a practical assessment of deployment issues, even if they are only partially addressable in implementation solutions. Interference Issues LTE as a solution has been designed to improve capacity and provide higher speeds at better QoS. Higher capacity mostly equates to higher transmission power and places a lot of importance on power control SO that cell edge users do not face too much interference. The farther the UE is from the eNodeB, the higher the power with which it needs to be transmitted. Also, even with the hierarchical layering of cells using different cell sizes (Macro, Pico, and Femto), interference still needs to be managed, both between the LTE cells and in an LTE cell (intracell interference) and with other radio access technologies that may exist (2G or 3G cells). Thus, intercell interference definitely increases in such scenarios. The tuning of such parameters in real

time is even more difficult as environmental conditions vary during the day and the traffic patterns experienced by the eNodeB can also vary by the time of the day. CHAPTER 3 DEPLOYMENT CHALLENGES IN EVOLVING 4G Note Interference issues and manual tuning for coverage and capacity are discussed in more detail in Chapter 1. The LTE standards have provided for the intercell interference coordination (ICIC) as part of Rel 8 and an improved enhanced ICIC (eICIC) feature as part of Rel 10 for heterogenous networks. Implementing eICIC requires self-organized networks to be implemented, including optimized algorithms to reduce interference. In a traditional macronetwork-based environment, ICIC improves interference only on traffic channels. Enhanced ICIC is dealt with in more detail along with self-organizing networks in Chapter 2. Spectrum Harmonization LTE as a solution has been designed to work with multiple bandwidth options and multiple frequencies. LTE supports 31 operating bands for the paired spectrum (FDD) and 12 operating bands for the unpaired spectrum (TDD). Typically, the bands are allocated to specific frequencies, which are popular in specific countries, such as in the United States and Japan, to help with better acceptance. Although the intention of supporting multiple bands and frequencies and providing for a wide range of supported bandwidths is to make an easier adoption of LTE with enough flexibility to support available bandwidth and frequencies, this definitely imposes problems for the seamless user experience that subscribers may tend to take for granted. A user with a GSM mobile, for example, could comfortably roam across the world without having to worry about compatibility and availability of services. A similar LTE user may not be able to do the same unless her phone supports the multiple bands as part of the services deployed by operators in such territories or countries. We have already encountered issues of users purchasing LTE phones in some countries only to be surprised disappointed that

the same phone is not supported by the LTE services offered by the operators in their home countries or countries of travel. If standardization efforts are not able to narrow down and harmonize the frequencies to be supported across geographies, the onus then tends to fall on the UE manufacturers to support multiple bands. This may increase the cost of production and maintenance for the manufacturers and increase processing capacity, which, again, affects and imbalances the ecosystem. Higher UE costs may imply a larger barrier for user adoption and will also make it imperative to create business plans that will recover investments and work against operators' efforts to provide the best services possible at the lowest cost per user or serviced bit. Note The cost per bit and other items like average revenue per user are discussed in more detail later in this chapter. Voice Over LTE Implementation In spite of the usage of data catching up with traditional voice offerings, as networks have evolved from 2G to 3G and LTE, voice still plays a prominent role in the product to be supported by an operator. Voice is of so much importance that LTE has evolved specific options and solutions to support it. For those operators who want to optimize their networks for higher data performance and offload voice processing, there is the option of circuit-switched fallback (CSFB). In this solution, the operator hands the UE over a 2G or 3G network to support voice, while suspending the active data connection with the LTE network. This implies CHAPTER 3 DEPLOYMENT CHALLENGES IN EVOLVING 4G that the operator needs to continue to support 2G and3G networks, including their legacy core network elements. Such solutions may not appeal to the service-conscious business user, for instance, who expects all services to be available without interruption all the time. Note See Chapter 1 for more details on the key KPIs defined by the Third Generation Partnership Project (3GPP) that need to be supported by systems to help manage QoS. Voice over L

TE (VoLTE) also comes with its own stringent requirements for minimal latency of voice packets, which may imply that even within services offered to the same user, voice may need to be given higher priority than pending data packets for the user. Normally, such complex conditions are met by implementation of scheduling algorithms that support guaranteed bit rate services such as the proportional fair scheduler (PFS), for example, but the implementation and testing of such solutions gets even more complicated with multiple users with differentiated priorities, classes, and services. There are more solutions for implementing VoLTE support using existing IP multimedia subsystem (IMS) based core networks, which uses single radio voice call continuity (SRVCC) to address voice requirements. The IMS is used to control the handoff of the user to legacy networks that support voice. Thus, the operator needs to plan the solution with IMS support in mind when starting deployment. Multivendor Interoperability The original goals for LTE when the standardization efforts were made was to make standards in such a way that network elements from different vendors can interoperate seamlessly. Such a solution would give operators the benefit of being able to mix and match network elements as they desire to give them competitive advantage and prevent overdependence on any particular vendor implementation. This would have allowed operators to be able to negotiate price and services and also allows for the market to possibly gravitate toward the best performing network elements for appropriate functions. Reality, as usual, is painting a different picture. It has been found that interoperability is not ensured between equipment from different vendors. In most cases, extensive testing is needed to conform that equipment and solutions are interoperable. Although the early entrant to the market has an advantage in terms of being the default implementation to which the later vendor needs to comply, the fact remains that true multivendor inte

roperability still remains a dream. Not only does in take too much effort and testing to ensure interoperability, but in some cases, it also implies that existing solutions provided by vendors need to be contractually maintainable and have support for further improvements or corrections for interoperability, which may not be the case for many implementations already deployed in the field. All of this ends up adding to the overall cost for deployment and maintenance, especially in some cases when specific vendor solutions need to operate together due to network planning in some particular sites or locations. Tracking such issues and managing them closely are the only options available to the operator to ensure successful deployments and good user experiences when they move between different vendor network elements. Issues Related to Backhaul LTE as a technology has been designed to provide very high speeds on the radio UE side. If we trace back to earlier technologies like 2G or 3G traditionally, the radio, including UE and NodeB, has been the side that has limited networks' throughput. With the removal of that limitation, experts perceived that the attention would shift back to the backhaul for being able to support the required bandwidth. Operators not only need to be able to design the backhaul to be able to support required load generated by LTE terminals with very high data capabilities, but they also need to be able to add required additional capacity as needed by the growing network and subscriber needs. CHAPTER 3 DEPLOYMENT CHALLENGES IN EVOLVING 4G Also, as an improvement in the architecture from 3G, LTE introduces the eNodeB as a single node that collects the functions of both the radio network controller (RNC) and the base station. In 3G networks, operators can plan a layered approach, with the RNCs aggregating the connections to the base stations and then being connected in a limited manner to the backhaul (core network). The LTE collapsing of the radio nodes into the eNodeB has also raised a related r

