MULTIPLE BASE SITE COVERAGE WITH OVERLAPPING AND OVERLAYING WIRELESS ARCHITECTURES by EDWARD SHIH-CHIA CHAO B.Sc, Queen's University, Canada, 1996 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF APPLIED SCIENCE in THE FACULTY OF GRADUATE STUDIES DEPARTMENT OF ELECTRICAL AND COMPUTER ENGINEERING We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA January 1998 ©Edward S. Chao, 1998 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department The University of British Columbia Vancouver, Canada Date DE-6 (2/88) ABSTRACT The demand for cellular radio and personal communications services (PCS) con-tinues to increase. Expectations for third generation wireless systems include higher data rates, improved quality of service, and ubiquitous coverage and access. Microcells with low base site transceiver antenna heights and small coverage areas are proposed to increase capac-ity. Microcellular architectures increase the frequency of handoffs and areas of low radio signal levels (radio "holes"). Accordingly, we propose multiple base site coverage with either extended overlapping microcells or macrocells overlaying microcells to reduce handoff rates and to enhance coverage and capacity. Analytical models accounting for call overflow to alternate base sites, increased co-channel interference, and radio holes are developed to evalu-ate system performance. The considered scenarios include an isolated group of microcells, a contiguous layer of microcells, and macrocell overlay with and without macrocell frequency reuse. Performance measures for assessment of the architectures include blocking probability, maximum supportable arrival rate, dropped call probability, handoff activity and carried traf-fic. In the absence of radio holes and co-channel interference, overlapping and over-laying architectures reduce call blocking and increase carried traffic. With co-channel inter-ference and adequate signal levels everywhere, these performance advantages are much diminished. Performance benefits of overlapping and overlay architectures are substantial, even with co-channel interference when radio holes are present. Handoff activity which is also evaluated is not much affected by architectural variations. ii TABLE OF CONTENTS ABSTRACT ii LIST OF TABLES vi LIST OF FIGURES vii ACKNOWLEDGEMENTS x CHAPTER 1: Introduction 1 1.1 Cellular Radio Communications 1 1.2 Motivation 2 1.3 Objectives 4 1.4 Outline of the Thesis 5 CHAPTER 2: System Issues 6 2.1 Cellular Engineering and Channel Reuse Efficiency 6 2.2 Wireless Growth Scenarios for Capacity Enhancements 7 2.3 Mobility Model 9 CHAPTER 3: Overlapping Microcells 11 3.1 Model for Overlapping Coverage Areas 11 3.2 System Description 13 3.2.1 State Characterization 13 3.2.2 Steady-State Distribution 16 3.3 Performance Measures for Overlapping Coverage 17 3.3.1 Blocking and Handoff Failure Probability 17 3.3.2 Dropped Call Probability 18 3.3.3 Handoff Activity 19 3.3.4 Carried Traffic 20 3.4 One Isolated Group of Overlapping Microcells 20 3.4.1 Numerical results 20 i i i 3.5 Contiguous Layer of Microcells 23 3.5.1 Contiguous Layer Growth (Options) 25 3.5.2 Numerical Results for Scenario I 26 3.5.3 Numerical Results for Scenario II 29 CHAPTER 4: Overlaying Microcells and Macrocells 32 4.1 Resource (Spectrum) Sharing/Partitioning 32 4.2 Call Admission and Handoff Strategies 34 4.3 System Description 35 4.4 System Characterization 37 4.5 Analytical Model of Macrocell/Microcell Overlay System 39 4.5.1 Microcell Level 39 4.5.2 Macrocell Level 41 4.6 Overlay Performance Measures 42 4.6.1 Effective Blocking Probability 42 4.6.2 Effective Dropped Call Probability 43 4.6.3 Effective Handoff Activity 46 4.6.4 Carried Traffic 49 CHAPTER 5: Results of Microcell and Macrocell Overlay 50 5.1 Analysis without Macrocell Frequency Reuse 50 5.1.1 Channel Assignment 50 5.1.2 Numerical Results 51 5.2 Analysis with Macrocell Reuse 57 5.2.1 Channel Assignment 57 5.1.2 Numerical Results 58 5.3 Summary 62 CHAPTER 6: Overlapping and Overlaying with Radio Coverage Holes 63 6.1 Microcell Propagation Environment 63 iv 6.2 Radio Coverage Holes in Overlapping Microcell Systems 65 6.2.1 Probability of Blocking (Channel and Signal) 67 6.2.2 Handoff Activity 72 6.3 Radio Coverage Holes in Overlaying Microcells and Macrocells 75 6.3.1 Probability of Blocking (Channel and Signal) 77 6.3.2 Handoff Activity 79 CHAPTER 7: Summary and Conclusions 81 7.1 Overlapping and Overlaying Architectures Compared 81 7.1.1 Blocking Probabilities and Maximum Supportable Arrival Rates . . 81 7.1.2 Comparison of Handoff Activity 83 7.2 Topics for Future Investigation 85 REFERENCES 86 APPENDIX A: Handoff Activity Simplification 89 APPENDIX B: List of Abbreviations and Acronyms 90 APPENDIX C: Frequency Reuse Derivation for Contiguous Microcells 91 LIST OF TABLES Table 3.1: Scenario I Cluster Characteristics 27 Table 3.2: Scenario I Numerical Results for GOS 2% 29 Table 3.3: Scenario II Cluster Characteristics 30 Table 3.4: Scenario II Numerical Results for GOS 2% 31 Table 5.1: Performance of P^Pdc a n d ^ Versus Channel Partitions and System Design Constraints 54 Table 6.1: Signal Outage Probability Versus Path Loss 65 Table 7.1: Blocking Probability and Maximum Supportable Arrival Rate Comparisons 82 Table 7.2: Handoff Activity Comparisons 84 vi LIST OF FIGURES Figure 2.1: Co-channel Interference with Six Interferers 6 Figure 2.2: Wireless Growth Scenerios for Evolving Cellular Networks 8 Figure 3.1: Overlapping Microcells 12 Figure 3.2: Handoff Priority Markov Chain 16 Figure 3.3: Probability of Blocking of One Isolated Group of Overlapping Microcells . 21 Figure 3.4: Probability of Dropped Call of One Isolated Group of Overlapping Micro-cells 22 Figure 3.5: Handoff Activity of One Isolated Group of Overlapping Microcells 22 Figure 3.6: Carried Traffic of One Isolated Group of Overlapping Microcells 23 Figure 3.7: Contiguous Microcell Growth Scenarios 26 Figure 3.8: Probability of Blocking for a Contiguous Layer of Overlapping Microcells -Scenario I 28 Figure 3.9: Probability of Blocking for a Contiguous Layer of Overlapping Microcells -Scenario II 30 Figure 4.1: Overlay System 36 Figure 4.2: Macrocell Markov Chain 42 Figure 4.3: Call Flow for Overlaying Microcells with Macrocells 43 Figure 4.4: Handoff Activity Cases 47 Figure 5.1: Channel Assignment for Macrocell Assigned 3 Channels per Microcell . . . 51 Figure 5.2: Blocking Probability of Overlay 52 Figure 5.3: Dropped Call Probability of Overlay 53 Figure 5.4: Handoff Activity of Overlay 55 vii Figure 5.5: Carried Traffic of Overlay 56 Figure 5.6: Channel Assignment for Overlay with Macrocell Frequency Reuse 58 Figure 5.7: Blocking Probability of Overlay with Macrocell Frequency Reuse 59 Figure 5.8: Dropped Call Probability of Overlay with Macrocell Frequency Reuse . . . . 60 Figure 5.9: Handoff Activity of Overlay with Macrocell Frequency Reuse 60 Figure 5.10: Carried Traffic of Overlay with Macrocell Frequency Reuse 61 Figure 5.11: Overlay Arrival Rates for Varied Channel Assignments 62 Figure 6.1: Microcell Coverage Area in an Urban Area 64 Figure 6.2: Channel and Signal Blocking Probability without Coverage Overlap 68 Figure 6.3: Channel and Signal Blocking Probability of One Isolated Group of Overlapping Microcells 69 Figure 6.4: Probability of Signal and Channel Blocking of Contiguous Overlapping Micro-cells with P h o l e = 0.10 70 Figure 6.5: Channel and Signal Blocking Probability of Contiguous Overlapping Micro-cells versus Probability of Signal Outage 71 Figure 6.6: Maximum Supportable Arrival Rate of Contiguous Microcells (Scenario I & II) with Radio Coverage Holes 72 Figure 6.7: Handoff Activity of One Isolated Group of Overlapping Microcells Versus Rel-ative Coverage Radius 73 Figure 6.8: Probability of Dropped Call for One Isolated Group of Overlapping Microcells Versus Relative Coverage Radius 73 Figure 6.9: Handoff Activity for Contiguous Overlapping Microcells Versus Relative Cov-erage Radius 74 viii Figure 6.10: Dropped Call Probability for Contiguous Overlapping Microcells Versus Rela-tive Coverage Radius 74 Figure 6.11: Call Flow for Overlay System with Radio Coverage Holes 76 Figure 6.12: Channel and Signal Blocking Probability of Overlay (with Macrocell Frequen-cy Reuse) with and without Macrocell Radio Coverage Holes 77 Figure 6.13: Channel and Signal Blocking Probability of Overlay Versus Probability of Sig-nal Outage 78 Figure 6.14: Maximum Supportable Arrival Rate Versus the Number of Assigned Macrocell Channels 79 Figure 6.15: Handoff Activity in Overlay Versus the Number of Assigned Macrocell Chan-nels 80 Figure 6.16: Dropped Call Probability in Overlay Versus the Number of Assigned Macrocell Channels 80 Figure C. 1: Common Cluster Configurations for Hexagonal Cellular Patterns 91 I X ACKNOWLEDGEMENTS First and foremost, I thank the Lord God Almighty, for His blessings, His love, and His grace. I thank my research supervisor Dr. R.W. Donaldson for his insights, guidance and encouragement. He carefully critiqued the work and showed me the importance of thoroughly understanding key principles. The wisdom he bestowed will remain with me well past the time I have spent working for him. This research was supported by a grant from the Canadian Institute of Telecommunications Research which was provided by Dr. R.W. Donaldson and by a scholarship from the Natural Science and Engineering Research Council (NSERC). I would like to thank my mom and dad for their love, support and encouragement. I would like to express my gratitude to my colleagues for their help throughout my studies. CHAPTER 1: Introduction 1.1 Cellular Radio Communications The growth of cellular communications has exceeded all early estimates [1]. Since 1983, when the first wireless phones went into service in the United States, the wireless industry has grown phenomenally with over 50 million U.S. subscribers in 1997 [2]. World-wide, the number of cellular subscribers has expanded to over 80 million [3]. With decreasing start-up costs, improved services and new installations in emerging Asia-Pacific countries, some analysts predict annual increases of 40% and 590 million subscribers worldwide by the year 2001 [4]. The first generation of wireless phones were built using analog technology with high transmission power levels and were primarily used for mobile (vehicular) communica-tion. Present cellular platforms, termed second generation mobile services, provide high qual-ity digital voice as well as circuit-switched data. Cell sizes are reduced and the subscriber base includes a growing proportion of pedestrian users. As part of the latter stages of second generation mobiles services, preliminary Personal Communications Services (PCS) systems, occupying the 1.8 - 2.2 GHz spectrum, were deployed across North America in 1997 [1]. In the early 1990's, governing standards bodies in Europe and North America began to consider third generation mobile communications under the project titles Universal Mobile Telecommunications Systems (UMTS) and International Mobile Telecommunications 2000 (IMT-2000) [5]. Third generation systems improve upon current systems with increased capacities, multimedia services, and tetherless access. Future services will require higher data 1 Introduction rates, higher quality of service, and the availability of services anywhere. A wireless user would be able to access the network globally from any country and locally from shopping malls, airports, office buildings and other common areas. To achieve ubiquitous coverage, mobile users should have access across multiple network technologies and standards. 1.2 Motivation The traditional technique of increasing coverage and capacity in cellular systems has been cell splitting. Other engineering methods, such as discontinous transmission (DTX), power control, sectorization with down tilting, and reuse partitioning were applied to reduce interference and support higher user densities. These techniques involve modifications of existing macrocellular architecture. Macrocell base stations can be characterized as having raised antennas (> 10 m), high power transmissions (2-10 W) and large radii (> 2 km). Microcells are the next technological step in achieving the goal of ubiquitous cov-erage and high capacity [6]. In Europe, microcells with small lightweight base stations have already begun to emerge [4]. With limited transmission powers, antennas below roof level, and small radii (< 500 m) microcells enable service for the high traffic demands of next gener-ation wireless systems. In addition to capacity improvements, microcell base stations cost significantly less than macrocell base stations. Microcells are deployed as an underlay cell to an existing macrocell. An appropriate deployment of one or two microcells within a macro-cell region enables regions of high traffic ('hot spot') to exhibit improved quality of service. Once microcell technology has matured with a large subscriber base, a contiguous layer of microcells can be deployed in urban areas with high population density. 2 Introduction Despite improved system capacity, new concerns arise with the installation of microcellular base stations. Low base station antennas relative to the surrounding buildings creates a propagation environment [7, 8] characterized by strong line-of-sight (LOS) propa-gations, pertaining to direct paths between the base sites and the handset, and weak non-line-of-sight (NLOS) propagation, pertaining to paths blocked by buildings. The combination of weak transmission and NLOS paths results in radio or coverage holes where the signal strength fails to meet the minimum requirements for establishing a communication link. As cell sizes shrink, mobiles cross cell boundaries more frequently and the handoff rate increases. As mobiles handoff between cells, the overhead associated with tearing down and setting up communication channels and the location updates will increase the signalling traffic. For high-speed vehicles, the signalling overhead will not correspond appropriately to the mobile's dwell time within a cell. Pedestrian traffic comprises a larger portion of traffic in downtown cores where microcells will be readily deployed. Handover rates of low mobility users, such as pedestrians, will be minimally changed by cell size reductions. Once a contiguous layer of microcells has been deployed, (i.e.: at every street cor-ner or intersection) it becomes difficult to further subdivide microcells because of the pres-ence of strong LOS signals. It is useful to consider methods of providing multiple base site coverage to regions to cover radios coverage holes, reduce the number of handoffs, and increase capacity. One method of achieving multiple base site coverage is by increasing the transmission power of base sites to expand the overlap between adjacent cells. Another method is to overlay a group of microcells with a macrocell. Both alternatives are considered in our study. Introduction 1.3 Objectives This thesis analyzes the capacity and coverage enhancements of overlapping microcells and overlaying microcells with macrocells. The objectives of this work are as fol-lows: 1) to analyze the blocking probably, dropped call probability, handoff off activity, and carried traffic as coverage overlap is varied for a single isolated group of microcells. 2) to propose growth scenarios for a contiguous layer of microcells with channel reuse and analyze the performance in terms of blocking probability, dropped call probability, handoff off activity, and carried traffic 3) to determine the improvement in signal and channel blocking probability and maximum supportable arrival rate when radio coverage holes are considered in overlapping microcellular environments. 