Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Performance of IEEE 802.11 MAC protocol over a wireless LAN with distributed radio bridges Chow, Cupid C. 1999

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1999-0207.pdf [ 5.4MB ]
Metadata
JSON: 831-1.0065166.json
JSON-LD: 831-1.0065166-ld.json
RDF/XML (Pretty): 831-1.0065166-rdf.xml
RDF/JSON: 831-1.0065166-rdf.json
Turtle: 831-1.0065166-turtle.txt
N-Triples: 831-1.0065166-rdf-ntriples.txt
Original Record: 831-1.0065166-source.json
Full Text
831-1.0065166-fulltext.txt
Citation
831-1.0065166.ris

Full Text

Performance of I E E E 802.11 M A C Protocol over a Wireless L A N with Distributed Radio Bridges by CUPID C. CHOW B. Sc. (Electrical Eng.) The California State University, Fresno, C A , USA, 1994 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 ENGINEERING We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA April 1999 ©Cupid C. Chow, 1999 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 of r^LJfecTRltAu £ -Nfr / iUfeg£/A^ The University of British Columbia Vancouver, Canada DE-6 (2/88) Abstract In this thesis, the performance, in terms of throughput and access delay, of the proposed distributed Wireless Loca l Area Network employing the I E E E 802.11 Med ium Access Control protocol is analyzed extensively by computer simulations. The proposed network is based on using the same frequency for the entire coverage area with multiple receivers or radio bridges (RBs), with or without capture effect at each receiver. Different channel environments, including the Additive White Gaussian Noise ( A W G N ) channel, log-normal shadowing channel, Rayleigh fading channel, and log-normal shadowing plus Rayleigh fading channel, are considered in the simulations. The effects of various parameters are also studied. The results show that the performance can be improved significantly by using more R B s except for A W G N channel. The performance improvement of multiple R B s over one R B depends on the system parameters. In the log-normal shadowing plus Rayle igh fading channel, the throughput with 4 RBs and no capture can be improved by at least 120% over one R B . It is found that multiple R B s are more effective in a system with large number of stations and for channels severely degraded by shadowing or fading. Wi th the use of capture, the performance can be improved by multiple RBs . But the performance improvement of multiple R B s is about the same as no capture cases. The effect of capture ratio is studied. It is found that the performance increases with decreasing the capture ratio. The performance of one and two dimensional models are found to be slightly different. The throughput does not change with packet size; however, the access delay degrades with packet size. Finally, it is found that the performance is not significantly affected by the length of the acknowledgment time-out. ii Table of Contents Abstract i i List of Tables vi List of Figures vii Acknowledgment xi Chapter 1 Introduction 1 1.1 Motivations and Objectives of the Thesis 2 1.2 Outline of the thesis 5 Chapter 2 Background 6 2.1 I E E E 802.11 Architecture Components 6 2.1.1 Adhoc Network 7 2.1.2 Infrastructure Network 7 2.2 I E E E 802.11 W L A N M A C Protocol 8 2.2.1 Distributed Coordination Function (DCF) 8 2.2.2 Inter-Frame Space (TFS) 9 2.2.3 D C F Access Procedure 12 2.2.4 Backoff Procedure 14 2.2.5 Directed M P D U Transfer Procedure using R T S / C T S 15 2.2.6 R T S / C T S Recovery Procedure and Retransmit Limits 17 2.2.7 Broadcast and Multicast M P D U Transfer Procedure 17 2.2.8 A C K Procedure 17 2.2.9 Duplicate Detection and Recovery 18 2.3 I E E E 802.11 Physical Layer 18 iii Chapter 3 Distributed W L A N Architecture and the Functional Evaluation of the I E E E 802.11 M A C Protocol 19 3.1 Distributed W L A N Architecture 19 3.2 Functional Evaluation of the Distributed W L A N Employing the I E E E 802.11 M A C Protocol 20 3.2.1 A n Example of the Distributed W L A N 21 3.2.2 The Advantages and Disadvantages of the proposed W L A N using I E E E 802.11 M A C protocol 23 3.3 Design of the Simulation Model 24 3.3.1 The O P N E T Simulation Environment 24 3.3.2 The Node Model 25 3.4 The Channel Models 26 3.4.1 Path Loss Model 26 3.4.2 Rayleigh Fading 27 3.4.3 Log-normal Shadowing 28 3.5 Station Location 29 3.6 Design of the I E E E 802.11 M A C Process 31 3.6.1 The Finite State Machine of the 802.11 M A C Protocol 32 3.6.2 The Simulation Parameters 35 3.6.3 Data Collection 37 Chapter 4 Simulation Results and Discussion - Part 1 39 4.1 Simulation Descriptions 39 4.1.1 Performance Measurement 39 4.1.2 Network Arrangement 39 4.1.3 Error-Free Channels 40 IV 4.2 Simulation Model Validation 41 4.3 A W G N Channel 42 4.3.1 Simulation Results 43 4.4 Rayleigh Fading Channel 46 4.4.1 Simulation Results 46 4.5 Log-Normal Shadowing Channel 55 4.5.1 Simulation Results 55 Chapter 5 Simulation Results and Discussion - Part 2 59 5.1 Log-normal Shadowing Plus Rayleigh Fading Channel With N o Capture 59 5.2 Log-normal Shadowing Plus Rayleigh Fading Channel With Capture 65 5.2.1 Capture Model 65 5.2.2 Simulation Results 67 5.3 Other System Parameters 74 5.3.1 Two-dimensional Model 74 5.3.2 Packet Length 80 5.3.3 Effect of ACK_Timeout 81 5.3.4 Effect of Capture Ratio 84 Chapter 6 Conclusions 86 Glossary 90 Bibliography 94 V List of Tables Table 2.1 Lists of the PHY Attributes of FHSS, DSSS and IR 11 Table 3.1 Summary of all states and transitions of the 802.11 MAC protocol 33 Table 3.2 The simulation parameters 36 Table 3.3 The Statistics collected by the process 38 VI List of Figures