equirement of eNodeBs themselves being connected to one another, along with the need to be connected to the core network elements, thus leading to a mesh structure. Hence, operators need to be very clear as to how they will achieve the required capacity in backhaul and yet be able to support such complex requirements for high capacity from LTE network deployments. This also means there needs to be sufficient investments made to realize and implement such a powerful backhaul capacity to handle the current and future LTE deployments. Environment Issues Aside from the issues discussed above, there are some challenges inherent to the technology and ecosystem that is evolving LTE development. This section addresses such issues, which are not only necessarily technology related, but also are inherent to the complexity of the technology and the environment under which such development is happening. We discuss issues relating to UE, issues with the execution of complex projects where all network elements, including UE, eNodeB, and the EPC, are evolving at the same time. Also, we highlighted the role that is played to a certain extent by standardization efforts and limitations. Some of these issues are not necessarily unique to LTE; but they need to be understood and acknowledged by the operator, if only to get a better handle on managing the challenging task of deployment under such conditions. UE Maturity During the development of any new technology or feature, there is always a lack of valid UE that could be used during network element (NE) development to validate the functionality of the NE or feature with assurance (from the eNodeB point of view). Mostly, this is due to the fact that the UE is also being developed in parallel to the eNodeB development timeline. In such scenarios, the NE development proceeds with the usage of appropriate simulation tools to simulate UE behavior. Note What is valid for the eNodeB is also valid for the LTE UE development. Hence, UE developers too should be managing the lack of an eNodeB

to validate the new feature. In the absence of a real NE, development features get validated against simulated NEs, with simulated interfaces and multiple assumptions about the feature and technology (where the specifications are not precise or leave room for interpretation). Hence, there are bound to be integration issues when the commercial UEs do become available. In cases where cooperation agreements between network development and UE vendors exist, such issues possibly get addressed better in extended integration testing. In most cases though, this implies issues when multiple commercial UEs diverge in their implementation or interpretation of standards. As a corollary to the above, where there are debug versions of UEs released, to help testing of other NEs, like eNodeBs, support may not be available for new features that are needed. Where there are commercial UEs available for testing, they may not provide enough debugging or logging capabilities to be of use during feature development and debugging. CHAPTER 3 DEPLOYMENT CHALLENGES IN EVOLVING 4G Feature Availability As discussed in the previous section, when all NEs are under development, planning integration testing and the rollout of a feature could be very difficult tasks. It could be that the standardization plans for the feature are made available for some projected period, requiring the network and UE vendors to commit to some date based on the same, so operators, in turn, could plan appropriate times for integration and field testing and deployment of the feature and a subscriber can be informed about the usage of such a feature and how it could benefit them in achieving their personal or business goals. Now, as always happens, Murphy's law could kick in, the best laid plans could go astray and delays could get introduced in all these processes. Essentially, the above process is like hitting a moving target and, depending on how much of a float is planned, the operator could absorb some of the delays. Being part of such a complex ecosystem inheren

tly creates such dependency risks, and all stakeholders need to work closely to be able to minimize big changes in plans as much as possible and keep surprises to a minimum. Of course, safe project planning in terms of planning additional time as a buffer could help mitigate the issues to a certain extent. Standardization Delays Much of LTE development has been rigorously driven by extremely dedicated people working as part of the standardization committees under the aegis of the 3GPP. It must be remembered though that standardization efforts are driven by people who work mostly for different participant companies that are trying to promote the agenda of their employer along with developing the standard. Hence, in spite of the best interests of the people involved, there may be issues with specifications for some standards and delays may be inevitable. It may also be that only when some standards are defined that further issues come up, which necessitates further specification efforts to be resolved. This is especially true for LTE where the standards are being developed for some problems that have not been solved at all until now, like SON, for example. Hence, SON will continue to be refined and be deployable only with LTE-Advanced, and such delays cannot be avoided. Operators must therefore be prudent and keep a wary eye on the standardization progress of important features in order to know well in advance if delays are imminent so that they can pursue alternate solutions wherever possible. Patent Costs If there is one moral that has been reemphasized through all the patent battles in recent times, it is that patents cost money and are really valuable. It is worth noting that companies pursue patents for many reasons. First, patents grant a company an advantage for an innovation that they have implemented and hence should provide avenues to monetize the same through licensing costs. Second, patents are useful to negotiate with a competitor who has a complementary set of patents to be able to cross-license the p

atents and make sure both are able to share advantages in a respectable manner. And third, patents are also used as a corporation weapon to destabilize competition by citing patent violations and ensuring that the competitor is unable to sells its product. This latter reason has been seen being actively played out in the past few years in courts all over the world between Samsung and Apple. Coming back to LTE though, it was recognized very early on that allowing such patents to be granted will make life very difficult for many companies to offer product solutions, and the innovations will not be coming forth as a few companies with specific and crucial patents will stifle innovation, increase costs of the UEs and solutions to unmanageable levels, and in the long term prevent LTE from being adopted, developed, or deployed. So, to control such intellectual property rights (IPR) costs, a few like-minded companies with key patents in LTE, namely Ericsson, NSN, Nokia, NEC, and Nextwave wireless, collectively agreed to minimum royalty payments for their shared patent use in single digits of the handset sale price and to not exceed more than $10 for LTE modems in Netbooks (seehttp://www.ericsson.com/news/1209031). It is also worthwhile to note that key players like Qualcomm are not part of this arrangement. Also, other big players in LTE like Huawei, Broadcom, and Texas Instruments are also not part of the above consortium. It is also clear that there are lots of innovation reports still being filed in all key and crucial areas of LTE, such as line interference management, SON, and handover optimizations, which implies that patent battles are far from over for LTE systems. CHAPTER 3 DEPLOYMENT CHALLENGES IN EVOLVING 4G This should be of much concern to operators as the mobile devices they are planning to roll out may become embroiled in patent wars and may get banned from being sold. Deployments could get delayed, UEs could end up costing more, and consumers may, in the meantime, find other offerings to be more interest