4) to analyze the blocking probability, dropped call probability, handoff activity, and carried traffic of microcells with overlay macrocells with and without channel reuse. 5) to determine the improvement in coverage blocking probability and maximum supportable arrival rate when radio coverage holes are considered in macro-cells overlaying microcells. This work differs from previous work in the following ways: • when microcell overlap, the resulting increase in interference is considered with proposals to accommodate the interference • radio hole effects are considered in overlapping and overlaying architectures 4 Introduction • a simplified overlaying model is developed and utilized • handoff activity is derived and calculated for the overlay 1.4 Outline of the Thesis Chapter 2 describes the frequency reuse concepts and the mobility model used in the overlapping and overlaying cell models. As well, the growth environment of the relevant wireless architectures are depicted. Chapter 3 describes the extended overlapping microcell model, and the results for an single isolated group of microcells is presented. The performance of two growth scenarios in a contiguous layer of microcells is described and analyzed. The overlaying of microcells with a macrocell model is considered in Chapter 4. Results of an overlaying system with and without macrocell frequency reuse are presented in Chapter 5. The effects of radio holes are considered in Chapter 6, where the benefits of over-lapping and overlay architectures are determined. Comparisons for the blocking probability, maximum supportable message arrival rate, and handoff activity are made between the various architectures and configurations. Chapter 7 summarizes the results and provides some suggestions for future research. 5 CHAPTER 2: System Issues 2.1 Cellular Engineering and Channel Reuse Efficiency Frequency reuse is the core concept behind cellular mobile communication sys-tems. Frequency reuse takes advantage of propagation losses over distance by reusing chan-nels in another cell a specified distance away. In this way, mobiles in different geographical regions use the same frequencies or channels simultaneously. With frequency reuse, the spectrum is more efficiently used but an improperly designed system results in serious interference. In general, a minimum amount of interference will be generated by the common use of channels and this is called co-channel interference. A primary issue in cellular engineering is balancing the maximization of spec-tral efficiency with the minimization the co-channel interference. Figure 2.1: Co-channel Interference with Six Interferers System Issues Although in practical situations, cell shapes are irregular, engineers draw hexago-nal-shaped cells to simplify the planning and design of a cellular system because it approaches the ideal circular power shape. The hexagonal cells are tessellated to form a regular pattern where reuse clusters can be formed. The limiting factor of cellular systems is the carrier-to-interference (C/I) ratio. A worst case scenario occurs when a mobile communicates with the base station at the edge of the cell boundary (R). The signal received from the communicating base station is propor-tional to R"Y, where y is the pathloss exponent determined from the terrain environment (y typ-ically varies between 2 and 4). Likewise, interference from the six co-channel inteferers is proportional to D"Y, where D is the distance between co-channel cells. Total carrier to interfer-ence ratio [9] from the six nearest co-channel cells is shown in equation (2.1). C R 1 1 6D~y (2.1) The parameter (D/R), is called the co-channel reuse ratio. With regular hexagonal cells, the co-channel reuse ratio is related to the frequency reuse factor (N) as follows: ^ = «/3N (for hexagons) (2.2) R The frequency reuse factor (N) is the number of cells the total channel set is shared between. 2.2 Wireless Growth Scenarios for Capacity Enhancements To meet the growing demand for wireless access, different technologies and engi-neering techniques are employed to enhance capacity and coverage. Figure 2.2 outlines the growth scenerios for evolving cellular networks. 7 System Issues 1. Traditional Macrocells (R > 2 km) o H—» CO 1_ CD c to c o CO o c CD D l T3 c o o 0> CO 2. Macrocells with reduced radius (R > 1 km) 5> 3. Macrocells + 'hot spot' microcells (microcell radius < 500 m) O CO c a> Figure 2.2: Wireless Growth Scenerios for Evolving Cellular Networks When cellular services were initially offered, macrocells (1) were deployed to pro-vide cost efficient coverage. Macrocells are characterized by base station heights above the surroundings and by large coverage areas. With a growing subscriber base, macrocells (2) were split to increase capacity. Decreased size maintained the same frequency reuse but with spatially denser traffic loads. Cellular engineering techniques were further applied to macrocellular base sites to System Issues improve capacity. Sectorization involved replacing omnidirectional antennas with several directional antennas and downtilting the antennas to reduce relevant interference. Sectoriza-tion enabled reduction of cell reuse from 12 to 7. Speech coding improved the quality while reducing the channel occupancy of calls by not transmitting during silence periods. Reuse partitioning assigned different frequency reuse scenarios to areas with different levels of inter-ference. Microcells (3) are the next technological step to accommodate wireless systems growth. With lower antenna heights and limited transmission power, 'hot spot' microcells ser-vice areas with high traffic densities or poor coverage. Macrocells continue to provide broad coverage in conjunction with the strategically placed 'hot spot' microcells. With larger market penetrations, a contiguous layer of microcells (4) will be needed to serve densely populated urban cores. Traffic will be composed of a higher propor-tion of pedestrians. Some regions might be served only by a layer of microcells while other areas would include both a microcell and macrocell layer. Investigations in this thesis will be concerned primarily with this stage of evolving wireless systems. Picocells (5) serving office buildings, shopping malls, and airports represent an additional layer in the multitier scenario. New sources of high density traffic will be gener-ated by the need for transmission of data generated in office environments. 2.3 Mobility Model A mobility model is used to determine the probability of handoffs for mobile users. Numerous models are available in literature with varying assumptions and consider-9 System Issues ations [11, 12]. In this thesis, tracking the specific direction and speed of a mobile within a cell is less of a concern than the determination of the cell boundary crossing rates of mobiles. Consequently, the mobility model in [11] is used to model handoff departures. In the mobility model, it is assumed that cells are circular, mobiles are uniformly distributed in the system, and mobiles in microcells move in straight lines with a direction uni-formly distributed between [0, 27t). Cell residence time (sojourn time) is defined as the length of time a mobile terminal resides in the cell. Mobile residence time for a call initiated cell and the residency time for an arbitrary cell is derived. The mean cell residence time E[T] in an arbitrary cell in terms of mean vehicle speed E[V] is found to be: E[T] = -*!L- (2.3) L J 2E[V] Mean boundary crossing rate per mobile and the mean outgoing handoff rate per calling mobile are defined as p b and p h respectively. Mean outgoing handoff rate is approxi-mated by the mean boundary crossing rate. In [13], it is shown the two are equivalent if there are no handoff failures. The resultant mean handoff departure rate (ph) is given by the recip-rocal of the mean cell residence time in (2.3) As with other mobility research [14,15], and for the sake of simplicity, the cell res-idence time is assumed to be a negative exponentially distributed (n.e.d) random variable. In subsequent chapters, the mean handoff departure rate is used to derive the mean handoff arrival rate. 10 CHAPTER 3: Overlapping Microcells In cellular radio communication systems, calls are normally served by the base sta-tion which provides the best signal quality. Within a base site's coverage area, a mobile might also establish an acceptable quality communication link with another base site. Under various circumstances, overlapping coverage areas can be used advantageously to alleviate base sta-tion failures or traffic 'hot spots'. Overlapping is inevitable for continuous coverage across multiple cells [16] and a minimum 50% coverage overlap is required between adjacent micro-cells. Under uniform and normal operating conditions and with appropriate system control, overlapping coverage can improve teletraffic system performance. In this thesis, the base sta-tion coverage was extended to produce large areas of overlap between microcells. Analysis was performed on the teletraffic performance in terms of blocking probabilities, dropped call probabilities, handoff activity and carried traffic. Both isolated clusters of microcells and a contiguous layer of microcells were investigated. 3.1 Model for Overlapping Coverage Areas A teletraffic model for overlapping microcells was formed, based on the percent-age areas in a microcell which were covered by 1, 2, or 3 base sites. The model for calculating overlapping areas of a hexagonal cell was previously derived [17]. 11 A1 Figure 3.1:Overlapping Microcells The coverage region in figure 3.1 was tessellated by microcells with a fixed hexag-onal cell radius Rh. The boundary where a mobile-base link is of acceptable quality defines the coverage radius R. The coverage radius exceeds the hexagonal radius for continuous cov-erage (R > Rh). Expanding R forms larger coverage overlap. Terminals can then access one, two, or three base stations, depending on the region where (Aj, A 2 , or A 3) the call originates. Total area of a hexagonal cell (H) is as follows, The angle 0 (figure 3.1) is one half the angle of the arc between points y and z. Angle a is the arc between points x and y. One, two or three base site coverage is computed as a function of H = 3-fR2h. (3.1) e. (3.2) Ax = 6J 2(73(sin0) 2-9+sinGcosG) ^ 2 = 9-^R2h-2A]-nR2 (3.3) (3.4) 12 Overlapping Microcells A 3 = nR2-3j3R2h + A^ ( 3 ' 5 ) The cell areas were computed as percentages, p 1 ; p2, and p3, by dividing A 1 5 A 2 , and A 3 , respectively, by the total hexagonal area (3.1). 3.2 System Description 3.2.1 State Characterization The system consists of a regular pattern of overlapping hexagonal cells with C t total channels and a frequency reuse factor (N). Each microcell has C m communications channels and C h channels reserved for handoffs. Analysis characterizes the states as the num-ber of calls served by a base site. For a mosaic or cluster of cells, the overall system state could have been represented by a multidimensional state system with each dimension corre-sponding to the number of calls served by a particular base station. A linear increase in the number of base stations in the system causes the number of system states to grow exponen-tially. For computational feasibility, microcell bases are decoupled from each other by apply-ing average new call and average handoff arrivals from adjacent cells. The statistical behavior of each base is modeled independently from all others. With a homogeneous microcell traffic scenario and an independent cell model, one overlapping microcell models the statistical behavior of the system [18]. States for the overlapping microcell model are driven by cellular mobile arrivals and departures. Arrivals include both new calls and handoffs while departures include call completions and handoff departures. Memory less traffic assumptions allow the problem to be 13 Overlapping Microcells formulated in the context of Markov chains with the following parameters: i) New call arrivals in any hexagonal microcells are poisson processes with mean new call arrival rate A n ii) Cell residence time % is the time a user is in the communication with a micro-cell base before a handoff occurs. As described in Section 2.3, x is a random variable with a negative exponential distribution (n.e.d.) and a mean equal to iii) The unencumbered session duration xe of a call is the average length of a com-pleted call. Unencumbered session duration is a n.e.d. random variable with mean equal to p. -1. iv) New call arrivals are uniformly distributed over the area of each microcell. In addition to serving calls from its own area, base A (figure 3.1) also accepts call requests from neighboring cells (B, C, D, E, F, G) within its coverage area. Defining r\ as the probability that at least C m - C h calls are in progress at a base, the fraction of new calls arising from region Aj is computed. The additional new arrivals from calls in blocked A 2 adjacent cell regions is p2*An*r). The A 3 region arrival rate from blocked adjacent microcells is more computationally complex and the reader is directed elsewhere for details [18]. Equations for the A 3 region arrival rate (A.3) is the following: ^3 = 2PlAnr]kn, (3.6) where: K = ri + (1 - r i ) | . (3.7) Combining the arrivals for A 1 ; A 2 , and A 3 yields the effective overlapping new call arrival rate 14 Overlapping Microcells ln = An(\+p2r\ + 2p3r\kn). (3.8) Notwithstanding the system model of microcells with independent call arrivals, handoff arrivals are tied to handoff departures. Any handoff departure from one microcell becomes a handoff arrival to a neighboring microcell. Handoffs occur uniformly over the cov-erage boundary. For a microcell with i calls in progress, the handoff departure rate is z*PHm-Accounting for handoffs from the different areas (A l 5 A 2 , A 3), equations (3.9) - (3.12) describe the handoff arrivals. Reasoning similar to new call arrival analysis was applied to the handoff arrivals except arc lengths (in figure 3.1) were used instead of percentage areas. Anal-ogous to kn, kh = v + ( l - v ) | , (3.9) where the probability of C m calls in progress is v. Handoff arrival rate as a function of the departure rate from an adjacent microcell is as follows: where a is the arc where A 3 handoffs arrive. An aggregate handoff arrival rate (Xh) for micro-cells was obtained by averaging the handoff arrival rates over the probability of i calls in progress (p(i)) and multiplying the result by the number of neighbors (6). Equating (3.10) to the handoff arrival rate (3.11) shows that this rate is proportional the expected number of calls (E[C]) in progress. cm K = 6^p(i)Ah(i) (3.11) ; = o = 6 £ [ C ] M ^ + 5!