Figure 2.1 Basic Service Set 6 Figure 2.2 A n infrastructure network 8 Figure 2.3 B asic Access Method 13 Figure 2.4 Directed M P D U transfer procedure 13 Figure 2.5 Backoff Procedure 15 Figure 2.6 Directed M P D U Transfer using RTS/CTS 16 Figure 3.1 The distributed architecture for W L A N , with multiple radio bridges 20 Figure 3.2 A n example of the W L A N employing RBs 21 Figure 3.3 The Time-line diagram of the example 23 Figure 3.4 The node model for the W T 25 Figure 3.5 Rayleigh C D F vs. Simulation results 28 Figure 3.6 Four sets of network arrangement 30 Figure 3.7 The state machine of the 802.11 M A C protocol 32 Figure 4.1 Normalized throughput versus normalized offered load for 5, 10 and 20 stations with C W m i n = 32 and CWmax = 256 for adhoc and infrastructure networks 41 Figure 4.2 Access delay versus normalized offered load with different values of CW(min/max) for 5 stations for adhoc and infrastructure networks 42 Figure 4.3 Normalized throughput versus normalized offered load for 15 W T s in A W G N channel using (a) Method A and (b) Method B 44 Figure 4.4 Normalized throughput versus normalized offered load for 30 W T s in A W G N channel using (a) Method A and (b) Method B 44 Figure 4.5 Access delay versus normalized offered load for 15 W T s in A W G N channel using (a) Method A and (b) Method B 45 vii Figure 4.6 Access delay versus normalized offered load for 30 W T s in A W G N channel using (a) Method A and (b) Method B . 46 Figure 4.7 Normalized throughput versus normalized offered load for 15 W T s in Rayleigh fading channel using (a) Method A and (b) Method B 48 Figure 4.8 Access delay versus normalized offered load for 15 WTs in Rayleigh fading channel using (a) Method A and (b) Method B 49 Figure 4.9 (a) Normalized throughput and (b) access delay versus normalized offered load for 30 WTs in Rayleigh fading channel using Method A 49 Figure 4.10 Saturated throughput versus transmitted power in Rayleigh fading channel using Method A for (a) 15 WTs and (b) 30 WTs 50 Figure 4.11 The coverage area of one R B for 15 WTs at low Tx 51 Figure 4.12 The coverage area for the 4 RBs with 15 W T S at low Tx 52 Figure 4.13 Access delay versus transmitted power in Rayleigh fading channel using Method A for (a) 15 WTs and (b) 30 W T s 53 Figure 4.14 The saturated throughput distribution on each of the 15 W T for (a) one R B with Tx = 50 mW, (b) one R B with Tx = 150 mW, (c) 4 R B s with T x = 50 m W and (d) 4 RBs with Tx = 150 mW. 54 Figure 4.15 Saturated throughput versus in log-normal shadowing channel for 15 W T s using Method A and Method B 55 Figure 4.16 Access delay versus in log-normal shadowing channel for 15 W T s using Method A : 57 Figure 4.17 (a) Saturated Throughput and (b) access delay versus using Method A for 30 WTs 58 Figure 5.1 Normalized throughput versus normalized offered load in log-normal shadowing plus Rayleigh fading channel for (a) 15 W T s and (b) 30 W T s . . . 60 Figure 5.2 Access delay versus normalized offered load iri log-normal shadowing plus Rayleigh fading channel for (a) 15 WTs and (b) 30 W T s 61 Figure 5.3 Saturated throughput as a function of transmitted power with dB in log-normal plus Rayleigh fading channel for (a) 15 W T s and (b) 30 W T s 62 viii Figure 5.4 Access delay as a function of transmitted power with dB for (a) 15 W T s and (b) 30 W T s 63 Figure 5.5 (a) Saturated throughput and (b) access delay as a function for 30 WTs . 64 Figure 5.6 The capture model 67 Figure 5.7 (a) Normalized throughput and (b) access delay versus normalized offered load in log-normal shadowing and Rayleigh fading channel with capture for 15 WTs using Method A 68 Figure 5.8 The throughput distribution of 15 WTs with no capture for (a) one R B , (b) 2 RBs , (c) 3 RBs and (d) 4 RBs 70 Figure 5.9 The throughput distribution of 15 WTs with capture for (a) one R B , (b) 2 RBs , (c) 3 R B s and (d) 4 RBs ....71 Figure 5.10 (a) The saturated throughput and (b) access delay as a function of transmitted power in log-normal plus Rayleigh fading channel at dB for 15 W T s using Method A with capture 72 Figure 5.11 (a) The saturated throughput and (b) access delay as a function of in log-normal plus Rayleigh fading channel at Tx = 100 m W for 15 W T s using Method A with capture 73 Figure 5.12 The station location for 15WTs and different number of R B s in a two-dimensional model 74 Figure 5.13 (a) Normalized throughput and (b) access delay versus normalized offered load in log-normal shadowing plus Rayleigh fading channel for no capture in two-dimensional model..... 75 Figure 5.14 The saturated throughput distribution for 15 WTs on each W T with no capture in a two-dimensional model for (a) one R B , (b) 2 RBs , (c) 3 R B s and (d) 4 RBs 76 Figure 5.15 (a) Normalized throughput and (b) access delay versus normalized offered load in log-normal shadowing,, and Rayleigh fading channel at Tx = 100 m W with capture for 15 WTs using Method A in a two-dimensional model 77 Figure 5.16 The saturated throughput distribution for 15 WTs on each W T with capture in a two-dimensional model for (a) one R B , (b) 2 RBs , (c) 3 R B s and (d) 4 RBs 79 Figure 5.17 (a) Saturated throughput and (b) access delay as a function of transmitted power in log-normal plus Rayleigh fading channel at dB for 15 W T s using Method A with capture in a two-dimensional model 80 Figure 5.18 (a) Saturated throughput and (b) access delay as a function of packet size in log-normal shadowing,, and Rayleigh fading channel at Tx = 100 m W with capture for 15 WTs using Method A 81 Figure 5.19 Saturated throughput as a function of ACK_Timeout for 15 W T s with no capture in log-normal shadowing plus Rayleigh fading channel using Method A and Method B 82 Figure 5.20 Saturated throughput as a function of ACK_Timeout for 15 W T s with capture in log-normal shadowing plus Rayleigh fading channel using Method A and Method B 83 Figure 5.21 Access delay as a function of ACK_Timeout for 15 W T s with no capture in log-normal shadowing plus Rayleigh fading channel using Method A (a) and Method B (b) 83 Figure 5.22 Access delay as a function of ACK_Timeout for 15 W T s with capture in log-normal shadowing plus Rayleigh fading channel using Method A (a) and Method B (b) 84 Figure 5.23 (a) Saturated throughput and (b) access delay as a function of c for 15 WTs with capture in log-normal shadowing plus Rayleigh fading channel.. 