ing. Also, vendors may have to find alternate ways to implement a feature or solution to a problem to avoid patent infringement, and the associated costs could end up delaying the release of a particular model or solution too. MICROSOFT AND THE ANDROID PATENT MONEY TRAIN To understand the cost of patents and the effect this can have on companies, we need go no further than Microsoft, whom some industry watchers project is making more money from Android devices than its own Windows phone division. Analysts estimate that Microsoft must be making nearly $2 billion every financial year, just from its Android licensing agreements s(see http://www.zdnet.com/microsoft-is-making-2bn-a-year on-android-licensing-five-times-more-than-windows-phone-7000022936/).So how exactly did Microsoft end up making so much money out of Android device sales? Microsoft has been claiming violations on its patents by open source Linux software for quite a long time. In 2007, Microsoft made claims about Linux violating close to 235 software patents that were owned by Microsoft (http://archive.fortune.com/magazines/fortune/fortune_archive/2007/05/28/100033867/inde htm?source=yahoo_quote). In the previous year, Microsoft had already signed an interoperability pact with Novell, which though touted at that time to be a deal to promote SUSE Linux, was widely seen by analysts as protection offered by Microsoft to SUSE Linux users from patent violation damages (http://www.microsoft.com/en-us/ news/press/2006/nov06/11-02msnovellpr.aspx). Novell had already been recognized as the owner of UNIX copyright in a case brought out by SCO against Linux vendors (http://www.groklaw.net/pdf2/Novell-846.pdf). So when Google started promoting the Android operating system as open source software, the fact that it is based on a Linux kernel made it vulnerable toward patent infringement suits. In 2011, Microsoft won a key patent decision that Motorola mobility violated Microsoft's patents in making its Android handsets http://www.theguardian.com/technology/2011/dec

article/2180926/software/microsoft-inks-licensing-deals-with-two-more-android-makers.ht In recent times, with the popularity of Android systems, many more Android device makers, including Amazon, have joined the list of companies who have entered into agreements with Microsoft. So, Microsoft's strategy to diversify its software patent portfolio is paying rich dividends to the tune of offsetting the losses it is facing from its own device businesses. More recently, Apple and Samsung are still embattled in further patent violation warfare, which could have more implications for Android device manufacturers in pushing up licensing costs (http://en.wikipedia.org/wiki/Apple_Inc._v._Samsung_Electronics_Co.,_Ltd..). Business Challenges In the previous section, we discussed issues related to LTE as a technology and how different aspects of the solution development could affect deployment and delay plans to roll out. This section will discuss issues related to business planning aspects of LTE deployments, which, though not numerous, are still very important in terms of impacts and in need of a thorough understanding and appreciation. CHAPTER 3 DEPLOYMENT CHALLENGES IN EVOLVING 4G Investment Issues As mentioned previously, LTE was designed from the ground up to provide the next futuristic network with very high data rates and strict targets to achieve QoS and latency issues. Also, there were strict performance criteria specified for various aspects of the solution as seen from the subscribers' perspective. Note These performance criteria are discussed in Chapter 1, with overall guideline requirements for LTE. In order to meet these stringent requirements, newer NEs were introduced in the LTE network architecture, both in the radio network and on the core network side, when compared with 3G. Thus, we have NEs, including UEs that are new, that have much more complex functionality to accomplish. This directly implies a few things. Operators mostly are unable to reuse their current networks and need to invest in building their

core and access networks with the newer NEs. Of course, some vendors are innovatively addressing such concerns by introducing newer 3G network elements that are LTE ready in the sense that they reduce the network deployment costs to just upgrades-for-LTE kind of solutions. Yet, predominantly, operators are required to invest upfront to make and deploy the LTE networks. In times of financial turmoil, as existed in much of 2012 in international markets, this would mean that a majority of these decisions cannot be absorbed by operators because huge investments in such a financial climate may not be feasible. Also, unless operators are of such a size as to have surplus cash available to plan and make such deployments, they are going to be dependent on external funding to enable such deployments. Such borrowings, however, carry periods for interest calculations and need to be factored into the plans for deployment, and they further constrain the timelines by which operators can plan and execute their deployment and to be able to fulfill commercial obligations to the creditors in a timely manner. Coupled with the inherent challenges discussed in the previous section, this really is an arduous task. Average Revenue per User and Return on Investment Periods To make any business plan work, there needs to be elements of cost and modeling of expected revenue to be worked out so that subsequent periods of business can be simulated and some sort of profit projections can be made. In the cellular world, the most common key performance indicator from revenues perspective is average revenue per user (ARPU). In simple terms, this could equate to the revenue generated by an operator divided by the number of users or subscribers supported by the operator. Hence, there is a way to find out how much one user is contributing to the margins of the operator and also determine business health through the service offering. The return of investment (ROI) is the period over which a business needs to be operational to be able to start makin

g enough money to tide over the investments made and hence start seeing profits. Businesses cannot expect to see any real profits until the ROI period is accomplished. The key challenge with LTE network rollouts for the operator will be generating enough revenue from the services to the users to be able to have a high enough ARPU so that the ROI period is low enough to sustain interest of the investors and creditors. This is a pretty difficult task! For the proliferation of flat rate plans for a fixed capacity offering (in terms of gigabytes), you have to take away the ability of operators to specialize their offering as a premium service. It has long been the case that call rates for voice services are being pushed down due to competition and pressure from market regulators linked to the government. Even roaming offerings are no longer able to generate revenue as agencies are trying to minimize the impact on consumers as much as possible. So, operators are forced to innovate and differentiate their offerings and services enough so users are ready to pay a premium price or top dollar for their services. Also, as the focus moves more to the content that users can use, operators need to work with content providers so they don't lose any more relevance as meaningful players who can offer services and make real money in doing SO. CHAPTER 3 DEPLOYMENT CHALLENGES IN EVOLVING 4G The recent trends for major telecommunication providers indicate the following concerns in profitability as seen in operating the business landscape: Earnings indicators like EBITDA have been falling for the past two years. Returns on capital employed are also showing a downward trend. Voice revenues continue to fall even while mobile broadband picks up pace. The Changing Marketplace As discussed previously, the investments being made and the services being offered with some specific targets in mind can be made by the operator, keeping in mind certain assumptions about user behavior and applications that are targeted. In the event that the marke