*a) (3.12) 15 Overlapping Microcells 3.2.2 Steady-State Distribution Pursuant to the system model, each microcell is characterized by the number of communicating calls in progress with generating poisson arrival streams (kn and ?ih). Up to the point where C m - C h channels are occupied, new call and handoff arrivals (A^ + A,h) are accepted for communication within the microcell. Up to the state with C m calls in progress, handoff arrivals (A,h) are accepted into the microcell. Mean channel holding time in a cell is derived by considering the mean outgoing handoff rate per calling mobile (p^-1) and the unencumbered call duration (p_1). The inverse of the sum of the two departure rates is the mean channel holding time (p u _ 1) in a microcell. Stated differently, the mean departure rate is the sum of the rates due to termination from call completion and handover. p"1 = + (3.13) The Markov chain state diagram [19] with the respective transition processes is shown in figure 3.2, where m = C m - C h and n = C m-( c h = u ) fch=lj fch=2j • • • Kh=m-n ( c h = m ) • • • fch=n-lj fch=n) Figure 3.2:Handoff Priority Markov Chain The steady-state distribution of the Markov chain is solved iteratively [10] with the following equations (3.14-3.16). Long-term probabilities of being in a state where i channels in the microcell are occupied is as follows: 16 Overlapping Microcells pt = prob{Ch = i} = py! „ n + K) K i ., if i < m if m < i < n (3.14) The reference for determining p; was p0, the probability of no active channels in a microcell. PQ = prob{Ch = 0} = u = o mr. i-m-\-l m + (3.15) For the handoff arrival (3.12),the mean number of calls in a macrocell E[C} is given by: E[C] = i = 0 (3.16) 3.3 Performance Measures for Overlapping Coverage 3.3.1 Blocking and Handoff Failure Probability Blocking probability is the average fraction of new calls which cannot gain imme-diate access to a channel. The steady-state probability that at least C m - C h calls are in progress is represented by TJ. For calls arising in region A 1 ; the blocking probability is T | . For new calls arising in A 2 , the blocking probability is the probability that both microcells were occu-pied, or T) . Likewise, for calls in A 3 , the probability is T) . Overall blocking probability (Pb) for a microcell is averaged over the regions. Pb = P\T\ + P2T\ + PJ\ (3-17) Probability of handoff failure (Phf) is the average probability that a communicating call requests a handoff and is denied a channel within the possible target cells. Handoff failure 17 Overlapping Microcells depends on the steady-state probability of all the channels being occupied (v) and the relative arc lengths for A 1 5 A 2 , A 3 handoffs. With a homogenous traffic distribution, the fraction of 20 oc handoff arrivals into region A] is — and the fraction into region A 2 is — (see figure 20 + OC 20 + OC 3.1). Probability of handoff failure was averaged for the fraction of calls with the respective blocking probability in each region: p * = w k ' + w r J ( 3 1 8 ) 3.3.2 Dropped Call Probability Calls are dropped when a call in progress is interrupted due to handoff failure at some time during the call. Dropped call probability (Pdc) relies on the derived handoff failure probability (3.18). At a reverse transition (from state n to state n-1) in the Markov chain (fig-ure 3.2), the mobile exits a cell either from call completion or handoffs. Given a mobile departure from a cell, the probability that the event was a handoff request is shown below (3.19) . Likewise, the probability that the departure resulted from a call completion is given by (3.20) . P(handoff request| handoff request or call completion) = (3.19) P(call completion |handoff or call completion) = — - — (3.20) Based on the probability of handoff request or call completion, the probability of / handoff requests (Phi) was found. The probability of i handoff requests is coupled with the probability of i successful handoff channel accesses and the probability that the last transition was a call completion. 18 Overlapping Microcells Prob{i handoffs} = PM = — — — . (3.21) U-Pdc) = X ^ i - V l T T l I ( 3 - 2 2 ) Substituting equation (3.21) into (3.22) produces a series equation (3.23) which is simplified to obtain the probability of dropped call (Pdc) oo I (1 - pdc) = I — ^ T T T ( 1 - V (3-23) r 1 + MV ^ c = .. *. ^ (3-24) 3.3.3 Handoff Activity Handoff activity is the expected number of handoffs which a non-blocked call experiences. There are exactly i handoffs if (1) the call fails at the ith handoff, or (2) it suc-ceeded at the ith handoff and successfully completes before the (i-t-l)th handoff. The proba-bility of the first event is shown in (3.25) and the probability of the second in (3.26). Prob{cz\\ fails on the ith handoff} = J(1 -Phfm '^^ hfm (3-25) Pro6{call completes after ith handoff} = ( ^ H m \ 1 - PMJ ^ (3.26) Consequently, handoff activity (HA) is a series which includes the two identified cases. A simplified equation for H A is obtained in (3.28). The simplifying steps are explained in Appendix A. = V / J ^ S _ ] ( i _ p J - l ( p + ( i _ p )_V—) (3.27) 19 3.3.4 Carried Traffic The carried traffic (CA) per frequency reuse cluster is the average number of chan-nels that are occupied, where p(- is the steady-state probability of / calls in progress in a micro-cell. oo i = 1 3.4 One Isolated Group of Overlapping Microcells An isolated single group of overlapping microcells is first considered. The micro-cell cluster might be in a small town located far from the next town or city with wireless com-munication coverage. Otherwise, the cluster could also be located in an urban setting where the microcell channels form a separate frequency band from channels in adjacent larger cells. In both sceneries, co-channel interference is not a limiting factor in the system design. In other words, microcell overlap can be increased without concern for an increased level of co-chan-nel interference in an adjacent cluster of microcells. Section 3.5 considers the contiguous microcell layer where the co-channel interference becomes critical to proper system design. 3.4.1 Numerical results For the purpose of generating practical numerical results, the following parameters are used. A total of one hundred and fourty channels are available (Ct = 140) with a frequency 20 Overlapping Microcells reuse of seven (N = 7). Each cell is allocated twenty channels (C = 20) and one channel (C h = 1) is reserved for handoff calls only. The unencumbered call duration is two minutes (1/p. = 120 sec) and the hexagonal cell radius is three hundred meters (Rh = 300 m). The mean vehi-cle velocity is 8 m/s (E[V] = 8 m/s). Microcell base station coverage radius (R) is increased over the hexagonal cell radius (R/Rh) by ratios of 1.0, 1.134, 1.309, and 1.500. Probability of blocking for the overlapping microcells appears in figure 3.3. Increasing overlap improves the allowable new call arrival rate for a given probability of blocking. For a two percent grade of service (GOS 2% = P b = 0.02), the arrival rate improves by 0.6, 1.2, and 2.1 calls/min/microcell. o PH, 60 e o o « <+H o o ' I R/Rh = 1.500 (simulation) - *-R/Rh= 1.309 (simulation) R/Rh =1.134 (simulation) R/Rh =1.0 (simulation) —*-R/Rh = 1.500 (theoretical) - - -R/Rh = 1.309 (theoretical) R/Rh =1.134 (theoretical) - -R/Rh = 1.0 fth£°J£jJ£^i_^ 8 9 10 11 12 Calls per minute per microcell 15 Figure 3.3 .-Probability of Blocking of One Isolated Group of Overlapping Microcells Figure 3.4 displays the probability of dropped calls for overlapping cells. For a two percent GOS, arrival rates of 6, 6.6, 7.2, and 8.1 calls/min/microcell are accommodated with R/Rh of 1.0, 1.134, 1.309, and 1.500, respectively. Dropped call probability improve-21 Overlapping Microcells merits occur similarly to the blocking probability as R/Rh increases. 0 u C H U a, Cu o cd X l o 6 7 8 9 10 11 12 13 14 15 Calls per minute per microcell Figure 3.4Probability of Dropped Call of One Isolated Group of Overlapping Microcells 2.0 F o •a CJ X e 3 1.0 h 0.0 R/Rh = 1.500 R/Rh = 1.309 R/Rh = 1134 R/Rh = 1.0 i i i i i i i j i i i _ 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0 13.0 14.0 15.0 Calls per minute per microcell Figure 3.5: Handoff Activity of One Isolated Group of Overlapping Microcells Figure 3.5 shows the effect of microcell coverage overlap on handoff activity. Handoff activity decreases with an increase in the relative coverage radius. Handoff activity decreases as the cell area increases At low call arrival rates (4 calls/min/cell), the number of 22 Overlapping Microcells handoffs is determined primarily by the longer cell dwell times. Figure 3.6 shows the traffic carried in microcells with overlapping coverage. Reduced blocking probabilities results in more calls gaining access to the microcells. Expanding the overlap increases the average number of calls carried by each microcell. As the arrival traffic increased, the carried traffic reaches closer to the limit of C m = 20 channels. 140.0 120.0 5 1 0 0 - ° u 80.0 60.0 40.0 20.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0 13.0 14.0 15.0 Calls per minute per microcell Figure 3.6: Carried Traffic of One Isolated Group of Overlapping Microcells 3.5 Contiguous Layer of Microcells Within a high-traffic geographical region, hexagonal microcells would be tessel-lated and grouped into frequency reuse clusters. Distinct channel sets would be allocated to each microcell and channels would be reused to maximize wireless coverage and capacity. With channel reuse, increased overlap generates stronger levels of interference from co-channel cells. Signal quality is directly associated with the carrier to interference 23 Overlapping Microcells (C/I) ratio. Consequently, signal quality degrades with increased overlap. Considering that the propagation loss is inversely proportional to a power of the propagation distance, the worst case carrier C/I is a function of the distance between co-channel base sites and the coverage radius. Expanded microcell coverage by increasing the overlap is achieved by transmitting at a higher power levels. When coverage radius (R) expands, the distance (D) between co-channel cells must also expand to maintain signal quality. In a frequency reuse system, expanding the distance between cells corresponds to increasing the frequency reuse factor (N). As mentioned in chapter 2, C/I is varies with the distance between co-channel cells (D) over the coverage radius (R) ratio. ( t = 2 - 7 " 4 ) ( 3 - 3 ° ) For extended overlap between microcells, the coverage radius (R) increases by a factor k, where k = R/Rh. The minimum ratio for the coverage radius (R) over the hexagonal radius (Rh) is one. R2 = kRx (3.31) To preserve the same C/I, the distance between co-channel cells increases proportionally to the larger radius: D2 = kDx. (3.32) The frequency reuse factor (N) is an integer within a certain set. It was computed with the fol-lowing equations [20] for hexagonal cells: N = i2 + ij + j2 (3.33) {i, j are integers} N> - irj <"4> 24 Overlapping Microcells Since R h is fixed, the reuse factor (N) is proportional to the square of the distance between co-channel cells. In chapter 2, equation (2.2) related N with D and R and is applicable in tradi-tional cellular engineering where the hexagonal radius and the coverage radius are identical. However, (3.34) is used to calculate the frequency reuse when the hexagonal radius no longer corresponds to the coverage radius. Equations (3.32) and (3.34) are substituted into the new frequency reuse factor (N2) which was found to increase in proportional to k . Appendix C explains the derivation of (3.36) fully. 3.5.1 Contiguous Layer Growth (Options) In a contiguous layer of microcells, the channel reuse separation distance grows as the amount of overlap expands. Consequently, the basic cell cluster contains more microcells but with fewer channels per microcell. There are two options in accommodating the growth in cluster size see figure 3.37. The cluster can (1) grow in size by encompassing a larger number of microcells and geographical area, or, (2) cover the same geographical area but with more and smaller microcells. The advantage of increasing the geographical area of the cluster is that new base stations need not be installed. Existing base sites are simply regrouped into dif-ferent clusters and channel sets. However, the total channel set ends up covering a larger region, effectively reversing the objective of increasing the number of channels available per unit area. The second option of maintaining a constant cluster size requires new base sites in the cluster region. Depending on the degree of overlap, the new coverage areas might or (3.35) (3.36) 25 Overlapping Microcells might not reuse some or all the previous transceiver sites. Installing new base sites might not be economically feasible. Figure 3.7: Contiguous Microcell Growth Scenarios 3.5.2 Numerical Results for Scenario I Increasing the overlap in a contiguous layer with constant hexagonal cell size expands the coverage of the cluster. The new frequency reuse factor (cluster size) is computed using equation (3.36). The total number of channels (Ct) is 140 and the initial frequency reuse factor (N) is 7. Overlap ratios (R/Rh) are chosen to be 1.0, 1.134, 1.309, 1.500; the respective frequency reuse factors of 7, 9, 12, 16 are obtained from (3.36). New call arrivals 26 Overlapping Microcells were in units of calls/min/km2. Table 3.1 lists the characteristics of the clusters in scenario I. Table 3.1:Scenario I Cluster Characteristics Microcell Radius (m) Frequency Reuse Factor Number of Channel/ Cell Microcell Area (km2) Cluster Area (km ) 1.0 300 7 20 .2338 1.64 1.134 300 9 15 .2338 2.10 1.309 300 12 11 .2338 2.81 1.500 300 16 8 .2338 3.74 Blocking probability is calculated and shown in figure 3.8. As the overlap and fre-quency reuse increases, the number of channels allocated to each microcell decreases. The result is losses of 7.5, 13, and 16 calls/min/km2 as the R/Rn ratio increments for a two percent GOS. Scenario I blocking probability degrades largely because of the expanded cluster area and the loss of channels. Fewer channels per microcell and larger cluster area outweigh the performance gain resulting from the capability to access multiple base sites. 27 Overlapping Microcells 10 10 10 10 10 10 ... „...~ Ph = 2% / ' / 7 / X / / / / / / / " / / _ / / / / R/Rh= 1.500 R/Rh - 1.309 R/Rh= 1.134 R/Rh = 1.0 J 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 , , I , , , , l I < 10 20 30 40 50 Calls per minute per km2 60 Figure 3.8 .-Probability of Blocking for a Contiguous Layer of Overlapping Microcells - Sce-nario I The results of the contiguous layer scenario I growth for the probability of dropped call, handoff activity and carried traffic are summarized in Table 3.2. Dropped call probability remains constant so long as the GOS is two percent. Number of handoffs per call reduces as the cell coverage increases. Handoff activity is largely determined by the cell area and decreases as R/Rh increases. Therefore, degradations in blocking and dropped call probabili-ties minimally affect the number of handoffs per call. Carried traffic performance degrades as the overlap increases. Losses in channels per microcell as the overlap increases, reducing the average number of calls in progress. At high arrival rates, the carried traffic per microcell cluster is limited by the maximum number of channels per microcell. Overlapping Microcells Table 3.2: Scenario I Numerical Results for GOS 2% Supportable Arrival Rate [calls/min/ km2] Dropped Call Probability (Pdc) Handoff Activity (HA) [handoffs/call] Carried Traf-fic (CA) [Erlangs] 1.0 25.66 0.01 2.04 81.44 1.134 18.5 0.01 1.78 75 1.309 13 0.01 1.54 71 1.500 10 0.01 1.34 70 3.5.3 Numerical Results for Scenario II Increasing overlap for contiguous microcells and constant cluster area reduces the tessellated microcell size. The new frequency reuse factor (N2) is computed using equation (3.34), except Dj is fixed and R h varies. . " . - " . ( j - J (3-37) To calculate the new cell sizes, a new hexagonal radius (Rh2)1S computed. Equation (3.36) is combined with equation (3.37). The new hexagonal cell radius is determined by dividing the old hexagonal radius (R n l = 300) by the coverage over hexagonal radius ratio: (s Table 3.4 lists the characteristics of the cluster in scenario II where the cluster area was constant. 