85 Acknowledgment I would like to express sincere gratitude to my research supervisor, Dr. V. C . M . Leung, who has provided me with many helpful suggestions, constant supervision, and invaluable guidance in both my research and the preparation of this document. I would like to thank Dr. E . Casas for his help in reviewing and verifying the simulation model. X I Chapter 1 Introduction Nowadays, Local Area Networks (LANs) are widely used to interconnect computers to each other, printers, servers, and the Internet. However, installing and maintaining the wiring of a L A N is very expensive, and in some instances technically very challenging. Wireless Local Area Networks (WLANs) provide an attractive alternative which enable flexible location of terminals and can avoid re-wiring when terminals are relocated. WLANs employ wireless communications for interconnecting wireless terminals among themselves and to wired LANs. A Medium Access Control (MAC) protocol is used by a W L A N to enable multiple terminals to efficiently share the same radio or optical wireless channel. For extending the connectivity of a W L A N , a backbone L A N can be employed. W L A N usually operates in the indoor environment. Fading and shadowing will degrade the performance of the system. Techniques such as diversity reception, equalization, and spread spectrum signalling can overcome the propagation effects to enable reliable communications [1]. When multi-access interference occurs, all the transmitted packets sent by several users collide at the receiver and all the packets are destroyed. Therefore, the throughput efficiency is limited. Fortunately, the capture effect allow the receiver to capture one of several colliding packets correctly due to different signal strengths when the packet reaches the receiver [2] [3]. Therefore, the capture effect can increase the throughput efficiency of the system. To provide coverage in a large building using low power radio transmissions, the central-ized architecture can be used. As in a cellular mobile telephone system, the building area is divided into cells which are connected through a backbone L A N . Different radio channels are assigned to adjacent cells to minimize interference. An alternative approach is the distributed 1 Chapter 1 Introduction 2 architecture proposed in [1][4][5] which is based on using the same radio channel over the entire coverage area. This architecture avoids the need for frequency coordination and does not need the complicated handoff mechanisms as required in the cellular architecture. Besides, it enables macro-diversity, and it is expected to give improved performance in the presence of propagation effects and hidden terminals. The performance evaluation of the distributed architecture by applying the I E E E 802.11 M A C protocol is the main goal of this thesis. 1.1 M o t i v a t i o n s a n d O b j e c t i v e s o f t h e T h e s i s W L A N is an integral and important component in the emerging wireless office. W L A N not only removes the hassle of installing and maintaining cables, but it also provides communica-tion where computers are no longer bounded to a particular location. Currently, W L A N s utilize two technologies, infrared and radio access. Both of them require proper access control protocols to allow multiple users to share the channel. In particular, radio access technology requires a proper knowledge of the environment in which the radio W L A N is to be deployed. Therefore, the performance of the radio W L A N systems are very much influenced by the radio propagation effects, especially in indoor environments, and by the M A C protocols as wel l . Accordingly, in 1990, the I E E E Project 802.11 W L A N Standards Committee was established to recommend an international standard [6] [7] for W L A N s . The scope of the work is to develop M A C layer and Physical ( P H Y ) layer specifications for wireless connectivity for fixed, portable and moving stations within a local area. The standard describes the M A C procedures to support the asynchro-nous and time-bounded M A C Service Data Uni t ( M S D U ) delivery services. The basic M A C protocol is a Distributed Coordination Function (DCF) that allows for medium sharing through the use of Carrier Sense Mult iple Access with Coll is ion Avoidance ( C S M A / C A ) and a random backoff time. Chapter 1 Introduction > 3 The architectures of the W L A N s can be addressed in two major areas. They are adhoc and infrastructure networks. Adhoc network is formed by a minimum of two Wireless Terminals (WTs) that are close enough to form direct connections without pre-planning. Both simulations and analysis have been done for this kind of networks. In [8], simulations results are presented for throughput and delays for 10 and 25 