t produces some stunning innovations, this could potentially challenge and turn the plans made by the operator upside down. For example, not many could have anticipated that the Android operating system would revolutionize the market for UEs in the way it has. Also, it brought a paradigm shift where users are ready to pay money for over the top (OTT) applications that are not necessarily basic services that work as a backbone. The longer a plan takes to roll out, the more chances that some disruptive innovation or application can change the market for better or for worse. In such an environment, operators are advised to target routes to better profitability in the following ways: Provide differentiated service and pricing models Partner more with OTT application providers Improve network efficiency significantly A Survey of LTE Deployments Around the World Having seen the challenges that exist for LTE deployments, let's assess the ways network operators around the world have gone about their deployments. Central to this analysis is a categorization of deployment strategies into single radio access network (RAN) VS. network overlay. In single RAN strategy, the network gets upgraded to support multimode, multistandard base stations that can help the operator deploy future proof technologies together. The following characteristics apply to the single RAN: Base stations deployed have multistandard capabilities, including 2G, 3G, along with LTE. In some cases the radio also has multimode support to work in TDD or FDD mode. There are reduced operational costs for managing the whole network. There are reduced site costs as converged nodes are deployed in the network. Overall upgrade of the networks may take time and also prove to be costly with the possibility of service disruption to existing subscribers. For LTE network overlay strategy, the LTE networks are deployed and operated side by side with the existing 2G or 3G network, without doing an upgrade. The following characteristics apply to the LTE network overlay: N

ew base stations are deployed for the LTE network. No immediate plans are made for upgrading 2G or 3G network base stations. There are reduced upgrade times and investment requirements. CHAPTER 3 DEPLOYMENT CHALLENGES IN EVOLVING 4G There is a need for separate systems to monitor existing 2G or 3G networks and new LTE networks. Possible upgrades are needed in the future depending on subscriber and services growth and the need for LTE expansion. In deciding on a particular strategy, network operators will have to consider the constraints described in the previous sections in terms of spectrum availability, existing network investments and services, investment, and ROI roadmaps. As per the data available in a majority of cases, network operators have preferred the strategy of network overlay for their deployment. In the following sections, we discuss the overall strategy with specific highlights about some countries' LTE deployment. South Korea In South Korea we looked at how SK Telecom, KT, and LG U+ went about their LTE deployments. SK Telecom adopted LTE network overlay as its strategy along with its existing 2G and 3G without going for an upgrade for the same. Its primary goal for the deployment was to meet the mobile broadband needs of its consumers. SK Telecom deployed its LTE services in 1.8Ghz with a bandwidth of 20Mhz. KT also adopted LTE network overlay to deploy LTE on the 800Mhz spectrum. It will also be able to utilize the 1.8Ghz spectrum it has after being allowed to discontinue the 2G services on the same. LG U+ went for a single RAN strategy using its existing CDMA infrastructure and deployed its LTE services using 2.1 Ghz spectrum. It did not have a 3G service operational and hence the strategy was probably easier to implement. Most Korean operators are looking to augment their LTE macro networks with small cell deployments to fill the gaps in residential areas and to continue to provide good coverage and services to its subscribers. Japan In Japan we looked at how NTT Docomo, KDDI, and SoftBank Mo

bile went about their LTE deployments. NTT Docomo adopted the LTE network overlay strategy for its LTE services deployment in the 2.1 Ghz spectrum. KDDI also adopted the LTE network overlay strategy for its LTE services in the 800Mhz spectrum band. KDDI is also planning to deploy in 1.5Ghz spectrum for services covering urban and suburban areas. SoftBank Mobile also adopted the LTE network overlay for deployment of LTE services in the 2.1 Ghz spectrum in FDD mode to coexist along with its high-speed packet access-plus services in the 900Mhz spectrum. SoftBank Mobile also plans for LTE TDD deployment in the 2.5Ghz spectrum. Australia In Australia we looked at how Telstra and Optus went about their LTE deployments. Telstra opted for LTE network overlay strategy to ensure LTE service deployment did not affect its existing 2G and 3G network operations. Telstra targeted its existing 3G subscribers for the LTE services by offering dual-mode LTE/ HSPA+ dongles. Telstra deployed LTE on the 1.8Ghz spectrum and plans to selectively refarm the 900MHz too. The network operator Optus also used a network overlay strategy to introduce LTE services in refarmed 1.8GHz 2G spectrum. Optus acquired WiMax operator Vivid wireless to be able to deploy TD-LTE network with 98MHz bandwidth in the 2.3GHz spectrum. For the TD-LTE deployment Optus mostly needs to use a single RAN strategy. United States In the United States, we looked at network operators Verizon, MetroPCS, AT&T, and Sprint Nextel for their LTE deployment strategies. CHAPTER 3 DEPLOYMENT CHALLENGES IN EVOLVING 4G Verizon wireless used a network overlay strategy for deploying its LTE network, for a faster and less expensive deployment, along with its existing 2G and 3G services in the 1GHz spectrum. Verizon also plans for a 700MHz LTE deployment with 10MHz bandwidth. The operator continues to upgrade its subscribers to its LTE network with continuing growth and expansion plans. MetroPCS also used a network overlay strategy to deploy LTE services on top of its existing 2G CDMA

services, totally skipping 3G. MetroPCS also worked on upgrading existing CDMA subscribers to a less costly enhanced voice data optimized Rev-A network to enable faster speeds without needing a costly LTE handset upgrade for the same. In the meantime, T-Mobile USA was upgrading its 3G infrastructure to offer 1Mbp speeds with HSPA+ service. T-Mobile USA was also acquiring AWS spectrum to be able to deploy LTE with its own network overlay strategy. But it also acquired MetroPCS to consolidate and become a leading provider. AT&T also deployed LTE in a network overlay strategy, along with its HSPA+ services using 700MHz spectrum and also AWS spectrum. Interestingly, AT&T also operated a wide network of Wi-Fi hotspots. Sprint is the only U.S. operator that deployed LTE with a single RAN strategy to enable it to upgrade its existing 2G and 3G networks. Sprint plans to invest in this strategy to improve on coverage, capacity, and overall data speeds. Sprint plans to optimally use the 5MHz of paired FDD spectrum in 1900MHz and needs to make long-term plans for the investments needed and to secure financial planning with a buy in from SoftBank of Japan. Also Sprint strategically acquired Clearwire, with its 4G network having 160MHz bandwidth in the 2.5GHz spectrum, to deploy a TD-LTE network. Traffic Profiles and Other Evolution Challenges When the operators plan and deploy solutions, they make assumptions about the kind of traffic that different subscribers want, the services that subscribers will be using, the bit rates that will be desired, and the number of devices that will be accessing the network. Note For a more rigorous treatment of these, refer to Chapter 1: parameters to be taken into account when dimensioning the network. Such parameters are also used for dimensioning the network to ensure QoS for all the service users and different applications. However, it is easy to see that the telecom market dynamics keep changing all the time. We see changes in subscriber patterns, devices usage, and many other aspects.