29 Overlapping Microcells Table 3.3:Scenario II Cluster Characteristics R/Rh Microcell Radius (m) Frequency Reuse Factor Number of Channel/ Cell Microcell Area (km2) Cluster Area (km ) 1.0 300 7 20 .2338 1.64 1.134 265 9 15 .1824 1.64 1.309 229 12 11 .1362 1.64 1.500 200 16 8 .1039 1.64 In figure 3.9, the probability of blocking for a scenario II contiguous overlapping layer is shown. As overlap expands, the supportable call arrival rate reduces; however, the loss is minimal when compared to scenario I (figure 3.7). For (R/Rh) ratios of 1.0, 1.234, 1.309, and 1.500, the respective new call arrival rates a GOS two percent are 26, 24.2, 23.3, and 23.3 calls/min/km . x On, C a o o s <4-l o •S o UH CM 10 10 10 10 10 10 Pb = 2% f / ' / / / ..' / / ' 1 R/Rh = 1.500 R/Rh = 1.309 R/Rh= 1.134 R/Rh = 1.0 1 I ' / \l 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 i i 1 i i i i l 10 20 30 40 50 60 Calls per minute per km Figure 3.9: Probability of Blocking for a Contiguous Layer of Overlapping Microcells - Sce-nario II 30 Overlapping Microcells Results for the dropped call probability, handoff activity and carried traffic for sce-nario II are summarized in Table 3.5. Table 3.4:Scenario II Numerical Results for GOS 2 % Supportable Arrival Rate [calls/min/ km2] Dropped Call Probability (Pdc) Handoff Activity (HA) [handoffs/call] Carried Traf-fic (CA) [Erlangs] 1.0 25.66 0.010 2.04 81.44 1.134 24.2 0.014 2.01 77 1.309 23.2 0.016 2.00 74 1.500 23.2 0.021 1.99 73 For a two percent GOS, the probability of dropped call, handoff activity and the carried traffic improve compared to scenario I. Probability of dropped calls increases for a given GOS. Handoff activity is constant because the coverage cell size remains the same. Carried traffic improves slightly relative to scenario I. Blocking probability of scenario II is much better than scenario I but at the cost of installing new base sites. 31 CHAPTER 4: Overlaying Microcells and Macrocells Third generation mobile communication systems will be comprised of multiple layers: macrocells for terrestrial coverage, microcells for dense traffic, picocells for indoor traffic, and satellite cells for global coverage. Microcells overlaid by a macrocell are the foun-dation for a multitiered network. Principles and experience obtained in the design and imple-mentation of a two-layer system would be applicable to a multiple layer network. In an overlay, the area covered by a cluster of microcell base stations is also covered by a macrocell base station, sometimes called an umbrella cell. With the simultaneous operation of two cell layers, efficient distribution of resources and traffic between the layers is essential. Ease of implementation, signal load minimization, spectral efficiency and cost are the basic design criteria. 4.1 Resource (Spectrum) Sharing/Partitioning A primary issue in two-tiered wireless system design is the allocation of frequency or time slot channels between microcells and macrocells. Spectrum sharing is one technique proposed, where different cell types concur-rently use the same spectrum [21]. For time division multiple access (TDMA) and frequency division multiple access (FDMA) systems, spectrum sharing actualizes into the sharing of time slots or frequencies between microcells and overlay macrocells. Sharing a channel pool avoids the reduction in trunking efficiency which otherwise occurs when channels are subdi-vided between cell layers. With sharing, complications do arise regarding the difference in 32 Overlaying Microcells with Macrocells transmission power between microcell and macrocell base sites. Macrocell bases transmit at higher power levels to cover a larger area. A handset communicating, on channel A in a microcell experiences severe co-channel interference from another handset communicating on channel A with a macrocell. In a spectrum sharing code division multiple access (CDMA) system, microcells and macrocells transmit and spread across the same bandwidth. Since CDMA is interference limited, an increase in transmission power to or from one mobile pro-duces additional interference for all other mobiles. Mobiles communicating with a macrocell at a high power level generate strong interference levels for all mobiles communicating in the underlaying microcells. A second alternative is to divide the channel set between cell layers such that each utilizes a separate set of frequencies or time slots. Trunking losses are incurred when the channel set is subdivided. Implementation is simpler because the power and interference lev-els are orthogonal between the layers. Performance comparisons of orthogonal sharing with TDMA in both cell layers, and spectrum sharing with TDMA in the microcells and CDMA in the macrocells appears in [20]. Orthogonal channel splitting maximizes resource efficiency because the strong cross-tier interference generated in spectrum sharing is avoided. A third option reuses the macrocell frequencies within selected microcells. An isolated microcell, within a macrocell, can reuse the channel set of another macrocell. Within a small radius around the microcell base station, the transmission power of the microcell sig-nal is stronger than the power level from macrocells using the same channels. The large prop-agation loss of the macrocell signal, despite the higher power levels, combined with the small propagation loss of the microcell signal allows the reuse of macrocell channels within the microcell. Implementations, however, are limited to a few, selected 'hot spot' microcells. 33 Overlaying Microcells with Macrocells 4.2 Call Admission and Handoff Strategies The criteria for call admission into microcells and macrocells determines the traf-fic distribution between cell layers. Call admission and handoff strategies specify those mobiles and conditions under which a communication link is established with a cell base sta-tion. As a practical example, a service provider may have initially installed a Global System for Mobile Communications macrocells operating at 800-900 MHz (GSM 900) and then deployed Personal Communications System microcells operating at 1.9 MHz (PCS 1900) [4]. In this system, there are three handset access capabilities. Older handsets (GSM 900) only access macrocells. Some new handsets (PCS 1900) only access microcells. A third handset type, dual mode, has the ability to access microcells and macrocells. In this case, traffic distri-bution among cell levels is determined by the number of subscribers with each handset type. A proposed solution to reduce high handover rates in small cells is to admit calls to macrocells based on mobile speed [10, 14, 22, 23]. Low mobility (low velocity) terminals are directed to communicate with microcells, while terminals in fast moving vehicles communi-cate with macrocells. When a call is setup, the system has no knowledge of the terminal's velocity. The network estimates the speed of the terminal once the call has started. The easiest method to estimate the speed is to measure and record the time a mobile spends within a cell (dwell time). Several strategies to sort terminals based on the dwell time of mobiles in cells have been proposed [10, 22, 23]. A mobile call is first admitted into a microcell [10, 21]. If the observed dwell time is lower than a threshold T, the mobile hands over to the macrocell. Otherwise, if the dwell time is greater than x, the mobile hands off to an adjacent microcell. In [22], the network esti-34 Overlaying Microcells with Macrocells mates the user's speed using maximum likelihood (ML), and minimum mean squared error (MMSE) estimates. In [10], mobiles estimate their own mobility by tracking cell dwell (sojourn) times. The mobile's estimate is compared with a threshold for microcell or macro-cell selection on the next handoff. By having mobiles track their own mobility, the network does not need to collect extensive data on every mobile's speed. A call handled by a macrocell may be permitted (reversible) or not permitted (non-reversible) to be handed down to a microcell. Most teletraffic analysis [14] assume non-reversible hierarchical cell architectures. In [14], calls were directed towards the lowest layer which gave acceptable coverage and if no channels were available, the request overflowed non-reversibly to a higher layer. To a certain extent, the hierarchical system inherently segre-gates users according to their mobility. Reversible systems are advantageous because they free up macrocell channels when microcell channels become available [24,25]. A situation could occur where a user trav-els to the office while making a macrocell call. Upon arrival at the office building in the downtown core, the user is low-mobility but continues to occupy a macrocell channel in a non-reversible system. In [25], analysis was performed on a reversible system. Five percent higher traffic (GOS = 1%) was accommodated compared to a non-reversible system, however, the handover rate increased by twenty percent. The primary disadvantage of reversible sys-tems is the higher signalling load. 4.3 System Description The system serves a large geographical region tessellated by hexagonal microcells 35 Overlaying Microcells with Macrocells (see figure 4.1). Every group of K microcells are overlaid by a larger macrocell. Gateways serving each cell are linked to the fixed terrestrial network where call routing and call hand-offs occur. The area is traversed randomly by a large number of vehicular and pedestrian ter-minals. A fixed bandwidth is allocated to provide C t total channels through either time or frequency separation. A frequency reuse, N , divides all the channels among N cells resulting in C m channels per microcell. Each microcell has C h channels reserved for handoff arrivals. g e o g r a p h i c a l r e g i o n Figure 4. L Overlay System Channels are partitioned between microcells and macrocells. Initially, C m = C t / N channels are assigned to microcells without overlay. With the addition of an overlay, channels are transferred from the N underlying microcells to the overlaying macrocell. Channels (C M ) supplied to the macrocell reduce the number of channels (C m ) per microcell. No channels are reserved for handoffs in the macrocells because of the low number of macrocell channels usu-ally available. The system operates under the following criteria. i) A l l mobile terminals are dual mode, with the capability to communicate with 36 Overlaying Microcells with Macrocells either microcells or macrocells. New call originations are directed to the clos-est microcell base station. If new calls are attempted in microcells where the number of channels used is equal to or greater than C m - C h , the call is for-warded to the overlaying macrocell. To minimize the system processing load, a terminal's admission to a layer is not based on its' velocity. ii) If all C M channels in the macrocell are occupied, the new call that was blocked at the microcell, is blocked again and cleared from the system. If fewer than C M calls are engaged in the macrocell, the new call begins communication with the macrocell base station. iii) A handoff request is initiated only if a terminal leaves the communication range of its serving cell. iv) A microcell call which initiates a handoff first searches for idle channels in the next closest (target) microcell. v) A microcell handoff request is directed to the overlaying macrocell if all C m channels in the target microcell are occupied. Handoff requests, from micro-cell or macrocell calls, to the macrocell are blocked and dropped if all C M mac-rocell channels are occupied. vi) The system is non-reversible. Handoffs from calls carried in the macrocell layer cannot be handed down to a microcell. 4.4 System Characterization The mathematical structure of the system is that of a multilayer hierarchical pois-37 Overlaying Microcells with Macrocells son overflow system. An exhaustive state description, using a multidimensional markov chain [26] is not practical due to the excessive number of states needed. For the overlay model in this thesis, a multidimensional markov chain would require up to 1.3469x1010 states. Since the time to calculate steady-state distribution grows exponentially as the number of states increases, the multidimensional markov chain method is not used. In the homogeneous model used, all microcells are statistically identical and all macrocells are statistically identical. The overall system is analyzed by focusing on one mac-rocell region. Statistical behavior is considered under the condition that a cell's neighbors exhibited their typical random behavior independently. With memoryless assumptions, the problem is defined in the context of markov chains with the following parameters: i) New call arrival processes offered to microcells are Poisson processes with mean new call arrival rate A,m. ii) Handoff arrival is a Poisson process with mean handoff arrival rate A, n m . The handoff arrival rate depends [10] on the handoff departure rate (2.3) (i.e. the inverse of the mean residence time) and on the expected number of calls in a macrocell: Km = HmE[C] (4.1) iii) The cell residence time is the period a user remains within the communications coverage of the serving cell. Cell residence time is a random variable having negative exponential distribution (n.e.d.) with mean equal to P-hm 1 f ° r micro-cells and p h M for macrocells. The means p.h m and p ^ conform with equation (2.3) except the microcell and macrocell radii are substituted for the respective departure rates. 38 Overlaying Microcells with Macrocells iv) Overflow microcell arrivals to macrocells are a Poisson process [15] with mean arrival rate A M . v) Macrocell handoff arrivals are a Poisson process [10] with mean handoff arrival rate Handoff arrival rate is found using (4.1) except p h M replaces M h^m-vi) The unencumbered session duration of a call is the full duration of a non-dropped call. The unencumbered session duration is a n.e.d. random variable with mean equal to p _ 1 . vii) New call arrivals are uniformly distributed within each microcell. Since each microcell is statistically identical, call arrivals to macrocells are also uniformly distributed. 4.5 Analytical Model of Macrocell/Microcell Overlay System Analysis of microcell with macrocell overlay begins at the microcell level and ascends to the macrocell level. Call arrivals to the macrocell depend on the number of calls blocked at the microcells. Performance measures are derived in the next section. 4.5.1 Microcell Level Microcells are characterized by the number of occupied communication channels. Up to the point where C m - C h channels are occupied, new call and handoff arrivals (X^ + X h m ) are admitted for communication into the microcell. Until C m calls are in progress, handoff 39 Overlaying Microcells with Macrocells arrivals (Xhm) are accepted into the microcell. A handoff departure from a microcell corresponds to a handoff arrival to another microcell. For a homogeneous system in statistical equilibrium, the average handoff arrival rate must equal the average handoff departure rate. Applying these principles allows micro-cells to be decoupled from each other by aggregating the handoffs to and from adjacent micro-cells. Essentially, each microcell functions independently of all others with an average handoff arrival and departure rate. The mean channel holding time in a cell is calculated by considering the mean out-going handoff rate per calling mobile (Phm ) and the unencumbered call duration (p ). The inverse sum of the two departure rates determines the mean channel holding time (p^ _ 1) in a microcell. H = (^ m + ^ ) _ 1 (4-2) The one-dimensional state transition diagram for a microcell is identical to the one shown in figure 3.2. The mean handoff arrival is now X^m instead of X^. The steady-state dis-tribution of the markov chain is calculated iteratively [10]. Long-term probabilities of being in a state i where (pj) channels in the microcell are occupied is given by equation (3.14). The reference for determining pj is p0, the probability (3.15) of having no calls in progress in a microcell. The steady-state distribution establishes the probability that a new call arrives when all new call channels (state / > C m - Cn) are occupied or the probability that a handoff is requested when all the channels (state i = C m ) are occupied. The probability of microcell blocking (Pbm) is given by: Overlaying Microcells with Macrocells Cm Pbm = Prob{new call blocking} = ^ pi . (4.3) i = cm- Ch The probability (Phfm) of microcell handoff failure is the probability that a currently commu-nicating call requesting a handoff is denied a channel within the prospective microcell. phfm = Prob{handoff failure} = pCm (4.4) 4.5.2 Macrocell Level The macrocell is also characterized by the number of occupied communication channels. The total number of macrocell channels is C M and no channels are reserved for handoffs. The mean channel holding time in a macrocell is found by accounting for the mean outgoing handoff rate per calling mobile (PhM_1) a n a " m e unencumbered call rate (p.