stations with different values of inter-frame spacing periods and transmission speeds. It shows that there is no significant difference in normalized throughput for various transmission speed, and the choice of inter- frame space values depends on the type of network. In [9][10][11], simple modifications for the backoff mechanism using weighted probability are presented. A simulation is done with 8 stations with or without hidden terminals, and the results show that the throughput increases by 25% with the modified backoff scheme. However, the optional Ready to Send/Clear to Send (RTS/CTS) mechanism does not completely solve the hidden terminal problem, and only some improvements can be achieved. In [12], voice traffic in the Contention Free Period (CFP) and data traffic in the Contention Period (CP) over an 802.11 W L A N are simulated. The results show that the cooperation of the C P and the C F P limits the number of possible voice conversations and the maximum payload size of the data. In [13][14], a simple analytical model to compute the saturation throughput performance in the presence of a finite number of terminals with the assumption of ideal channel is presented. Besides, an adaptive contention window mechanism is proposed and the optional R T S / C T S is investigated with hidden terminals by simulations. The results show that the basic C S M A / C A suffers from several performance drawbacks. With the adaptive contention window mechanism, the system becomes stable, and it outperforms the standard protocol when the network load and the number of mobile stations are high. In [15][16], the possibility of power capture, and presence of hidden stations is considered. System throughput is computed and fairness properties are Chapter I Introduction 4 evaluated. The impact of spatial characteristics on the performance of the system and that observed by individual stations is determined. The simulation results obtained for throughput for varying values of capture parameter and probability of hidden stations validate the analytical approach. In contrast to the adhoc network, infrastructure networks contain Access Points (APs) which are stations that provide access to the backbone L A N . The centralized architecture of the infrastructure networks is used as in a cellular mobile telephone system, with the building area divided into cells which are connected through a backbone L A N . This type of network is studied in [17] with Rayleigh fading, shadowing and power capture effect by using the M A C protocols based on I E E E 802.11. The channel throughput and packet delay are analyzed. In [18] [19], the natural hidden terminals problem of this network is analyzed. The analysis includes the effects of the difference in the coverage of the A P and the mobile terminals as well as the capture effect caused by the near far problem. Mos t of the research and commercial products employ the centralized architecture, such as the Wireless In Bui ld ing Network (WIN) by Motoro la [20], which can handle a combined network traffic demand of approximately 3 Mbps. In [21] [22] [23], the performance of the M A C layer is determined by simulating asynchro-nous data for both adhoc and infrastructure networks. In [21], the sensitivity of network geometry, traffic flow, queueing size, channel models, hardware, backoff parameters and hidden terminal problem on performance are studied. It turns out that the connectivity and the hidden terminal parameter can be used to predict the performance of the system. For [22] [23], performance results are provided for packetized data and a combination of packetized data and voice over the W L A N . The results show that the 802.11 W L A N can achieve a reasonably high efficiency when the Chapter 1 Introduction 5 medium is almost error-free, but may degrade under fading, and it also shows that time-sensitive traffic such as packet voice can be supported together with packet data. In [24] [25], they both present the modeling and performance analysis of a distributed W I N with multiple receivers and Direct Sequence Spread Spectrum (DSSS) . The first one use the Carrier Sense Multiple Access and the second one use the slotted A L O H A medium access control protocol. Besides, in [26], a one-dimensional network model with multiple access ports is studied by using slotted A L O H A medium access control. Therefore, the proposed distributed W L A N s has not been studied using the I E E E 802.11 standard M A C protocol. The objective of this thesis is to investigate by simulations the throughput and delay performance of the I E E E 802.11 M A C protocol employing C S M A / C A in W L A N s employing distributed radio bridges (RBs) [1]. We focus on the uplink channel between W T s and the multiple R B s , subject to Rayleigh fading and shadowing, with and without capture effect at the receivers. 1.2 Outline of the thesis In this thesis, an overview of the I E E E 802.11 W L A N M A C protocol is given in Chapter 2. Chapter 3 presents the architecture and a functional evaluation of the W L A N employing distributed R B s with the I E E E 802.11 M A C protocol. A s we l l , the simulation model using Optimized Network Engineering Tools (OPNET) [27] is presented. Chapter 4 discusses the model validation and the simulation results for the Additive White Gaussian Noise ( A W G N ) channel, the Rayleigh fading channel and the L o g normal shadowing channel. Chapter 5 presents the simula-tion results on the performance of the combination of the A W G N , Rayleigh fading and shadowing channel with and without capture effect. Conclusions are given in Chapter 6. Chapter 2 Background In this chapter, the architecture components to provide I E E E 802.11 W L A N that support station mobility transparently to upper layers w i l l be described. Then, the functional description of the M A C layer w i l l be presented. 