This section initially takes a look at some trends that can help us understand this dynamic nature, before we try to analyze the future trends to take into account. Recent Trends in Telecom Customer Profiles Evolution Usage of Dongles VS. Smartphones Originally, subscribers who looked for connectivity to continue to have access to their work and hence needed mobile broadband access with their laptops purchased dongles instead of a smartphone, which was a device that helped one get e-mail and other important applications like messaging. Now comes the interesting part. Depending on how the smartphone works in terms of idle time handling, smartphones could end up generating a lot of signaling traffic in comparison to dongles. One such study by NSN (albeit in 3G networks) found that even though dongles accounted for 60% of the data traffic load, they generated only 1% of the signaling load. This is due to the fact that the dongles were getting connected and staying connected, whereas smartphones were aggressively optimizing power management and hence were becoming a new state called fast dormancy handling, where they were switching into RRC idle states very quickly. CHAPTER 3 DEPLOYMENT CHALLENGES IN EVOLVING 4G Another key aspect of the dongle tends to be that most of them support only data connections VS. smartphones that have to support traditional customer service-based services of voice, text messages, and data connections. Hence, dongle implementations tend to be simpler as the complexity of connection management and other settings is managed by the connection management software in the laptop. This remains true even in cases of dongles that support voice. Hence, dongles cost a lot less than the equivalent smartphones. By keeping track of a variety of devices getting connected in the network, the operator can continue to remain alert about trends of subscriber usage. For example, the following observations were made on Vodacom LTE networks in 2011(http://mybroadband.co.za/news/broadband/39919-data-usage-smartp

hones-vs-dongles.html): The average data consumed by dongles grew only by 10% year over year, whereas data consumed by smartphones saw a 100% increase. The ratio of dongles to smartphones changed from roughly 80:20 to 65:35 by the end of the year. Vodacom had a total of 1.1 million dongles and 4.1 million active smartphone subscribers by the end of the year. Smartphones are definitely going to continue to grow over dongles, but they will still be affected by price and data plans support provided by operators. Further studies in consumer behavior and sentiment (http://www.marketingmagazine.co.uk/ article/1054529/research-charts-death-dongle-smartphone), give credence to dongles being preferred by business savvy subscribers only. In this survey made in the United Kingdom in 2011, the number of respondents who preferred a dongle fell drastically from 20% in the previous year to 7%. At the same time, smartphone penetration keeps increasing, with 35% owning a smartphone in 2009, and doubled the same amount in the previous year. Also, 60% of the respondents indicate that they would be buying a smartphone. Trends in Data VS. Voice Usage Initial LTE networks were deployed with data-only plans and used existing 2G or 3G networks for voice support. Currently though, VoLTE has been specified and also implemented by many vendors as part of their solutions. It is interesting to note that with additional smartphones getting introduced into the market, the usage of traditional voice keeps coming down in comparison to data usage in same periods. One study of Finnish iPhone users shows data over a 10-month period in 2011-2012 (http://www.forbes.com/sites/terokuittinen/2012/10/15/ as-iphone-mobile-data-usage-soars-voice-calls-dive/), where data usage has seen a spike of over 68% per subscriber, as shown in Figure 3-1. Average data MB per subscription per month 2,000 1,500 1,000 Figure 3-1. Average date per month CHAPTER 3 DEPLOYMENT CHALLENGES IN EVOLVING 4G Also, in the same period the average voice minutes per subscriber dropped

by over 13%. It was observed that traditional text messaging too declined by the same percentage (Figure 3-2). Average voice minutes per subscription per month Figure 3-2. Average voice minutes It has to be noted that this study is based on iPhone usage patterns alone. Also, the trend of similar data usage is being observed across high-end smartphones in Finland, the United States, and the United Kingdom, independent of network operators and across different networks. It does appear that newer smartphones, which allow higher data rates to be enabled, coupled with newer OTT applications make subscriber data usage trends more uniform. Elsewhere, as illustrated in another study, the trend is that data are becoming the premium service along with decreasing voice spend(http://www.worldwideworx.com/mobile-data/). Average user cell phone spending has increased from 8% to 12% over an 18-monthperiod.Andvoicespendduring the same period went down from 77% to 73%. The demography of mobile subscribers also plays a crucial role in a trend of decline in voice usage. This is shown by a recent survey of mobile data usage by teens in the United States http://www.nielsen.com/us/en/ newswire/2011/new-mobile-obsession-u-s-teens-triple-data-usage.html). Average mobile data usage has increased by 256% for teens and shows a definite increase across all age groups in general, as shown in Figure 3-3. Monthly Data Usage by Age (MB) Q3, 2010 vs. Q3, 2011 % increase +256% +147% +118% +133% +126% +133% 13-17 18-24 25-34 35-44 45-54 55-64 Q3, 2010 Q3, 2011 Source Nielsen nielsen Figure 3-3. Monthly data usage by age CHAPTER 3 DEPLOYMENT CHALLENGES IN EVOLVING 4G Not surprisingly, voice usage went down for the same group by 17%, from an average of 685 minutes to minutes. Most of the respondents attributed this to messaging being "faster, easier and being more fun." Growth of Internet and Smartphone Usage Across the World As a running theme in the telecom landscape, consumer growth patterns and traffic growth continue to be driven strongly by a