-1). The inverse sum of the two departure rates determines the mean channel holding time in a macro-cell ( P M - 1 ) -As described in the previous section, microcell traffic levels are statistically inde-pendent from each other. All microcells have equal arrival rates, departure rates and number of channels resulting in the same statistical equilibrium. The results of one microcell repre-sents the behavior of all the underlying microcells. The macrocell arrival rate (AM) is comprised of those calls not accepted into the N underlying microcells. The first portion of A M includes the effects of blocked new microcell calls, the combinations of i of N microcells, and the new call arrival rate. The second part of the macrocell arrival rate is similar except it relates to the microcell probability of handoff failure and the handoff arrival rate. Applying geometric distribution simplification and substi-41 Overlaying Microcells with Macrocells tution yields the macrocell arrival rate in (4.6). N N -ifN i=1 7=1 J AM = P b m K N + P h f m K m N The resulting markov chain is shown in figure 4.2 with respective arrival and (4.5) (4.6) departure rates. A " Figure 4.2.-Macrocell Markov Chain The steady state distribution of the second markov chain can be calculated to yield the proba-bility (PbM) of blocking for the macrocell: r ^ s = f / ( c „ ! ) PbM = V J (4.7) M fhM + X h m \ i 4.6 Overlay Performance Measures 4.6.1 Effective Blocking Probability Blocking and handoff failure probabilities are first calculated for the microcells. The macrocell blocking probability is computed once the microcell overflow traffic is found. Figure 4.3 illustrates traffic flow to the microcell and macrocell overlay and the relationship 42 Overlaying Microcells with Macrocells with the blocking and call dropped probabilities. The effective blocked traffic for new and handoff calls accounts for the blocked traffic from the microcells and the blocking probability at the macrocell. ^Vnew mi JA_n.ew htm. MICROCELLS BLOCKED CALLS mi rhfm I i MICROCELLS rbM XhM MACROCELLS Figure 4.3: Call Flow for Overlaying Microcells with Macrocells For a new call to be blocked in a microcell with macrocell overlay, successive attempts to access microcell and macrocell channels are blocked. Thus, the overall blocking probability P b e is the product of the probability of blocking at each layer: Pbe = PbMPbm (4-8) 4.6.2 Effective Dropped Call Probability In an overlay system a call is dropped if the macrocell has no channel available at a microcell handoff failure or at a macrocell handoff request. Dropped call probability P d c e includes the handoff failure probability at the microcell level and the blocking probability of 43 Overlaying Microcells with Macrocells the macrocell. As in chapter 3, the dropped call probability is derived by finding the probability that the call is not dropped. There are three cases where a call is successfully completed (see 4.9 - 4.11): (1) a call begins and completes all communication in the microcell layer, (2) a call admitted to the microcell experiences a microcell handoff failure but successfully completes all subsequent macrocell handoffs, and (3) a call is admitted directly to the macrocell and completes the call at a macrocell, in which case all macrocell handoffs are successful. Equa-tions (4.9 - 4.12) present the probability of the call not being dropped where the three cases are considered. Pi = v,- = o (4.9) oo / oo ^ ;=0V=0 J ^j^r^-pbM)i+l (4-io) ^ 3 = Pbm I Hj-^—(l-PbM)j (4.11) (1-Pdce) = Pl+Pl + Pl (4-12)-Computing the probability of / microcell handoff requests (hj) and j macrocell handoff requests (Hj) depend on the relevant conditional probabilities (4.13 - 4.16). Given that a departure from a microcell has occurred the probability of a handoff request is (4.13) and the probability of call completion is (4.14). Similar conditional probabilities are found for macrocell handoff requests and call completions (4.15 & 4.16). 44 Overlaying Microcells with Macrocells P(micro handoff request | micro handoff request or call completion) = ^ h m (4.13) V-hm + ^ P(micro call completion I micro handoff request or call completion) = — ^ — (4.14) P(macro handoff request | macro handoff request or call completion) = ^hM (4.15) VhM + ^ P(macro call completion | macro handoff request or call completion) = — ^ — (4.16) Applying the conditional probabilities, the probability of / microcell handoffs requests and j macrocell handoffs requests are found. i Prob{i microcell handoffs} = Phi = — . (4.17) Prob{j macrocell handoffs} = PHj = — — — . (4.18) (H + i W 7 The probability of i successful microcell handoffs is the product of the probability (Phi) of i handoff requests and the probability of / successful channel accesses, (1-Phfm)1- The probabil-ity of successful macrocell handoffs is similarly determined with PHj and (l-P b My. Substitut-ing the handoff probabilities into equations (4.9 - 4.11) results in the expanded series equations (4.19 - 4.21). i = 0 oo / oo • \ p2 - a -/>*,) i s fejd - w-V -pbMy <4.2o, c o 45 Overlaying Microcells with Macrocells Applying the simplification principles used in Appendix A, the equations for the three cases of successful call completion are: Pi = O - n J , ^ ; (4-22) Phfm (M + M//J — — M l~Phfm (M- + M7im^fc/m)X(M-+M'/iM^>Af) i t *M + ^hMPbM Combining the simplified results (4.22 - 4.24) with the dropped call equation (4.12) yields the P2 - (l~Pbm\ D MY 5 w / 5 \ (4.23) P3 = Pbm * n (4-24) effective dropped call probability. Pdce = 1 - ^ 1 - ^ 2 - ^ 3 (4-25) Numerically determined blocked and dropped call probabilities are compared with a reference network where only microcells exist. In the absence of overlaid macrocells, the dropped call probability is given by (4.22) with the (1 -Pbm) t e r m e q u a l t 0 unity since this performance measure only considers admitted calls. 4.6.3 Effective Handoff Activity Handoff activity is the expected number of handoffs a non-blocked call experi-ences. There will be exactly i handoffs if (1) the ith handoff fails at both the microcell and macrocell, or (2) it succeeds at the ith handoff and successfully completes before the (i+l)th handoff. There will be (i + j) handoffs if the call fails at the ith microcell handoff but obtains access to a macrocell channel and attempts j subsequent macrocell handoffs. At the macrocell layer: (3) the call fails at they'th macrocell handoff or (4) successfully completes before the (/+l)th handoff. Calls which are blocked at the microcell but are accepted to the macrocell can also: (5) fail at the ith macrocell handoff or (6) complete after the z'th macrocell handoff. 46 Figure 4.4 illustrates all six cases. Overlaying Microcells with Macrocells ^handoff failure / < successful handoff or call completion ^channel access failure Figure 4.4-.Handoff Activity Cases The handoff activity (HA) is a summation of the six possible cases as indicated below by equation (4.26). The expected number of handoffs for cases (1) to (6) are repre-sented by C], C 2 , C 3 , C 4 and C 5 6 respectively. The probability that (/ - 1) successful micro-cell handoffs occur with a handoff failure on the ith handoff and a successful macrocell channel access is denoted by PS; in (4.27). HA = Cx + C2 + C3 + C4 + C5 6 PS; = J- 1 (4.26) (4.27) Equations representing the six call scenarios are presented below (4.28 - 4.32). Cases (3) and (4) represent the situations where the call fails a microcell handoff but is able to establish a communication channel with a macrocell. Cases (5) and (6) represent the scenario where the new call is initially blocked but engages in communication with a macrocell. i = 0 (4.28) 47 Overlaying Microcells with Macrocells i = 0 7 - 1 ( M W M Y~* c3 = U-PbJl,J S ( ^ ) j = 0 U = 1 °° r j 7 = 0 U = l ( 1 _ PbM^ P f c M M / / M Y y / - * M-M / / M + M J FOM M / / M + M-OO j c5,e = pbm^j{^^yi-pbMy \ p b M + ( 1 - ^ M ) M ^ M M + M / /M (4.29) (4.30) (4.31) (4.32) Applying series simplification rules to the above equations results in new equations contain-ing the common factors (M l s M 2 , M 3 , M 4 , M 5) defined below. Simplified results for the six cases are shown in equation (4.34), (4.35), (4.40), (4.41) and (4.42). M, = M// m (M + M t f m ) (M + M / / „ A / m ) 2 (4.33) c2 = U-PhJ M M + M//, • d-V Mi (4.34) (4.35) M2 = M3 = MffO T(l -Phfn) (M + M / / w ^ / m ) (M + \ i H M P b M ) MA Mi M5 = Mn ( 1 - M 2 ) 2 ( 1 - M 3 ) 2 (4.36) (4.37) (4.38) (4.39) 48 Overlaying Microcells with Macrocells C 3 = ( . - ^ ) ^ f ( ^ ) M 5 (4.40) (p + \LHMPbM) C 5 > 6 = ,.. . T T , (4-42) The summation of all handoff activity cases (4.26) combines the simplified results of the six call scenarios. For comparison, the handoff activity with no macrocell coverage uses C] and C 2 only, with the microcell blocking probability P b m equal to zero (to only account for non-blocked calls) and the macrocell blocking probability P b M equal to unity. 4.6.4 Carried Traffic The carried traffic is the mean number of calls carried in the macrocell region. The expected number of calls in an overlay region includes the expected number of calls per microcell multiplied by the number of microcells per macrocell together with the expected number of macrocell calls. CM CM Ca = Nj^ipm(i)+^jpMU) (4-43) i=0 j=0 49 CHAPTER 5: Results of Microcell and Macrocell Overlay 5.1 Analysis without Macrocell Frequency Reuse 5.1.1 Channel Assignment In chapter 4, a general analytical model was derived for the microcell architecture with macrocell overlay. Section 5.1 considers the scenario where frequency reuse has not been applied to the macrocell channels. Similar to the isolated group of overlapping micro-cells, co-channel interference in the macrocell layer is not a limiting factor. Each macrocell covers the area of seven (K = 7) microcells. A total of one hun-dred and forty (Ct = 140) duplex communication channels are available for each macrocell overlay region. Frequency reuse is also seven (N = K = 7) which simplifies the channel assignments between the two layers. Initially, each microcell is allocated twenty communica-tion channels (C m = 20) with one channel (C^ = 1) is reserved for handoff arrivals to the microcell. Channels are assigned to the macrocell by transferring channels from the N under-lying microcells. If q channels are removed from each of the N underlying microcells, then a total of q*N channels are available at the macrocell. Figure 5.1 shows the channel assign-ments where three channels from each microcell are set aside for the macrocell. 50 Results of Microcell and Macrocell Overlay Figure 5.1 :Channel Assignment for Macrocell Assigned 3 Channels per Microcell 5.1.2 Numerical Results For the purpose of generating practical numerical results, the previous values from section 3.4.1 for 1/p, R h, and E[V] are used in the overlay system. Additionally, the macrocell hexagonal radius is 1000 meters. Figure 5.2 shows the effect of macrocell overlay channels on the probability of blocking (Pb) in the system. One channel is re-assigned from each of the N = 7 microcells to the macrocell resulting in a (19,7) configuration with 19 channels per microcell and 7 channels per macrocell. As well, two and three channels are transferred from the microcells resulting in (18,14) and (17,21) channel partitions. With a (19,7) channel assignment, a new call originating in a particular microcell first attempts to access one of the 18 microcell channels (1 channel is reserved for handoff priority). If one of the 18 microcell channels can not be accessed, the call attempts communication through one of the 7 macrocell channels. The macrocell supports the overflow of calls blocked at the microcell level; these 51 Results of Microcell and Macrocell Overlay macrocell channels are shared across a larger region and a larger number of users. 10 X PH e M o o s <+H o >1 X cd X O 10 10 10 10 10 / / 17 micro chan, 21 macro chan (simulation) 18 micro chan, 14 macro chan (simulation) 19 micro chan, 7 macro chan (simulation) 20 micro chan (simulation) 17 micro chan, 21 macro chan (theoretical) 18 micro chan, 14 macro chan (theoretical) 19 micro chan, 7 macro chan (theoretical) 20 micro chan (theoretical) _i_L 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 Calls per minute per microcell 12.0 13.0 14.0 15.0 Figure 5.2:Blocking Probability of Overlay For a two percent GOS, the allocation of one channel per microcell to the overlay-ing macrocell yields a significant 1.0 calls/min/microcell improvement in the maximum sup-portable arrival rate. This is a 17% improvement at the original arrival rate of X = 6.0 calls/ min/microcell. Allocation of three channels per microcell increased the arrival rate by 1.75 calls/min/microcell for a 46% gain. The initial transfer of one microcell channel incurred the largest percentage gain in maximum allowable arrival rate, subsequent microcell transfers returned diminishing improvements. The gains in supportable arrival rate indicates that the pool of macrocell channels is large enough such that the loss from channel sharing is minimal. 52 Results of Microcell and Macrocell Overlay Probability of dropped calls is displayed in figure 5.3. Handoff calls have access to all the C m channels in the microcell as well as all C M macrocell channels. 