2.1 I E E E 8 0 2 . 1 1 A r c h i t e c t u r e C o m p o n e n t s The basic building block of an 802.11 L A N is the Basic Service Set (BSS) which is a set of stations controlled by a single coordination function. Figure 2.1 shows two B S S s , each of which has two stations. The ovals is used to define a basic service area (BSA) as the coverage area over which the stations can communicate with each other. The independent B S S is the most basic type of 802.11 L A N . A minimal 802.11 L A N can consist of only two stations. BSS 1 802.11 Components BSS 2 Figure 2.1 Basic Service Set. 6 Chapter 2 Background 7 The architectures of the W L A N s can be addressed in two major areas. They are adhoc and infrastructure networks. 2.1.1 A d h o c N e t w o r k Figure 2.1 shows two independent BSSs . In each B S S , the stations are close enough to form a direct connection without pre-planning. This type of operation characterizes as adhoc network. Since, all the stations must be close enough for communications, the range of this type of network is limited. 2.1 .2 I n f r a s t r u c t u r e N e t w o r k In contrast to the adhoc network, infrastructure networks can provide wireless users with specific services and range extension. A n infrastructure network contains one or more A P s which are stations that provide access to the Distribution System (DS). The D S is used to interconnect a set of BSSs to create an Extended Service Set (ESS). Stations within an E S S can communicate and mobile stations may move from one B S S to another (within the same ESS) . The B S S s may overlap or be disjoint. The Extended Service Area (ESA) which is not smaller than a B S A , is the area within which members of an ESS can communicate. The D S can be thought of as a backbone network that is responsible for M A C level transport of M S D U s . A n E S S can provide access for wireless users into a wired network via a device known as a portal. The Station Services (SS) include authentication, deauthentication, privacy and M S D U delivery by every station. The Distri-bution System Services (DSS) include association, disassociation, distribution, integration and reassociation and are used to cross media and to address space logical boundaries. Figure 2.2 shows a simple ESS with two BSSs, a D S , and a portal access to a wired L A N . Chapter 2 Background 8 Figure 2.2 An infrastructure network. 2.2 I E E E 8 0 2 . i l W L A N M A C P r o t o c o l The basic M A C protocol specified in I E E E 802.11 is a D C F that allows for medium sharing through the use of C S M A / C A and a random backoff time. The D C F shall be implemented in all stations and A P s . A n alternative access method is a Point Coordination Function (PCF) which may be implemented on top of the D C F . This access method uses a point coordinator to determine which station currently has the right to transmit. Since this paper mainly focuses on the D C F part, the system modeling and performance analysis w i l l be based on the D C F . However, both the D C F and the P C F may coexist without interference. 2.2.1 D i s t r i b u t e d C o o r d i n a t i o n F u n c t i o n ( D C F ) The fundamental access method of the 802.11 M A C is a D C F knows as C S M A / C A . The Chapter 2 Background 9 C S M A / C A is designed to reduce the collision probability between multiple stations accessing a medium. Carrier sensing (CS) can be performed both through physical and virtual mechanisms. The physical CS mechanism shall be provided by the P H Y and the details of CS are provided in the individual P H Y specification section [6] [7]. The virtual CS mechanism is achieved by reserv-ing the medium as busy through the exchange of special small RTS and C T S frames before the actual data frame. The RTS and C T S frames activate the Network Allocation Vector ( N A V ) which contains the duration field for the period of time that the medium is to be reserved to transmit the actual data frame. This information is distributed to all stations within detection range of both the source and the destination stations in order to solve the hidden station problem. However, this can only be used for directed frames. When multiple destinations are addressed by broadcast/ multicast frames, then this mechanism is not used. 2.2 .2 I n t e r - F r a m e S p a c e ( I F S ) The time interval between frames is called the IFS. According to the priority levels for access to the wireless media, there are four different IFSs. The first one is the Short Inter-frame Space (SIFS), and it has both a minimum and maximum specification which depends on the P H Y layer. This SIFS is used for an acknowledgment ( A C K ) frame, a C T S frame, and between frames. The second type is the P C F - I F S (PIFS) and is used only by the P C F to send any of the C F P frames. After a station operating under the P C F detects the medium free for the period PIFS, it is allowed to transmit. The third type is the DCF- IFS (DIFS) and is used to transmit asynchronous M A C Protocol Data Units (MPDUs) . After a station operating under the D C F detects the medium free for the period DIFS, it is allowed to transmit, as long as it is not in a backoff period. Extended Inter-frame Space (EIFS) is used by the D C F when the P H Y has indicated to the