ccess to the Internet. Mobile broadband services and access of the same through smartphones continue to be critical profitability factors for operators. The proportion of Internet users across the world gives us a good idea about where the penetration is already at a high level. Markets that have high Internet penetration tend to be more mature in their use of the Internet. Also, they have a higher demand for advanced value-added services in comparison to more developing areas, where primary access to the Internet is still developing. The statistics presented in Figure 3-4 from Internet World Stats provide a good picture of the proportion of Internet users across the world by the end of 2013. Internet Users in the World Distribution by World Regions - 2013 Q4 11.4% 10.8% Asia 45.1% Europe 20.2% 20.2% North America 10.7% Lat Am / Caribb 10.8% 45.1% Africa 8.6% Middle East 3.7% Oceania / Australia 0.9% Source: Internet World Stats - www.internetworldstats.com/stats.htm Basis: 2,802,478,934 Internet users on Dec 31, 2013 Copyright © 2014, Miniwatts Marketing Group Figure 3-4. Internet use by the end of 2013 It is clear from Figure 3-4 that bigger demographies have a larger presence in the Internet. Asia dominates as a highly populated region and hence it is more interesting to see the proportion of population that has access to the Internet. As illustrated in Figure 3-5, from the same period we see a worldwide total penetration of Internet at only 39%. CHAPTER 3 DEPLOYMENT CHALLENGES IN EVOLVING 4G World Internet Penetration Rates by Geographic Regions - 2013 Q4 North America 84.9% Europe 68.6% Australia / 67.5% Oceania Latin America / 49.3% Caribbean Middle East 44.9% World, Avg. 39.0% 31.7% Africa 21.3% Penetration Rate Source: Internet World Stats - www.internetworldststs.com/stats.htm Penetration Rates are based on a world population of 7,181,858,619 and 2,802,478,934 estimated Internet users on December 31, 2013. Copyright©2014, Miniwatts Marketing Group Figure 3-5. Worldwide penetration of the Internet Also, t

he top 20 countries that use the Internet contribute to 70% of the total users, and further Internet penetration among the population gives us an indication of the maturity of the market in that country and the potential for differentiated services(http://www.internetworldstats.com/top20.htm).From the same data, it is very clear that some countries like India and China have had phenomenal growth rates from 2000 to 2013; yet their penetration rates are still low, indicating much scope for further growth potential. Going further, as per Figure 3-6 from the IDC, as of 2012, 53% of smartphone consumers over the world access the Internet through their smartphones. Combining the Internet penetration with the smartphone access gives a very good indicator of the potential for smartphone-based mobile broadband access across different countries. This should be a critical part of operator assessments for business potential in that region. CHAPTER 3 DEPLOYMENT CHALLENGES IN EVOLVING 4G Figure 3-6. Percentage of smartphone owners using Internet on phone daily Furthermore, figures from Google Analytics for overall mobile traffic access numbers as a percentage of overall web traffic numbers give very encouraging inputs for operators: Total mobile traffic was up 139% on average from Q4 2011 to Q4 2012. Eighteen percent of total web traffic was coming from mobile devices by Q4 2012. Mobile traffic as a percentage of total website traffic nearly doubled from 10% to 18% from Q4 2011 to Q4 2012. In fact, as per the projections from Morgan Stanley, mobile Internet users are poised to overtake desktop Internet users as of end of 2014, as shown in Figure 3-7. Mobile Users > Desktop Internet Users Within 5 Years Global Mobile vs. Desktop Internet User Projection, 2007 - 2015E 2,000 1,600 1,200 Mobile Internet Users Desktop Internet Users 2007E 2008E 2009E 2010E 2011E 2012E 2013E 2014E 2015E Morgan Stanley Morgan Stanley Figure 3-7. Mobile Internet users are poised to overtake desktop Internet user CHAPTER 3 DEPLOYMENT CHALLENGES IN EVOL

VING 4G More interestingly, the growth rates of early desktop Internet access in the 1990s (as measured with AOL and Netscape adoption) in comparison with mobile usage growth in the 2000s (as measured primarily by following Apple devices: iPhone/iTouch) leads to the following observations. Adoption of Apple devices is almost 11 times faster than AOL and several times faster than Netscape. The 3G services act as an inflection point in helping enable access to more than 20% of cellular users to help accelerate mobile Internet access. This proves to be inspirational for LTE operators as LTE services could serve to take mobile Internet usage acceleration to the next level. This is especially true with the increasing popularity of services that demand higher bandwidths, like video, which NSN forecasts will account for 60% of all mobile traffic in 2013. The IDC Predications 2014 (http://www.idc.com/research/Predictions14/index.jsp)furthercement some of the observations made in the earlier sections: Smartphone and tablet sales are outgunning PC sales by nearly 2.5 times or 250%. Apple continues to see strong iPhone and iPad sales to retain a value advantage of nearly double its competition, but it still is outnumbered by Android devices with a volume of one- third or a ratio of 1:3. Google Play Store catches up with the App Store to narrow down revenue differences. Cloud spending is expected to surge to nearly $100 billion, for a growth rate of 25%. Data volumes are expected to reach 6 trillion terabytes, pushing the spending on big data analytics by 30%. The Internet of Things (IoT) remains very promising, with 30 billion endpoints and $8.9 trillion in revenue expected by 2020. The ITU-T ICT facts and figures for 2014(http://www.itu.int/en/ITU-D/Statistics/Pages/facts/default.aspx) add some very interesting insights into some of the items we have been discussing in the preceding section. Some of these include: There are 2.3 billion mobile broadband subscriptions, with developing countries contributing up to 55%. Mobile

broadband is growing at double the rate in developing countries as compared to developed countries. Mobile broadband penetration rate is expected to reach a global value of 32% by the end of 2014, with developing countries still only being 21% covered VS. 84% in developed countries. The penetration rate is almost double that of 2011. Africa leads in mobile broadband growth from 2% in 2010 and to 20% in 2014. Mobile subscriptions will reach almost 7 billion by the end of 2014, with a penetration rate of 96%, with more than half of these coming from Asia-Pacific region. Mobile cellular growth rates are at the lowest global level of 2.6%, which indicates saturation in the markets globally, with developing countries still expanding twice as much as developed countries. Fixed broadband growth is slowing down in developing countries. By the end of 2014, nearly 3 billion people (40% of world population) will be using the Internet, with 66% of the same hailing from developing countries. While 44% of world households will have an Internet connection, only 10% of households in Africa enjoy this technology. The percentage of fixed broadband users who enjoy more than 10Mbit/s speeds still remains lower in developing countries when compared to developed countries. CHAPTER 3 DEPLOYMENT CHALLENGES IN EVOLVING 4G All of the above indicators imply that there is still good potential for enabling high-speed mobile broadband access and applications that exploit the same. For more detailed analysis of traffic profiles and dimensioning inputs, see Chapter 1. Future Traffic Profiles and Trends Having seen how the Internet and mobile Internet devices are all shaping up, it is interesting to see what sort of trends are seen in the near horizon of five to seven years and what changes will be happening in that same timeframe. is very important for operators to understand that all of these trends represent real opportunities for growth and profitability and hence need to be factored in when planning for 4G network, and beyond, deployment.