10 10 10 10 10 10 _L / . / / // / l / i i i i l 17 micro chan, 21 macro chan 18 micro chan, 14 macro chan 19 micro chan, 7 macro chan 20 micro chan i I i i i i l i i i i l i i i i l i i i i l 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0 13.0 14.0 15.0 Calls per minute per microcell Figure 5.3.'Dropped Call Probability of Overlay Overlay results in a significant reduction in dropped call probabilities. At an arrival rate of X = 6.0 calls/min/cell, the probability of dropped call is 0.010, 0.00148, 0.00064, and 0.000003 respectively for 0, 1,2, and 3 microcell channels transferred to the macrocell. The probability falls by almost an order of magnitude for each microcell channel assigned at the given arrival rate. For a GOS of two percent, the maximum supportable arrival rates to a microcell for (20,0), (19,7), (18,14) and (17, 21) channel configurations are 6.0, 7.01, 7.45 and 7.75 calls/min/cell, respectively. At those arrival rates, the probabilities of dropped call are 0.010, 0.0158, 0.0167 and 0.0170, respectively. Probabilities of dropped calls exceeds P^ = 0.01 when the arrival rates are allowed to increase in conjunction with the reduced blocking probabilities. Dropped call probability can be reduced in two ways. The 53 Results of Microcell and Macrocell Overlay number of handoff priority (hp) channels can be increased, or the maximum allowable call arrival rate can be constrained to meet a P d c = 0.01 requirement. Table 5.1 shows the results of varying arrival rate with constraints on either P b or P d c . Table 5.1: Performance of Pb>Pdo a n t * ^ Versus Channel Partitions and System Design Constraints Channel Partition (micro, macro) Para-meter h p = l P b = 0.02 h p = 2 P b = 0.02 hp=3 P b = 0.02 h p = l P d c = 0.01 (19,7) Pb 0.02 0.02 0.02 0.0131 P d c 0.0158 0.0151 0.0115 0.01 X 7.01 6.68 6.17 6.75 (18,14) Pb 0.02 0.02 0.02 0.0119 P d c 0.0167 0.0134 0.0132 0.01 X 7.45 7.03 6.67 7.15 (17,21) Pb 0.02 0.02 0.02 0.0120 P d c 0.0170 0.0134 0.0124 0.01 X 7.75 7.32 6.92 7.45 In comparison to increasing the number of handoff priority channels with a P b = 0.02 constraint, the supportable arrival rate is higher for the P d c = 0.01 constraint with one handoff priority channel. Additionally, P b is below 0.02 with P d c constrained to 0.01. Figure 5.4 displays the handoff activity (HA) for the overlay system. The call duration, handoff rates, handoff failures and blocking probabilities all affect the number of handoffs per call. 54 Results of Microcell and Macrocell Overlay Figure 5.4:Handoff Activity of Overlay As the arrival rate increases, the handoff activity of both the overlay and non-over-lay system diminishes. In the non-overlay system, the number of calls blocked or dropped increases due to higher blocking and handoff failure probabilities. Handoff activity fell because a smaller percentage of calls established communication or completed all handoffs. At an arrival rate of X = 6 calls/min/cell, the numbers of handoffs for (20,0), (19,7), (18,14), and (17,21) are 2.01, 1.97, 1.94, and 1.90, respectively. At higher arrival rates (from 10 to 15 calls/min/microcell), both the handoff failure and dropped call probabilities for overlay and non-overlay systems converge (see Figure 5.3). On the other hand, for the same arrival rates, handoff activities tend to diverge (see Figure 5.4). The dichotomy between handoff activity and dropped call results from more calls communicating with the macrocell base stations where the longer cell residence time causes fewer handoffs per call. 55 Results of Microcell and Macrocell Overlay Figure 5.5 shows the carried traffic in a macrocell coverage region. Carried traffic includes the average number of calls in the overlaid macrocell and in each of the N microcells. 140.0 - 120.0 S 100.0 W 80.0 60.0 40.0 17 micro chan, 21 macro chan 18 micro chan, 14 macro chan 19 micro chan, 7 macro chan 20 micro chan _L_L < 1 1 I i i i i l ' ' ' < l i ' ' ' l ' ' i 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0 13.0 14.0 15.0 Calls per minute per microcell Figure 5.5:Carried Traffic of Overlay Overlay supports more calls than a one-layer microcellular system. At low arrival rates (A, < 6 calls/min/microcell), all four channel assignments have similar carried traffic. However, at higher arrival rates (A, > 6 calls/min/microcell), more calls in the non-overlay sys-tem are blocked when compared to the overlay system. Additional macrocell channels increases the carried traffic. For an arrival rate of 6 calls/min/microcell, the carried traffics for (20,0), (19,7), (18,14), and (17, 21) are 81.44, 83.67, 83.99, and 84.00 Erlangs, respectively, representing a potential improvement of approximately 3%. 56 Results of Microcell and Macrocell Overlay 5.2 Analysis with Macrocell Reuse 5.2.1 Channel Assignment Complications arise in channel assignment when the system is a contiguous layer of microcells and a contiguous layer of macrocells. Techniques can be applied to mitigate the high levels of co-channel interference between macrocells. Dynamic channel assignment (DCA) [27] assigns channels to cells depending on the measured level of interference. With this arrangement, the channel with minimum interference is assigned to a requesting mobile. Because macrocells will not carry a large portion of the total traffic, the use of antennas arrays [28] is a possible solution. Such an array can form independent beams to cover different mobiles or groups of mobiles. Both techniques, however, require new equipment and/or addi-tional signal processing. Fixed channel allocation (FCA) using frequency reuse is the sim-plest method to control macrocell co-channel interference. Because FCA designs for the worst case interference scenarios, it provides a reference for comparison with future channel assign-ment techniques. Assuming the same (C/I) requirements in the macrocell layer as the microcell layer, the macrocell layer uses the same frequency reuse factor (N = 7). Figure 5.6 shows the arrangement and assignment of channels to microcells and macrocells when three microcell channels are moved to the macrocells with frequency reuse. 57 Results of Microcell and Macrocell Overlay Set Channels A 1-17 B 21-37 C 41-57 D 61-77 E 81-97 F 101-117 G 121-137 M1 18,19,20 M2 38, 39, 40 M3 58, 59, 60 M4 78, 79, 80 M5 98, 99, 100 M6 118, 119, 120 M7 138, 139, 140 Figure 5.6: Channel Assignment for Overlay with Macrocell Frequency Reuse Each microcell is initially assigned 20 channels with 1 channel reserved for hand-off arrivals. In figure 5.6, every seven microcell channels transferred are split into seven mac-rocell channel sets (Ml, M2, M3, M4, M5, M6, M7). A mobile could access up to 17 microcell channels and 3 macrocell channels for a (17,3) channel partition. The number of macrocell channels available to a particular mobile has been markedly reduced from section 5.1, when mobiles could access 21 macrocell channels. 5.1.2 Numerical Results The previous values from section 3.4.1 for C t , C m , C h , 1/p, R h, and E[V] are used in the overlay system with macrocell frequency reuse. Additionally, the macrocell hexagonal radius is 1000 meters. Blocking probabilities for an overlay system with macrocell frequency reuse is shown in figure 5.7. The performance gains of the overlay achieved in the section 5.1 no longer exist. For a GOS of two percent, the allocation of 19 channels per microcell and 1 58 Results of Microcell and Macrocell Overlay channel per macrocell (19,1) results in a 0.05 calls/min/microcell loss in maximum support-able arrival rate. With (18,2) and (17,3) partitions, the maximum arrival rates further reduces by 0.25 and 0.45 calls/min/microcell, respectively, which corresponded to 4% and 7.5% losses. Under the assumed conditions, the reduced supportable arrival rates which result from trunking efficiency losses outweigh the gain in handling traffic overflow. 0 X CjH, oo c M o o s M—< o o 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0 Calls per minute per microcell 13.0 14.0 15.0 Figure 5.7:Blocking Probability of Overlay with Macrocell Frequency Reuse The dropped call probability displayed in figure 5.8 behaves similarly to the block-ing probability. For a system requirement of GOS 2%, the dropped call probabilities are 0.010, 0.0137, 0.01409, and 0.01493 for channel partitions of (20,0), (19,1), (18,2) and (17,3) respectively. Not only does the maximum supportable arrival rate diminish, but the probabil-ity of dropped call rises notwithstanding the reduced arrival rates. 59 Results of Microcell and Macrocell Overlay T3 OH as U -o u Cu Cu o o >1 133 O 17 micro chan, 3 macro chan 18 micro chan, 2 macro chan 19 micro chan, 1 macro chan 20 micro chan ' ' ' ' i i i i i i i i i i i 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0 Calls per minute per microcell 13.0 14.0 15.0 Figure 5.8:Dropped Call Probability of Overlay with Macrocell Frequency Reuse o -o e 03 Z 3.0 2.0 1.0 17 micro chan, 3 macro chan 18 micro chan, 2 macro chan 19 micro chan, 1 macro chan 20 micro chan i i i i i i 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0 13.0 14.0 15.0 Calls per minute per microcell Figure 5.9: Handoff Activity of Overlay with Macrocell Frequency Reuse 60 Results of Microcell and Macrocell Overlay Figure 5.9 presents the handoff activity of calls in the overlay systems. The num-ber of handoffs per call diminished with additional overlay channels. This further demon-strates that handoff activity is influenced primarily by the cell residence time and less by the probability of handoff failure. Carried traffic in the macrocell system is shown in figure 5.10. With macrocell fre-quency reuse, the carried traffic drops relative to the non-overlay system as additional chan-nels are allocated to macrocells. Channels in the macrocell layer are reused over a distance larger than for microcells, resulting in reduced number of channels per macrocell coverage region. The total number of channels in a macrocell region is K*(20 - q) + q, where q is the number of channels given up by a microcell. For a system with three channels per microcell given up, the total number of channels per macrocell region decreases to 122 channels com-pared with 140 channels in the one layer system. 140.0 c o '5b cj = 120.0 CJ o o 100.0 OO c W 80.0 ca CJ u 60.0 40.0 17 micro chan, 3 macro chan 18 micro chan, 2 macro chan 19 micro chan, 1 macro chan 20 micro chan I ' ' ' ' I ' ' ' ' I _1_L. 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0 Calls per minute per microcell 13.0 14.0 15.0 Figure 5.10: Carried Traffic of Overlay with Macrocell Frequency Reuse 61 Results of Microcell and Macrocell Overlay 5.3 Summary The microcell with macrocell overlay improves in all performance measures when macrocell frequency reuse is not considered. With macrocell frequency reuse, only the hand-off activity continues the improved performance in comparison with the non-overlay system. The most telling measure is the probability of blocking or the maximum supportable arrival rate for a given Pb. Dropped call probability behaves similarly to P b and the carried traffic reflects the calls accepted into the system (i.e.: calls not blocked). Figure 5.11 shows the overlay maximum supportable arrival rate (for a P b = 0.02) as channels are transferred from microcells to macrocells, with and without macrocell fre-quency reuse. 10.0 o 3 o L i i i i I i i i i I i i i i 0.0 1.0 2.0 3.0 Channels per microcell assigned to macrocell Figure 5.11 .-Overlay Arrival Rates for Varied Channel Assignments 62 CHAPTER 6: Overlapping and Overlaying with Radio Coverage Holes 6.1 Microcell Propagation Environment Wireless personal communication networks of the future will require massive cov-erage in terms of large capacity and ubiquitous access [6]. Microcells can support the high traffic density. The propagation environment and the supporting technology in microcells dif-fers from that of conventional macrocells; microcell base stations transmit at lower power levels and the antennas are placed at street level (approximately lamp post height), and are often below the roof tops of surrounding buildings. Coverage areas are smaller (300 - 500 m). The combination of low antenna height and tall buildings creates coverage which is shaped by the streets and surrounding buildings. In macrocells, the Rayleigh distribution is commonly used to describe the envelope of the multipath propagation [9]. Microcells are characterized by a direct line of sight signal (LOS) and blocked non-line of sight (NLOS) signal [7, 8]. LOS propagation is characterized by Rice probability distributions while NLOS propagation is characterized by Rayleigh prob-ability distributions. With rectilinear building and street patterns [11], the coverage of micro-cells tends to be concave diamond shaped, with elongated LOS signals along the main streets as shown in Figure 6.1 63 Overlapping and Overlaying with Radio Coverage Holes Figure 6.LMicrocell Coverage Area in an Urban Area Low antenna heights and larger probability of signal blocking by buildings reduce the co-channel interference thereby improving the reuse efficiency. At 900 MHz, the reuse efficiency improved by 50%; however, at 2 GHz there is no apparent improvement in reuse efficiency [7]. The outage probability was studied for varying pathloss thresholds, cell sizes, and antenna heights in a microcellular propagation environment [8]. Probability of signal outage was the cumulative distribution of the pathloss over the coverage area. The probability den-sity function of the pathloss [8] was: r max rt T / \ f(PLl\(ni,n2,a))= J _ e x p ( - - ^ ))f{r)dr (6.1) r . ' min The variable nj is the propagation loss exponent before the fresnel breakpoint, n2 is the prop-agation loss exponent after the fresnel breakpoint, a is path loss standard deviation, pj is the 64 Overlapping and Overlaying with Radio Coverage Holes reference pathloss in decibels at one metre, and r is the distance between mobile and base sta-tion. Pathloss function (PL]) considered the fresnel breakpoint and the transmit and receive antenna heights [8,29]. For the overlay system, results from equation (6.1) can be used to compare the probability of signal outage for the varying antenna heights and coverage sizes of microcells and macrocells in an urban setting. Table 6.1 shows the signal outage probabilities for varying signal thresholds where one base station antenna height (h) was 13.3 m with a cov-erage radius (r) of 1 km and the other base station antenna height (h) was 3.7 m with a cover-age radius (r) of 300 m. Signal threshold of the larger cell was altered to account for the higher transmission compensating for the larger pathloss. The results could be applied to a microcell and a small macrocell. The signal outages for the two base stations at a specified pathloss are on the same order of magnitude. Table 6.1: Signal Outage Probability Versus Path Loss Path Loss (dB) Signal Outage Probability h = 3.7 m r = 300 m h = 13.3 m r = 1000 m 98 0.10 0.03 108 0.01 0.012 118 0.0003 0.0008 6.2 Radio Coverage Holes in Overlapping Microcell Systems Radio coverage hole effects are considered for the overlapping microcell model of chapter 3. To receive service through a microcell base station, the mobile must establish a link of acceptable quality (signal availability) [30]. In addition, a channel must be available to 65 Overlapping and Overlaying with Radio Coverage Holes accommodate the call (channel availability). With the handoff priority model (section 3.2.1), channel availability is measured by probability variables r\ and v. The probability of a base site being in a state in which a new call or handoff could gain access to a channel is (1-T|) and (1-v) respectively. The signal availability is determined by the probability of signal outage (Poutage)' also called the probability of a radio coverage hole. The propagation environment experi-enced by a mobile communicating with two different microcell base sites can be significantly different; as a result, the signal quality of links with different base stations is assumed to be statistically independent. To establish a communication link with a microcell base site, both signal quality and channel availability conditions must be met [30]. New calls attempting communication with a microcell base station must not be in a radio hole; the probability of this event is (1 -Poutage)- As w e U ' m e r e m u s t D e l e s s m a n C-C h calls; this probability is (1 - r\). The variable TJ ' , defined below, denotes the probability that a new call cannot communicate with a particu-lar base site either because a channel is unavailable or signal quality is inadequate. (1 -TT) = d-^(l-P0Utage) (6.2) •H' = t\+Poutage-t\Poutage (6-3) Variable, u', defines the probability that a handoff request cannot communicate with a partic-ular base site based on either signal quality or channel availability considerations. V' = V + Poutage- V o^utage (6-4) The new variables T | ' and u', are substituted into the appropriate formulas for overlapping microcells (3.6), (3.9), and (3.8) and the following equations are derived: K = -n'-t- ci--n')|. (6.5) 66 Overlapping and Overlaying with Radio Coverage Holes k' = V' + (1 - v ' ) £ 2. (6.6) Likewise, the effective overlapping arrival rate (3.8) and the handoff arrival rate (3.12) are altered to include r\', D ' and Poutage-V = A „ ( 1 + P2r\f + 2p 3 Tl V ) ( 1 - Poutage) (6-7) V = 6 £ [ C ] p ^ + ^v)l-^Mtflge) (6-8) Equations (6.7) and (6.8) are substituted into equation (3.14-3.16) and the steady-state distri-bution is determined. The new overall channel and signal blocking probability (Pt,sc) for over-lapping microcells and the handoff failure probability (Phfsc) account for signal outage. pbsc = / > l l ^ ' + /vr^2 + / 7 3 1 ^ , 3 ( 6 - 9 ) Phf = -^—v' + —^-\'2 (6.10) fsc 20 +a 20 +a v ' The new performance measures (6.9 & 6.10) are substituted into the previous equations for dropped call probability (3.24), handoff activity (3.28), and carried traffic (3.29). 