M A C that a frame transmission was begun that did not result in the correct reception of a complete M A C Chapter 2 Background 10 frame with a correct Frame Check Sequence (FCS) value. The SIFS Time and Slot Time are defined as follows, and are fixed per P H Y . SIFS Time = RxRFDelay + RxPLCPDelay + MACProcessingDelay + RxTxTurnaroundTime. Slot Time = C C A Time + RxTxTurnaroundTime + AirPropagationTime + MACProcessingDelay. (2.1) The PIFS and DIFS are derived by: PIFS = SIFS Time + Slot Time DIFS = SIFS Time + 2 x Slot Time (2.2) The EIFS is derived from the SIFS and the DIFS and the length of time it takes to transmit an A C K frame at a Mbit/s by the following equation: EIFS = SIFS Time + (8 x A C K S i z e ) + PreambleLength PLCPHeaderLength + DIFS. ( 2 ' 3 ) where A C K S i z e is the length, 14 bytes, the SIFS Time, Slot Time, RxRFDelay , R x P L C P D e l a y , MACProcessingDelay, RxTxTurnaroundTime, C C A Time, AirPropagationTime, MACProces s -ingDelay, PreambleLength and PLCPHeaderLength are different for the three P H Y layer specifi-cations. The following table lists the default values: Chapter 2 Background Table 2.1 Lists of the PHY Attributes of FHSS, DSSS and IR. PHY Attribute Default Value Slot Time 50 [is CCA Time 27 us RxTxTurnaroundTime 20 [is RxPLCPDelay 2 [is SIFS Time 28 [is FHSS RxRFDelay 4 \is MACProcessingDelay 2 [is PreambleLength 96 \xs PLCPHeaderLength 32 p.* AirPropagationTime 1 [IS CW T,min 15 CW ^ T" max 1023 Slot Time 20 [is CCA Time < 15 [is RxTxTurn arou ndTime <5[is RxPLCPDelay implementation dependent SIFS Time 10 [IS DSSS RxRFDelay implementation dependent MACProcessingDelay not applicable PreambleLength 144 bits PLCPHeaderLength 48 bits CW " min 31 CW ^- " max 1023 Chapter! Background 12 PHY Attribute Default Value Slot Time 8 \id SIFS Time 7 [is RxTxTurnaroundTime 0 [IS RxPLCPDelay 1 [is CCA Time 5 lis IR RxRFDelay implementation dependent MACProcessingDelay 2 lis PreambleLength 16 lis (1 Mbps) 20 lis (2 Mbps) PLCPHeaderLength 41 lis. (1 Mbps) 25 u.s (2 Mbps) CW 63 CW -^ " max 1023 2.2 .3 D C F A c c e s s P r o c e d u r e The basic access refers to the mechanism a station uses to determine whether it has permission to transmit. A station with a pending M P D U may transmit when it detects a free medium for greater than or equal to a DIFS time. If the medium is busy, the random backoff time algorithm w i l l be followed. Figure 2.3 shows the basic access mechanism. Figure 2.4 shows the basic transmission procedure for the source, destination and other stations. The source station follows the basic access procedure and transmits the data frame. The destination station returns an A C K frame and other stations defer and follow the backoff procedure. Chapter 2 Background 13 Immediate access when medium is free >= DIFS DIFS Busy Medium DIFS PIFS SIFS Contention Window «< * Next Frame Defer Access Select Slot and Decrement Backoff as long as medium is idle Figure 2.3 Basic Access Method Source DIFS DATA Destination SIFS ACK Other DIFS Contention Window //// Defer Access Backoff after Defer Figure 2.4 Directed MPDU transfer procedure. Chapter 2 Background 14 2.2 A B a c k o f f P r o c e d u r e When a station intends to transmit an M P D U and finds that the medium is busy, the backoff procedure w i l l be followed. This procedure has a backoff timer which is selected from (2.4). The timer wi l l decrement only when the medium is free for DIFS time and it w i l l be frozen during the busy period. When the timer goes to zero, M P D U transmission begins i f the medium is still idle. Figure 2.5 shows the backoff procedure of five stations trying to send their own M P D U s and some of them have to go through the backoff procedure while one of them is sending its own M P D U . In order to provide fairness of access for all the stations, once a station has transmitted a frame and has another frame ready to transmit, it has to perform the backoff procedure. The backoff time is determined as follows: where the Slot Time is defined in equation (2.1) and the default values are listed in Table 2.1. The Contention Window ( C W ) has an ini t ial value of C W m i n for every M P D U and is increased exponentially after every retransmission attempt up to C W m a x . This is done to improve the stabil-ity of the access protocol under high load conditions. The default values for the C W m i n and C W m a Y are in Table 2.1. Backoff Time = Random () * Slot Time (2.4) where: Random () = Pseudo random integer drawn from a uniform distribution over the interval [ 0, C W ] C W = A n integer between C W m i n and C W (2.5) Chapter 2 Background 15 = Contention Window | | = Backoff DIFS DIFS = Remaining Backoff = M P D U arrival DIFS Frame Station A Station B i Defer 4 Deferi Station C Station Defer Station E Backoff Frame Dbfed Frame Frame Figure 2.5 Backoff Procedure 2.2 .5 D i r e c t e d M P D U T r a n s f e r P r o c e d u r e u s i n g R T S / C T S Figure 2.6 shows the directed M P D U transfer procedure with the use of R T S / C T S . A station shall use an RTS/CTS exchange for directed frames only and the length of the M P D U wi l l Chapter 2 Background 16 be used to determine whether R T S / C T S is used. If the length is greater than a threshold, RTS_Threshold = 3000 bits [7], the R T S / C T S exchange is used. From Figure 2.6, the source station sends a RTS frame using the basic access procedure and the backoff procedure. Within the CTS_Timeout, the destination station returns the C T S frame after the SIFS. A l l other stations wi l l set their N A V to the duration of the whole transmission