In the following sections, we list the major trends as forecasted from the following listed reports: Vision 2020 whitepaper final from NSN (http://networks.nokia.com/file/26156/ echnology-vision-2020-white-paper) Traffic and Market report June 2012 from Ericsson (http://www.ericsson.com/res/ docs/2012/traffic_and_market_report_june_2012.pdf) IDC predictions 2013-competing on the 3rd platform (http://www.idc.com/research/ Predictions13/downloadable/238044.pdf) Internet Trends by Mary Meeker, Morgan Stanley April 12, 2010 (http://www.slideshare. het/malaparte/morgan-stanley-internet-trends-mary-meeker-2010041 Cisco Visual Networking Index: Forecast and Methodology, 2012-2017 //www.slideshare.net/andrewwilliamsjr/cisco-visual-networking-index- forecasting-and-methodology-2012-2017) McKinsey-MGI_Disruptive_technologies_Executive_summary_May2013 (http://www.mckinsey.com/~/media/McKinsey/dotcom/Insights%20and%20pubs/MGI/ Research/Technology%20and%20Innovation/Disruptive%20technologies/MGI_Disruptive_ echnologies_Executive_summary_May2013.ashx) MGI_IT_enabled_trends_Report_May 2013_v2(http://www.mckinsey.com/insights/ high_tech_telecoms_internet/~/media/mckinsey/dotcom/insights/high%20tech%20 telecoms%20internet/ten%20it-enabled%20business%20trends%20for%20the%20decade%20 ahead/mgi_it_enabled_trends_report_may%202013_v2.ashx) One thing that is also very clear is that in the following sections, we are trying to do some crystal ball gazing to really delve into what the future holds in terms of areas of interest and impact to the telecom operators, especially given that 4G deployments are going to have much more momentum. While crystal ball gazing is good for getting a future outlook, one should also be clear upfront that our projections may not come true or may come true only to a partial extent. Also, it is important to note that in the telecom industry, as in other sectors, it is not easy to estimate the impact of certain technologies from the buzz existing about the same technology in Internet, social blogging, and oth

er technology-related news sites. Clearly, mobile Internet is generating enough revenue currently and has impacted many businesses already by enabling online customers through mobile devices. Other key technologies have enough potential and impact for the 4G operator to consider, such as mobile Internet, Cloud technology, the IoT, and automation of knowledge work. CHAPTER 3 DEPLOYMENT CHALLENGES IN EVOLVING 4G The Internet of Things The term Internet of Things was originally coined by Kavin Ashton as part of a presentation to Procter and Gamble (http://www.rfidjournal.com/articles/view?4986).What started originally as a method of managing things and devices with radio-frequency identification (RFID) enablement has now grown to cover a multitude of things as and more analysts have joined the bandwagon. As of now, the Internet of Things is meant to include everyday objects that have the following properties: objects can be read, located, addressed, and controlled via the Internet. The technologies used for accessing the Internet could be through RFID, WLAN, wired networks, and any other methods, which could encompass things like ultra light SIMs for other radio access technologies like 3G or LTE, which could be embedded into the devices. Some recent research from McKinsey even projects the Internet of Things to have morphed into what they call the Internet of All Things, emphasizing the nature of the trend itself, to encompass all things that come onto the Internet, including items that we would normally not associate with the Internet, like clothing, food, and even shelter. The paper defines the Internet of All things, as follows: Linking machinery, equipment, and other physical assets with networked sensors and actuators to capture data and manage performance, enabling machines to collaborate and even act on new information independently. The following application areas are seen as having major potential for the IoT: Remote monitoring of assets, systems, and people Performing preventive maintenance and improving

systems management with data collected in real time Autonomous optimization of systems with complex closed-loop decision making Health applications where people can get a quantified assessment of their overall health, well- being, and monitoring Additional areas that are still being explored for more impact in applications include: Sensor networks made up of distributed sensors, which are still evolving, some as small as smart dust that could help collect more information about buildings, vehicles, and other places that need to have Internet access and applications to use the same Ubiquitous positioning applications to locate and use objects residing indoors or outdoors or even underground, where reach through normal signaling may be inadequate Biometrics applications, where systems and things could identify people using their biometric footprint like fingerprints, facial scans, iris scans, etc. Machine vision using objects enabled with cameras that could interwork with applications like Augmented Vision, to give more context-based information. For example, Google Glasses is one such application. Where all these applications lie in relation to IoT evolution is more evident in the roadmap shown in Figure 3-8. As per the current state of practice, we have barely scratched the surface of IoT applications, and a lot more improvements and capabilities are planned for the coming decades. CHAPTER 3 DEPLOYMENT CHALLENGES IN EVOLVING 4G TECHNOLOGY ROADMAP: THE INTERNET OF THINGS Software agents and advanced sensor fusion Technology Reach Miniaturization, power- efficient electronics, and available spectrum Teleoperation and telepresence: Ability to monitor and control distant objects Ability of devices located Physical-World indoors to receive geolocation signals Locating people and everyday objects Cost reduction leading Ubiquitous Positioning to diffusion into 2nd wave of applications Surveillance, security, healthcare, transport, food safety, document Demand for expedited management logistics Vertical-Market Applicatio

ns RFID tags for facilitating routing, inventorying, and loss prevention Supply-Chain Helpers Source: SRI Consulting Business Intelligence Figure 3-8. Roadmap of the future of the Internet of Things As can be seen in Figure 3-8, Supply Chain management and inventory applications have already started using IoT to manage their products using RFID-based solutions. What is interesting in this roadmap is that positioning applications are about to get more traction toward the middle of this decade. But the real growth of IoT will come when teleoperation and the telepresence of objects become enabled as it would mean decisions on the following: How the objects would access the Internet: Would they directly access the Internet or would they talk to a local information collector that would bridge the Internet? What would be the power consumption of the objects? What would be the form factor of such objects? What kind of data would need to be transmitted and received? How often would such data transmission need to happen? Would the objects communicate directly to other objects or people also? What would be the protocol followed in such communications? Would there be only data communicated or some form of voice communication also posited? What would be the number of such objects on the network? What is the cost of failure to the objects and to the applications? CHAPTER 3 DEPLOYMENT CHALLENGES IN EVOLVING 4G We will address these questions later in this section in the context of their impact on 4G for such Internet of Things. Also, we will define the traffic profile for the IoT in that same discussion. The following section will focus more on the top impediments to the IoT as of now so you get an idea as to the state of practice of the IoT. Top Challenges for Internet of Things Standardization Standardization has been identified as one of the key challenges facing a larger adoption of the IoT by a ITU-T report on this in 2007, and it is still one the major concerns in further development of the IoT. Even a recent consortium