6.2.1 Probability of Blocking (Channel and Signal) The previous values of C m , C h , p, Rh, and E[V] are used for performance calcula-tions in the system with overlapping microcellular coverage and radio coverage holes. Radio hole effects are first considered with no cellular overlap. Figure 6.2 shows the results with various signal outage probabilities. Radio holes limit the minimum obtainable channel and signal blocking probability. In this thesis, grade of service (GOS) refers specifically to the channel blocking probability. At two percent GOS and an arrival rate of 6 calls/min/micro-cell, the channel and signal blocking probabilities were 0.0294, 0.0686, 0.118, 0.265 for the respective outage probabilities of 0.01, 0.05, 0.10, and 0.25. 67 Figure 6.2: Channel and Signal Blocking Probability without Coverage Overlap The dropped call probability behaved similarly to the blocking probability in figure 6.2. Handoff activity decreased because of higher blocking probabilities which resulted when signal outage was considered. 68 Overlapping and Overlaying with Radio Coverage Holes 10 u a c u -o a ca "3 a OH a o o S H-l o o 10 4 10 Arrival rate = 6 calls per min per microcell when GOS = 2% for R/Rh = 1.00 CP .o' - O " , © - ' . -o-' R/Rh= 1.500 — o-R/Rh= 1.309 -&-R/Rh= 1.134 - - G - - -R R h = 1.0 — • — — o 0.1 0.2 Signal Outage Probability, Poutage Figure 6.3-.Channel and Signal Blocking Probability of One Isolated Group of Overlapping Microcells Figure 6.3 shows the P b s c for various probabilities of radio hole (Poutage) for an isolated groups of microcells. The probability of channel and signal blocking is plotted for an arrival rate X = 6 calls/min/microcell where the non-overlapping microcellular architecture had a GOS of two percent in the absence of radio holes. Isolated microcells already exhibit improved blocking probabilities without holes and the trend is seen to continue as the radio hole probability increases. 69 Overlapping and Overlaying with Radio Coverage Holes 10 -a c cd "3 c a ca X ! o X ) OH c o o S o X I X I o 10 10 10 R/Rh= 1.500 RR h= 1.309 R/Rh= 1-134 Scenario II Scenario I R/Rh=1.0 Scenariol&n 10 20 30 40 50 60 Calls per minute per km Figure 6.4 .-Probability of Signal and Channel Blocking of Contiguous Overlapping Microcells with Phole = 0.10 For the two contiguous microcell growth scenarios articulated and analyzed in sec-tion 3.5.2 and 3.5.3, the blocking probabilities are higher for a given arrival rate as the overlap increased. Figure 6.4 displays the probability of channel and signal blocking when the proba-bility of radio hole is 0.10. For a two percent GOS, the non-overlapping microcell system accommodated an arrival rate of X - 25.66 calls/min/km2. The corresponding channel and sig-nal blocking probability was 0.1176. Blocking probabilities for scenario I continued to per-form worse than those of the non-overlap system for any given arrival rate. The lost performance due to channel set splitting and larger cellular coverage area outweigh the improved signal coverage. Considering only channel performance, scenario II blocking prob-abilities are higher than the non-overlapping system. When signal availability is accounted 70 Overlapping and Overlaying with Radio Coverage Holes for, scenario II demonstrated improved performance. 0 10 cj c a ca JS U T 3 a ca 13 e 60 CH 60 C CJ O S <+H o o Arrival rate = 25.66 calls per min per km h when GOS = 2% for R/Rh = 1.00 [&--Q--10 10 Q - -_ - - A - - - - - - -R«h = = 1.500, Scenario II — A— R/R„: = 1.309, Scenario II — A-RR h = = 1.134, Scenario II RR h = 1.500, Scenario I — ED— R/Rh = 1.309, Scenario I - B -R/Rh = 1.134, Scenario I R/Rh = 1.0, Scenario I & II — • — 0.1 0.2 Signal Outage Probability, Poutage Figure 6.5.'Channel and Signal Blocking Probability of Contiguous Overlapping Microcells versus Probability of Signal Outage Figure 6.5 shows the scenario I and II blocking probabilities (channel and signal) of contiguous overlapping microcells with varied radio hole probability. Scenario I had higher blocking probabilities for the R/Rh and P o u t a g e values considered. For probability of coverage hole greater than 0.03 the channel and signal blocking probabilities of scenario II are lower than the non-overlapping system. Figure 6.6 displays the maximum supportable arrival rate for a given P b s c as R/Rh is varied. The P b s c value shown for a respective P o u t age 1S m e Pbsc f ° r P ^ h = 1 00 at two per-cent GOS. Scenario I shows improvement in maximum arrival rate only when the P o u t a g e l s high at 25 percent. The maximum supportable arrival rate for scenario II increases for P o u t a g e greater than 0.03. At P o u t a g e = 0-10, the maximum supportable arrival rate increases from X = 71 Overlapping and Overlaying with Radio Coverage Holes 25.66 calls/min/km2 to X = 34.77 calls/min/km2 for a 35% gain for R/Rh=l. 134, and remained above 11% for all R/Rh values between 1.134 and 1.500. B 60.0 CJ Cu C cj Cu S 04 cd 50.0 40.0 30.0 h 20.0 10.0 1.2 1.3 Relative Coverage Radius, R R h Poutage = 0.25, P b e s c = 0.25001, Sc II Poutage = 0.10, P b e s c = 0.10151, SC II Poutage = 0.05, P b e s c = 0.05606, Sc II Poutage - 0.00, P b e s c = 0.02, Sc II Poutage = 0.25, P b e s c = 0.25001, Sc I Poutage = 0.10, P b e s c = 0.10151, Sc I Poutage = 0.05, P b e s c = 0.05606, Sc I Poutage = 0.00, P b e s c = 0.02, Sc I -0---A--• G -Figure 6.6:Maximum Supportable Arrival Rate of Contiguous Microcells (Scenario I & II) with Radio Coverage Holes 6.2.2 Handoff Activity Depending on the cause, a low handoff activity may or may not be desirable. A reduced handoff activity might indicate more dropped calls or handoff failures. On the other hand, for a constant number of calls dropped, a reduced handoff activity is preferred. Examin-ing handoff activity together with dropped call probability can help differentiate between the two cases. 72 Overlapping and Overlaying with Radio Coverage Holes Handoff activity for a single isolated group of overlapping microcells is presented in figures 6.7 and 6.8. The dropped call probability decreases with increasing R/Rn, resulting in more successfully completed calls. Normally, this would increase the total number of handoffs in the system; however, the handoff activity actually drops for larger R/Rn values. Fewer handoffs are generated because of the larger cell coverage areas. 2.5 cd U UH C J CH O § X C J 3 0.5 0.0 Arrival rate = 6 calls per min per cell when GOS = 2% for R/R h = 1.00 See Dropped Call Probability for Legend J i i i I i i i i I i i _L _l 1 I L. 1.0 1.1 1.2 1.3 Relative Coverage Radius, R/Rh 1.4 Figure 6.7: Handoff Activity of One Isolated Group of Overlapping Microcells Versus Relative Coverage Radius o CH C3 u T3 cj CH CM O X ca X o 10 Arrival rate = b calls per min per cell . .when G Q S = 2% fo^R/Rh = 1.00 10 10 10 10 10 •A-_l I I | _ 1.0 — A -P — 0 9S 1 outage — v-'-z—' r outage Poutage = 0.05 • outage = 0.10 - • --••A--= 0.00 - • -1.1 1.2 1.3 Relative Coverage Radius, R R h 1.4 <> {] i\ Figure 6.8:Probability of Dropped Call for One Isolated Group of Overlapping Microcells Versus Relative Coverage Radius Handoff activity for contiguous microcellular systems are shown in figures 6.9 and 73 Overlapping and Overlaying with Radio Coverage Holes 6.10. For both scenarios, the dropped call probabilities increase with expanded R/Rn when there are no radio holes but decrease or remain constant when radio hole probabilities of 0.05, 0.10 and 0.25 are considered. In scenario I, the number of handoffs decreases when R/Rh increases. In scenario II, handoff activity remains constant when there are no radio holes not-withstanding the increase in dropped calls; therefore, more handoffs are completed success-fully but fewer handoffs are generated overall. 2.5 Arrival rate = 25.66 calls per min per km z when GOS = 2% for R/R h = 1.00 See Dropped Call Probability for Legend i ' i i I i i i i L J i i i_ 1.0 1.1 1.2 1.3 Relative Coverage Radius, R/Rh 1.4 Figure 6.9-.Handoff Activity for Contiguous Overlapping Microcells Versus Relative Coverage Radius ca 10 O -IE I 10 /£ — • 10 10 10 10 Arrival rate = 25.66 calls per min per km when GOS = 2% for R/R h = 1.00 _i i i i I i i i i I i i Poutage = 0.25, SC II Poutage = 0.10, SC II POL . = 0.25, Sc I = 0.10, S c i r outage Poutage = 0.05, SC I - B- -- A- -- • -- A -1.0 1.1 1.2 1.3 Relative Coverage Radius, R/R h 1.4 Figure 6.10:Dropped Call Probability for Contiguous Overlapping Microcells Versus Relative Coverage Radius 74 Overlapping and Overlaying with Radio Coverage Holes 6.3 Radio Coverage Holes in Overlaying Microcells and Macrocells Radio hole effects are now considered in the microcell with overlaying macrocells model (chapter 4). The signal quality of the links with microcell and macrocell base stations are assumed to be statistically independent. The signal availability for the mobile-microcell link is Pmoutage and for the mobile-macrocell link is P]vioutage-The effective new call {Xm') and handoff (Km) arrivals to the microcells are reduced as a result of calls blocked due to radio holes. The microcell blocking (4.6) and handoff failure (4.7) probabilities include the following modified arrival rates. K' = KV-Pmoutage) &.\ 1) Km = Km(l~ Pmoutage) (6-12) Arrivals to the macrocell are comprised of not only calls blocked at the microcell (4.9) but the calls from mobiles in microcell radio holes. The effective macrocell new call (AM') and handoff 0\hM') arrival rate further account for macrocell radio coverage holes (P]yioutage)-A M = i(Pbm^n + Phfm^hm)(l ~ Pmoutage) + (^n + ^hm)Pmoutage]N(l ~ Pltfoutage) (6-13) KM = KMV ~ P'Moutage) (6.14) Figure 6.11 illustrates the flow of arrivals to the microcells and macrocells with radio cover-age holes. Calls blocked from channel availability are represented by the solid line while calls in radio holes are shown by the dashed lines. 75 Overlapping and Overlaying with Radio Coverage Holes * M- p v moutage * pbm vVnew Witage JAhm * ^ifm y 1 " pmoutape' ^ " pmoutageJ ^outage J\hm * Phfm*^" pmoutaae)l MICROCELLS moutage BLOCKED CALLS X h M Figure 6.11: Call Flow for Overlay System with Radio Coverage Holes New microcell blocking (Pbm') and handoff failure (Phfm') consider the probability of micro-cell signal outage ( P m o u t a g e ) -P ' = P + P —PP bm bm moutage bm moutage P ' — P + P — P P hfm hfm moutage hfm moutage (6.15) (6.16) The macrocell channel blocking probability is found by considering new call arrival rates (6.13 & 6.14) with (4.10). Similar to Pbm', the macrocell channel and signal blocking proba-bility (PbM') is modified. p ' _ p , p _ p p bM bM moutage bM moutage (6.17) Since the microcell and macrocell radio coverage holes are statistically independent, the effec-tive channel and signal blocking probability (Pbesc) f ° r m e overlay is the product of (6.15) and (6.17). The new performance measures (6.15 - 6.17) are substituted into the previous equa-tions for dropped call probability (4.29), handoff activity (4.30), and carried traffic (4.47). 76 Overlapping and Overlaying with Radio Coverage Holes 6.3.1 Probability of Blocking (Channel and Signal) The previous values of K, N, C t , C h , p., Rh, and E[V] in chapter 5 are used for per-formance calculations in the overlaying system with radio coverage holes. Figure 6.12 com-pares the effects of considering and not considering macrocell radio coverage holes. Frequency reuse was applied to the macrocell channels and the outage probability was 0.10. The results indicate that the macrocell coverage minimally affects the overall blocking proba-bility because of the low proportion of traffic and channels in the macrocell layer. Conse-quently, further analysis considers only microcell probability of outage. C T3 a C3 13 a e cd JS o U X> CU a o « o >1 •8 o 10 10 Pmoutage — 0.10 17 micro chan, 3 macro chan, PMoutage = 0.00 — *— 18 micro chan, 2 macro chan, PMoutage = 0.00 — *- -19 micro chan, 1 macro chan, PMoutage = 0.00 17 micro chan, 3 macro chan, PMoutage = 0.10 18 micro chan, 2 macro chan, PMoutage = 0.10 19 micro chan, 1 macro chan, PMoutage = 0.10 20 micro chan ' ' ' i i i i i i I i i i i i i _L _L 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0 13.0 14.0 15.0 Calls per minute per microcell Figure 6.12: Channel and Signal Blocking Probability of Overlay (with Macrocell Frequency Reuse) with and without Macrocell Radio Coverage Holes Figure 6.13 shows the channel and signal blocking probability as the microcell probability of signal outage varied. At the given arrival rate X = 6 calls/min/microcell, the overlay system without macrocell frequency reuse always outperforms the non-overlay sys-77 Overlapping and Overlaying with Radio Coverage Holes tem. For the frequency reuse macrocell overlay system, the channel and signal blocking of the (19,1) channel and signal configuration improves when the signal outage is greater than 0.04. The performance of the (17,3) configuration improves when the signal outage is greater than 0.06. 10 ca e 60 •o X U CJ X CM 60 C o o s o ca x o 10 10 Arrival rate = 6 calls per min per microcell when GOS = 2% for non-overlay system Plvloutage = 0.00 cr 17 micro chan, 3 macro chan — A — 18 micro chan, 2 macro chan — A- -19 micro chan, 1 macro chan 17 micro chan, 21 macro chan — EJ -18 micro chan, 14 macro chan — a- -19 micro chan, 7 macro chan 20 micro chan —•— 0.0 0.1 Signal Outage Probability Pmoutage 0.2 Figure 6.13 .-Channel and Signal Blocking Probability of Overlay Versus Probability of Signal Outage Figure 6.14 displays the maximum supportable arrival rate for various microcell channel assignments to the macrocell and varied radio hole probability. The reference P b e s c for the respective P o u t a ge w a s m e Pbesc ° f t n e non-overlay system when the GOS is two per-cent. The overlay system without frequency reuse overlay outperforms the non-overlay sys-tem. At a signal outage probability of 0.05, the maximum arrival rate for the macrocell frequency reuse overlay begins to out perform the non-overlay system. For signal outage probabilities of 0.10 and 0.25, the maximum supportable arrival rate is nearly constant when one to three microcell channels are assigned to a macrocell. At P m o u t a g e = 0.10, adding one 78 Overlapping and Overlaying with Radio Coverage Holes channel to the macrocell improves the maximum arrival rate by 0.74 calls/min/microcell or 12% while adding two channels causes a 13% improvement. B 60.0 cj Cu ej Cu c3 CJ <& > 'E < O Cu Cu 3 t/3 50.0 40.0 30.0 h 20.0 3 £ I O . O 0.0 1.0 2.0 Microcell Channels assigned to Macrocell • outage = 0.25, Pbesc = 0.25001, Macro Frequency Reuse - • -Poutage = 0.10, Pbesc = 0.10151, Macro Frequency Reuse Poutage = 0.05, Pbesc = 0.05606, Macro Frequency Reuse Poutage = 0.00, Pbesc = 0.02, Macro Frequency Reuse = 0.25, Pbesc = 0.25001 • outage Poutage = 0.10, Pbesc = 0.10151 Poutage = 0.05, Pbesc = 0.05606 Poutage = 0.00, Pbesc = 0.02 O-- A -Figure 6.14:Maximum Supportable Arrival Rate Versus the Number of Assigned Macrocell Channels 6.3.2 Handoff Activity The handoff activity is examined in figures 6.15 and 6.16. For both overlay sys-tems, the handoff activity for a given probability of signal outage decreases by a small amount. Notwithstanding the constant or improved blocking and handoff failure probabilities, macrocell overlay without frequency reuse causes handoff activity to decrease. Compared to the overlay without frequency reuse, the overlay with macrocell fre-79 Overlapping and Overlaying with Radio Coverage Holes quency reuse has a slightly lower handoff activity because of higher dropped call probabili-ties. cd U Cu o -o a cd X ) s 3 2 2.5 2.0 1.5 1.0 0.5 0.0 • Moutage ' •-<> Figure 6.15'-.HandoffActivity in Overlay Versus the Number of Assigned Macrocell Channels cd a .