process. When the source station receives the C T S frame, it wi l l transmit the data frame after SIFS. Finally, the destination station wi l l send back an A C K frame upon successful reception of data after the SIFS. DIFS SoUIX£_ RTS SIFS Destination Other Stations CTS DATA SIFS SIFS ACK DIFS NAV (RTS) Contention Window NAV (CTS) Defer Access Backoff after Defer Figure 2.6 Directed MPDU Transfer using RTS/CTS. Chapter 2 Background 17 2.2 .6 R T S / C T S R e c o v e r y P r o c e d u r e a n d R e t r a n s m i t L i m i t s Due to a collision or interference, CTS may not be returned after the RTS transmission. If after a RTS is transmitted, the CTS fails to return within a CTS_Timeout. The Short Retry Count is incremented and then a new RTS will be generated following the backoff procedure. This process will continue until the transmission is successful, or until the Short Retry Count reaches a ShortRetryLimit, and the MSDU or MPDU is discarded. The ShortRetryLimit is 7. If a directed data frame has been transmitted without RTS/CTS, but an A C K frame has not been received within an ACK_Timeout, it will go through the backoff procedure. The data frame will be retransmitted when the backoff time has reached zero. This process will continue until the Long Retry Count reaches a LongRetryLimit, and then the M S D U or M P D U is discarded. The LongRetryLimit is 4. 2.2 .7 B r o a d c a s t a n d M u l t i c a s t M P D U T r a n s f e r P r o c e d u r e In the DCF, when broadcast or multicast MPDUs are transferred from a station to stations, only the basic access procedure is used. Regardless of the length of the frame, and no A C K is returned by any destination stations. However, any broadcast of multicast MPDUs transferred from a station to the DS should follow the basic access method and the rules for RTS/CTS exchange. The broadcast/multicast message shall be distributed into the BSS. 2.2 .8 A C K P r o c e d u r e An AP shall always generate an A C K frame, and a station shall return an A C K frame when it receives a unitcast data frame or management frame. If the source station has not received the A C K frame within the ACK_Timeout, it will conclude that the MPDU was lost. Chapter 2 Background 18 2.2 .9 D u p l i c a t e D e t e c t i o n a n d R e c o v e r y Each frame has an M P D U ID which is a 16 bit hash of the 2 Octet Network ID field, 6 Octet source address and a 1 Octet sequence number maintained by the source station. A destina-tion station checks the M P D U ID with the IDs kept in the M P D U _ I D _ C A C H E and discards the duplicate frame, but it wi l l still follow the A C K procedure. 2.3 I E E E 802 .11 P h y s i c a l L a y e r The draft standard includes two types of P H Y layer specifications for the 2.4 G H z I S M band, F H S S and DSSS, and Infrared (IR) P H Y specification. The F H S S P H Y parameters are used in the simulation model in this thesis and the capture model is mainly based on the frequency modulation. 1 Mbit/s and 2 Mbit/s are the only rates currently supported, and the 1 Mbit /s data rate is used in the simulation model. The number of transmit and receive frequency channels is 79 for the North America. Detail information on P H Y layer can be found in [7]. Chapter 3 Distributed WLAN Architecture and the Func-tional Evaluation of the IEEE 802.11 MAC Pro-tocol This chapter describes the architecture of the distributed W L A N employing multiple R B s and how the frames received by the RBs are forwarded over the L A N . B y reviewing the architec-ture of the distributed W L A N and the I E E E 802.11 M A C methods, this chapter provides an evaluation of the functional aspects of the distributed W L A N using the standard M A C protocol. Moreover, a simulation model of the proposed network using O P N E T w i l l be presented. Then, the design of the I E E E 802.11 M A C state machines wi l l be described. 3.1 D i s t r i b u t e d W L A N A r c h i t e c t u r e Figure 3.1 shows the distributed architecture. This architecture is based on using the same radio channel over the entire coverage area. Each W T is interconnected with multiple R B s to a backbone L A N . This architecture avoids the need for frequency coordination and does not need the complicated handoff mechanisms as required in the cellular architectures. In Figure 3.1, this W L A N consists of multiple WTs transmitting (uplink) to the R B s and receiving (downlink) from the R B s . Each W T consists of a work-station equipped with a wireless transceiver, and each R B incorporates a processing unit which buffers and forwards incoming frames according to a bridging algorithm [1]. The W T employs a M A C protocol to communicate with the R B s within its range over the wireless channel. The locations of the mobile terminals are variable. Sufficient number of R B s are attached to the backbone L A N at fixed locations to provide multiple-site radio coverage of all possible locations of the WTs . The proposed architecture can be expanded easily to increase network capacity and reconfiguration of network topology. W T s can be added in any existing network coverage area but the number of WTs that can be accommodated within a certain 19 Chapter 3 Distributed WLAN Architecture and the Functional Evaluation of the IEEE 802.11 MAC Protocol 20 area is limited by the network throughput capacity available over that area. The throughput may be increased by reducing the coverage of each RB in that area, and increasing the number of RBs. The coverage area of the network can be expanded by attaching more RBs to the backbone L A N to cover new areas. BACKBONE L A N W T W T W T WT = Wireless Terminal RB = Radio Bridge Figure 3.1 The distributed architecture for WLAN, with multiple radio bridges. 