on the IoT held in 2013 observed that standardization in Cloud access technologies could be the major impairment for the IoT. In this regard, a 2012 report, from the European Research Cluster on IoT, identified the following aspects for architecture development: Structure reference frameworks for logical hardware and software components still need to be developed and refined. Questions of object identification, virtualization, and decentralization still need to be finalized. The interoperability of the IoT across application sectors needs to be ensured. Security and Privacy As with everything else related to the Internet, issues about security and privacy dominate the IoT. With the number of devices that will be getting on to the Internet directly or indirectly, there are key questions about security aspects that need to be addressed. Risks remain about vulnerabilities related to security issues that could lead to annoyances and also losses of an economic nature to the businesses involved. Also, one can only imagine what havoc can be wrecked by those who could take control of such devices via the Internet and could harm or impair many systems by creating problems that range from household issues to complex industrial shutdowns. As with sensitive information collected by social networks today, questions remain about privacy concerns for the huge amount of data that would be gathered, collected, analyzed, and disseminated to a wide variety of manual and automated systems. With the prospect that additional devices could in turn be using the data collected in other ways, the complexity of the problems and rules that govern gain more importance. Also, different countries and businesses could have different laws regarding privacy of data, which also may need to be aligned and adhered to in making the IoT applications work together safely. Lifecycle processes that cover the data also need to be defined and adhered to. Interoperability Testing Standardization efforts are further needed to ensure that all the systems part

icipating in the IoT remain compatible to one another. Interoperability testing is seen to be pretty challenging in such environments. By the nature of the IoT, objects and systems are meant to identify one another and operate together in increasingly manifold ways, placing a larger burden on testing efforts. Large-scale testing and pilots to confirm interoperability and further improvements to standards are expected to be crucial and essential in making things simpler. Environmental Concerns There are jokes doing the rounds that the latest touch-based smartphones with the biggest screens around are not so smart after all, as sometimes they run out of battery and hence remain "dumber than normal phones." Energy requirements and management would play key roles in future electronic devices. With the number of devices that may end up online on the Internet and power consumption requirements for these, there are valid questions that need to be answered about how energy impacts due to the IoT will be managed. CHAPTER 3 DEPLOYMENT CHALLENGES IN EVOLVING 4G Also, electronics have a way of becoming outdated very easily. The industry itself seems to be fueling this rapid replacement of current models with newer ones, leading to tons of e-waste being generated around the cities of the modern world. It is expected that for whatever reasons (faulty devices, lower lifespan, becoming outdated), the IoT could end up adding to the e-waste generation and would hence need to be managed more actively and responsibly with the help of businesses and governments. 4G and the Internet of Things The Case for LTE-Advanced and loT McKinsey estimates that there has been over a 300% increase in machine-to-machine device connections over the past five years. Clearly, there is an increasing trend of objects getting on to the Internet and interconnecting with other devices on the Internet. LTE as the leading 4G technology is primed to enable the IoT in the following ways: Nearly a trillion things could be connected to the Internet and more than

100 million machine-to-machine connections could be present. LTE networks are able to address so many devices due to close interworking with IPv6 for standardization. Billions of connected subscribers could be generating as much as 1GB data every day, and LTE networks are primed to be able to carry such high data volumes. LTE networks, with their very low latency targets, should be able to meet with real-time data delivery requirements of the IoT applications. LTE and Small Devices It is expected that objects and devices that want to get on to the Internet would need cost-effective solutions that would enable rapid adoption and deployment of LTE-based devices. NSN believes that LTE has great potential to enable all objects and networked things to come on to the Internet and manage the data explosion that would ensue. Earlier we asked the question about how the objects would access the Internet. For the sake of our discussions here, we'll restrict ourselves to the following model for devices and objects to come on to the IoT, as it is more complex and more interesting. The objects or devices shall have embedded communication capability to implement one or more of the wireless access technologies, namely the LTE access technology. Customer applications that work on the IoT that are to be implemented on devices that control the data and format in which it must be sent to the Internet. These models would imply that there are a lot of innovations waiting to happen in such a market where there is a huge demand for simplified "system on chips," which could be used to design and add IoT capability to any normal device or equipment discussed above. For example, SEQUANS communications has a brand of product lines based on Streamlite LTE (http://www.sequans.com/products-solutions/streamlitelte/) As seen in the following quote, the product is designed to be used with lots of daily devices: Firefly is designed for connection to CE and M2M devices such as digital cameras, e-readers, security cameras, vending machines, digital

signage, power meters, and health monitoring devices. It is also perfectly suited for integration into surface-mountable modules. CHAPTER 3 DEPLOYMENT CHALLENGES IN EVOLVING 4G Some of the technologies that have been discussed in this chapter are concerned with power consumption, and having a small footprint for the device can be seen in some excerpted product highlights such as: Throughput up to 50Mbps DL/25 Mbps UL 40 nm CMOS technology 4G-EZ ultra-low power technology High-performance LTE RFIC supplied by Fujitsu Semiconductor Dual RF ports The Fujitsu MB86L13A, which the Firefly solution uses, is itself marketed as an IoT ready LTE-optimized transceiver, and though it is primarily targeted for LTE mobile handsets, the vendor does claim support for additional devices. These devices, with the key use being a mobile Internet device, include dual RF ports, mobile phones, cards, and embedded modules. It is to be noted here as elsewhere that LTE as a standard has evolved to better cater to trends like the IoT for better interworking and optimization, much of which is part of LTE-Advanced specifications (Rel 10 and above), which is beyond the focus of this book. Also, it can be expected that there will be plenty of innovation in integration of IoT with LTE networks. The following section gives some pointers to aspects of the LTE RAN, which could be tuned in anticipation of devices that participate in IoT: Data usage and frequency: IoT devices that connect to the Internet directly could work on the following two modes: The devices could be more like sensors that need to send small chunks of data in real time in periodic intervals: Traffic profiles: In such scenarios, the traffic and the bearers in uplink alone could be modeled after VoLTE bearers that also have similar characteristics. Data dimensioning needs to be done for the size of the packets and the rates of the bearers assigned to ensure QoS is satisfied. Latency requirements need to be checked from the point of view of the application, and this may also dete