£? T3 a u a a cd X ! o o OH X cd X O UH OH "cd U u o< Cu O 10 o •- • - U J ^ . - r r . : 10 ,, 10 10 10 10 A -- Z 1 Arrival rate = 6 calls per min per microcell when GOS = 2% for non-overlaying system See Handoff Activity for Legend _L 1 2 Microcell Channels assigned to Macrocell Figure 6.16: Dropped Call Probability in Overlay Versus the Number of Assigned Macrocell Channels 80 CHAPTER 7: Summary and Conclusions Microcells are essential to satisfying the demands for high capacity, multimedia services and tetherless access to third generation wireless communication systems. This the-sis investigates extended overlapping microcells and overlaying microcells with macrocells to improve coverage, capacity and accessibility. Analytical models to evaluate performance were developed for overlapping microcells as well as for overlaying microcells with macro-cells. The advantages of the overlapping and overlaying architectures are most evident when signal availability considerations are included. 7.1 Overlapping and Overlaying Architectures Compared The various architectures for multiple base-site coverage are described in detail in previous chapters. Comparison of the performance of the overlapping and overlaying archi-tectures in terms of blocking probabilities, maximum supportable arrival rates and handoff activity are presented below for representative system parameter values. 7.1.1 Blocking Probabilities and Maximum Supportable Arrival Rates The blocking probabilities and maximum supportable arrival rates for the different overlapping and overlaying architectures are summarized in Table 7.1 for specific representa-tive system parameters. Dropped call probability performs similarly to the blocked call prob-ability and carried traffic represents calls not blocked. 81 Summary and Conclustions Table 7.1: Blocking Probability and Maximum Supportable Arrival Rate Comparisons P A outage = 0 P A outage = 0.10 Architecture Pbat X = 25.66 calls/ min/ km2 X (calls/ min/ km2) at Pb = 2% Pbsc a t x = 25.66 calls/ min/ km2 X (calls/ min/ km2) at p — x bsc ~~ 0.1015 Microcells (no overlap, no overlay) 0.02 25.66 0.1015 25.66 Single Isolated Group of Overlapping Microcells (R/ R h = 1.500) 0.001823 34.64 0.01136 35.50 Contiguous Scenario I Over-lapping Microcells (R/Rh = 1.134) 0.078465 18.90 0.095920 26.42 Contiguous Scenario II Overlapping Microcells (R/ Rh= 1.500) 0.032354 23.20 0.04217 34.43 Overlay (17,21) < le-5 36.82 0.01351 41.91 Overlay with Macrocell Fre-quency Reuse (17,3) 0.035610 23.74 0.089782 28.48 For overlapping microcellular coverage, the highest performing R/Rh ratios con-sidered, namely 1.134 and 1.500, are shown. For overlaying microcells with macrocells, the highest performing channel configurations, with up to three channels assigned per microcell, are shown. Without radio holes, blocking probabilities and maximum supportable arrival rates improve only by using an isolated group of overlapping microcells or the overlay without macrocell channel reuse. The gains achieved in these two special cases are attributed to the absence of any co-channel interference. The performance degradation of contiguous overlap-82 Summary and Conclustions ping microcells and overlay with macrocell frequency reuse is caused by the reduced trunking efficiency from the subdividing of channel sets to mitigate co-channel interference. There-fore, in the absence of radio holes and with full co-channel interference, the overlapping and overlaying architectures are not beneficial in terms of call blocking and carried traffic. When radio coverage effects are considered, the channel and signal blocking prob-abilities and maximum supportable arrival rates for both overlapping and macrocell overlay architectures improved performance. The isolated groups of overlapping microcells had a slightly lower P b s c compared to that for the (17,21) overlay; however, the (17,21) overlay sup-ported a higher message arrival rate (63% improvement) while the isolated overlapping micro-cell's arrival rate improved by only 38%. For the other architectures with full co-channel interference, the scenario II contiguous overlapping microcells had the lowest P b s c and the highest maximum supportable arrival rate (with a 35% improvement). Scenario I perfor-mance improved minimally. The maximum supportable arrival rate for an overlay with mac-rocell frequency reuse rose by 11%. An important additional factor is the installation and setup costs for each architec-ture. Scenario II growth requires the installation of, from two to nine, additional microcell base sites per cluster. Overlaying architectures require one new macrocell base station instal-lation per cluster. 7.1.2 Comparison of Handoff Activity Table 7.2 summarizes handoff activity performance of the architectures previously described at X = 25.66 calls/min/km . 83 Summary and Conclustions An isolated group of overlapping microcells is the only overlapping/overlaying configuration which significantly reduces the handoff activity below that which results from no overlap or overlay. Overlapping microcellular architecture with scenario I growth and the macrocell overlay systems reduce the handoff activity by a small amount. Scenario I is impractical because of the high blocking probability. With full co-channel interference sce-narios, only the macrocell overlay shows significantly reduced handoff activity. To achieve further reduced handoff activity, terminal (vehicle) speed sensitive channel allocation would have to be implemented. Table 7.2: Handoff Activity Comparisons P — 0 x outage ~~ " p — * outage — 0.10 Architecture H A (number of handoffs per call) H A (number of handoffs per call) Microcells (no overlap, no overlay) 2.01 1.69 Single Isolated Group of Overlapping Microcells (R/Rh =1.500) 1.35 1.31 Contiguous Scenario I Overlapping Microcells (R/Rh= 1.134) 1.71 1.56 Contiguous Scenario II Overlapping Microcells (R/Rh= 1.500) 1.97 1.68 Overlay (17,21) 1.90 1.65 Overlay with Macrocell Frequency Reuse (17,3) 1.88 1.56 84 Summary and Conclustions 7.2 Topics for Future Investigation As wireless communications moves closer towards multi-tiered cellular architec-tures, further research may be pursued in overlapping microcells and overlaying macrocells with microcells. The overlay investigated in this thesis was applicable to a TDMA or FDMA system where the channels are orthogonal. In CDMA systems, channels are not orthogonal. Additional research is required for the implementation of overlay in CDMA. Dynamic chan-nel assignment which assigns channels based on the measured levels of interference can potentially be more spectrum efficient. For overlapping microcells, the effects of DCA and the high co-channel interference could be investigated. DCA in the overlay macrocell chan-nels allows the system to adapt to non-homogeneous traffic. Adaptive antenna arrays enhance the system traffic performance of both overlap-ping microcells and overlaying microcells with macrocells. A directed beam not only uses the transmission power more efficiently but reduces co-channel interference. Further analysis of overlapping and overlaying cells would consider multimedia traffic, consisting of delay sensi-tive voice and error sensitive data. Speed estimation and speed-sensitive channel allocation could more significantly reduce the handoff activity of the overlay. 85 REFERENCES [I] "Mobile Warfare," Time Magazine, Vol. 149, No. 21, May 26, 1997. [2] "Wireless to Hit Growth Landmark," Wireless Week, July 21, 1997. [3] X. Lagrange, "Multitier Cell Design," IEEE Communications Magazine, Vol. 35, No. 8, August 1997, pp. 60-64. [4] M. Madfors, et. al, "High Capacity with Limited Spectrum in Cellular Systems," IEEE Communications Magazine, Vol. 35, No. 8, August 1997, pp. 38-45. [5] J.C.S., Cheung, M.A. Beach, and J.P. McGeehan, "Network Planning for Third-Genera-tion Mobile Radio Systems," IEEE Communications Magazine, Vol. 32, No. 11, November, 1994, pp. 54-55. [6] L.J. Greenstein, etal, "Microcells in Personal Communications Systems," IEEE Com-munications Magazine, Vol. 30, No. 12, December 1992, pp. 76-88. [7] M.V Clark, V. Erceg, and L.J. Greenstein., "Reuse Efficiency in Urban Microcellular Net-works," IEEE Transactions on Vehicular Technology, Vol. 46, No. 2, May 1997, pp. 279-287. [8] M.J. Feuerstein, et. al, "Path Loss, Delay Spread, and Outage Models as Functions of Antenna Height for Microcellular System Design," IEEE Transactions on Vehicular Technology, Vol. 43, No. 3, August 1994, pp. 487-497. [9] W.C.Y., Lee, Mobile Telecommunications Systems, McGraw Hill Publications, New York, 1995. [10]K.L.,Yeung, and S. Nanda, "Channel Management in Microcell/Macrocell Cellular Radio Systems," IEEE Transactions on Vehicular Technology, Vol. 45, No. 4, November 1996, pp 601-612. [II] A.J. Goldsmith, and L.J. Greenstein, "A Measurement-Based Model for Predicting Cover-age Areas of Urban Microcells," IEEE Journal on Selected Areas in Communications, Vol. 11, No. 7, September 1993, pp. 1013-1022. [12] M.M. Zonoozi, and P. Dassanayake, "User Mobility Modeling and Characterization of Mobility Patterns," IEEE Journal on Selected Areas in Communications, Vol. 15, No. 7, September 1997, pp. 1239-1252. • [13]H. Xie and S. Kuek, "Priority handoff analysis," Proceedings of 43rd IEEE Vehicular Technology Conference 1993, (VTC'93), New Jersey, May 1993, pp. 855-858. 86 [14]L.-R. Hu, and S.S. Rappaport, "Personal Communication Systems Using Multiple Hierar-chical Cellular Overlay," IEEE Journal on Selected Areas in Communications, Vol. 13, No. 2, February 1995, pp. 406-415. [15]B. Jabbari, and W.F. Fuhrmann, "Teletraffic Modeling and Analysis of Flexible Hierarchi-cal Cellular Networks with Speed-Sensitive Handoff Strategy," IEEE Journal on Selected Areas in Communications, Vol. 15, No. 8, October 97, pp. 1539-1548. [16] A.J.M. Ransom, "Handoff considerations in microcellular systems planning," Proc. 6th Int'l Symposium on Personal Indoor and Mobile Radio Communications, Toronto, Canada, September 1995, pp. 805-808. [17]N. Srivastava, and S.S. Rappaport, "Models for Overlapping Coverage Areas in Cellular and Micro-Cellular Communication Systems," Proc. IEEE Globecomm Conference 1991, Phoenix, USA, pp. 890-894. [18]T.-P. Chu, and S.S. Rappaport, "Overlapping coverage and channel rearrangement in microcellular communication systems," IEE Proc.-Commun., Vol. 142, No. 5, October 1995, pp. 323-332. [19]D. Bertsekas, and R. Gallager, Data Networks, Prentice-Hall Inc., Upper Saddle River, New Jersey, 1992. [20]M.D. Yacoub, Foundations of Mobile Radio Engineering, CRC Press, Boca Raton, Flor-ida, 1993. [21]C.-L. I, L. Greenstein, and R.D. Gitlin, "A Microcell/Macrocell Cellular Architecture for Low- and High-Mobility Wireless Users," IEEE Journal of Selected Areas in Commu-nications, Vol. 11, no. 6, 1993, pp. 885-891. [22]C.W. Sung and W.S. Wong, "User speed estimation and dynamic channel allocation in hierarchical cellular system," Proc. IEEE Vehicular Technology Conference 1994, Stockholm, Sweden, Vol. VTC-94. [23]X. Lagrange and P. Godlewski, "Performance of Hierarchical Cellular Network with Mobility-Dependant Hand-Over Strategies," Proceedings of the IEEE Vehicular Tech-nology Conference, Atlanta, 1996, pp. 1868-1872. [24]S.A. El-Dolil, W.-C. Wong, and R. Steele, "Teletraffic performance of Highway Micro-cells with Overlay Macrocells," IEEE Journal of Selected Areas in Communications, Vol. 7, No. 1, 1989, pp 71-78. [25]R. Beraldi, S. Marano, and C. Mastroianni, "A Reversible Hierarchical Scheme for Micro cellular Systems with Overlaying Macrocells," Proceedings of the IEEE Infocom, San Francisco, 1996, pp. 51-58. 87 [26]S.S. Rappaport and L.-R. Hu, "Microcellular Communication Systems with Hierarchical Macrocell Overlays: Traffic Performance Models and Analysis," Proceedings of IEEE, Vol. 82, No. 9, 1994, pp. 1383-1397. [27]I. Katzela, and M. Naghshineh, "Channel Assignment Schemes for Cellular Mobile Tele-communication Systems: A Comprehensive Survey," IEEE Personal Communications Magazine, June 1996, Vol. 3, No. 3, pp. 10-31. [281L.C. Godora, "Applications of Antenna Arrays to Mobile Communications, Part I: Perfor-mance Improvement, Feasibility, and System Considerations, " Proceedings of the IEEE, Vol. 85, No. 7, July 1997, pp. 1031-1060. [29]H. H. Xia, et. ai, "Radio Propagation Characteristics for line of sight microcellular and personal communications," IEEE Transaction on Antenna Propagation, Vol. 41, No. 10, October 1993, pp. 1439-1447. [30JT.-P. Chu, and S.S. Rappaport, "Generalized Fixed Channel Assignment in Microcellular Communications Systems," IEEE Transactions on Vehicular Technology, Vol. 43, No. 3, August 1994, pp. 713-721. 88 APPENDIX A: Handoff Activity Simplification The handoff activity of overlapping microcells was determined in section 3.33. Equation (3.6) considered the two cases for handoff failure and successful call completion. The equation is copied and the simplication steps are shown: ( = 0 Phfm + ( 1 _ Phfm) ^ Mm V ~ M n > + ngJ - ( p t f m The following well know series equation is applied. P~ + ^HmPhfrr, (A.2) X na" = — * — 2 (A.3) „ = 0 ( 1 - ^ Substituting (A.3) into (A.2) results in the following. (A.4) = (Phfm(» + VHm) + V-PhfJV) -2 (A.5) (» + VHm-VHm(l-Phfin)) (A.6) The simplified equation (A.6) for handoff activity of overlapping microcells is shown in chap-ter 3. 89 APPENDIX B: List of Abbreviations and Acronyms CDMA Code Division Multiple Access DCA Dynamic Channel Assignment DTX Discontinuous Transmission FCA Fixed Channel Assignement FDMA Frequency Division Multiple Access GSM Global System for Mobile Communications IMT-2000 Internation Mobile Telecommunications 2000 LOS Line of Sight ML Maximum Likelihood MMSE Minimum Mean Square Error n.e.d. Negative Exponential Distribution NLOS Non-line of Sight PCS Personal Communications Services TDMA Time Division Multiple Access UMTS Universal Mobile Telecommunications Systems 90 APPENDIX C: Frequency Reuse Derivation for Contiguous Microcells Various cellular cluster configurations are shown in figure C l , where R n is the hexagonal radius. Figure CJ .'Common Cluster Configurations for Hexagonal Cellular Patterns The area of a hexagonal cell (H) is: U _ 3^3^ 2 ( C l ) as seen in figure C l . Likewise the area of the corresponding hexagonal cluster area (C) is shown in equation (C.2) where D is the distance between the centers of two co-channel cells. C = ijlf D/2 V 2 V ( V 3 ) / 2 (C.2) 91 The number of cells per cluster is obviously: N = | . (C.3) Substituting (Cl) and (C.2) into (C.3) and cancelling common terms gives: " • f f l -If the hexagonal radius is fixed, then: < N = mD2, (C.5) where m is a constant. The result of equating the constant m from two systems with different N and D is shown in equation (C.6). The simplified result (C.7) is used for equation (3.35). = (C6) D\ D2 N2 = NX (C.7) 92