3.2 F u n c t i o n a l E v a l u a t i o n o f t h e D i s t r i b u t e d W L A N E m p l o y i n g t h e I E E E 8 0 2 . 1 1 M A C P r o t o c o l A L A N consists of a collections of devices that share the network's transmission capacity. The function of the M A C protocol is to control access to the transmission medium to provide an orderly and efficient use of that capacity. The IEEE 802.11 M A C protocol is designed to reduce Chapter 3 Distributed WLAN Architecture and the Functional Evaluation of the IEEE 802.11 MAC Protocol 21 the collision probability between multiple stations accessing a wireless medium by using the carrier sensing mechanism and the random backoff time mechanism as described in Chapter 2. To illustrate the operation of the distributed network employing the IEEE 802.11 M A C protocol, the following example is given. 3.2.1 A n E x a m p l e o f t h e D i s t r i b u t e d W L A N In Figure 3.2, there are two RBs and three WTs. Each WT can start transmission when it senses the channel free for DIFS time. Besides, each WT can choose either to send the actual data with RTS/CTS or without RTS/CTS. For simplicity reason, all the RBs and WTs in Figure 3.2 are within range. Therefore, there are no hidden terminal in this example. Figure 3.2 An example of the WLAN employing RBs. Consider that W T A has a data frame waiting in its transmit queue ready to be sent. W T A first needs to find out whether the payload length exceeds the RTS_Threshold, and if so RTS will be sent before the actual data frame. The RTS_Threshold can be set from 0 (which means no Chapter 3 Distributed WLAN Architecture and the Functional Evaluation of the IEEE 802.11 MAC Protocol 22 MPDU is delivered without the use of RTS/CTS) to maximum M P D U length, and it can be set differently in each station. In this case, suppose all the MPDU needs to be sent with RTS/CTS. An A C K frame shall be returned by the destination station upon successful reception of a data frame within the ACK_timeout. W T A first listens to the channel and if it is free for longer than DIFS, then transmits a RTS. Once the transmission has started, all other stations, if they sense the channel busy will defer until the medium is free for DIFS and go into backoff. All the RBs and WTs are within range and receive the RTS at different times because of the propagation delay. R B t and R B 2 b ° m receive the RTS at different times and try to return a CTS, and each RB needs to wait for SIFS time and senses the channel before sending the CTS. If RBj receives the RTS first, then RB] will return a CTS first. In that case, if R B 2 senses the channel while RB, has already started transmitting the CTS, R B 2 will delay it transmission and will go back to the wait state. This situation can be vice versa. However, RBj and R B 2 sense the channel at the same time and find out that it is free, and then they both return CTS back to W T A . As a result, two CTSs collide and both are lost. And after the CTS_timeout has expired, W T A goes into backoff and waits for the next available time for retransmission of the RTS frame. In other words, if the CTS frame is successfully received by W T A , then the data frame will be sent after SIFS time. The same approach will be followed when RBs receive the data frame and return A C K frame after SIFS. Figure 3.3 shows the timing relationship for the system. Chapter 3 Distributed WLAN Architecture and the Functional Evaluation of the IEEE 802.11 MAC Protocol 23 W T A W T B W T C RBj R B 2 RTS Figure 3.3 The Time-line diagram of the example. 3.2 .2 T h e A d v a n t a g e s a n d D i s a d v a n t a g e s o f t h e p r o p o s e d W L A N u s i n g I E E E 8 0 2 . 1 1 M A C p r o t o c o l The distributed W L A N uses the same radio channel over the entire coverage area. This can eliminate the need for frequency coordination, and can avoid the complicated handoff mechanism. Each WT does not depend only on one particular RB. If a RB fails to work properly, the whole network can still operate without that RB, since each WT needs not access a fixed AP (RB). Besides, when fading occurs, the packets may not reach the destination, if there is only one RB that the W T can communicate with. Multiple RBs can enhance the chances of successful packet reception. However, the disadvantage is that the packet waiting time of a W T is longer because more WTs trying to access the same channel. In other words, a station has a packet to transmit must wait until the channel is free, and therefore, the access delay is high. Moreover, Chapter 3 Distributed WLAN Architecture and the Functional Evaluation of the IEEE 802.11 MAC Protocol 24 when the multiple RBs receive a data frame, they will either all return ACKs frames or one of them returns A C K frame, and that will delay the next data packet transmission. From the example shown in Figure 3.2, a relatively high system throughput can be achieved since multiple receptions can guarantee at least one packet to be successful received by the destination station. When the RTS/CTS exchange is successful, the channel is reserved for both the source station and the destination station, and the data packet has a very high probability to be received correctly. 3.3 D e s i g n o f t h e S i m u l a t i o n M o d e l Simulation plays an important role in the performance evaluation of M A C protocols. In the simulation environment, different models are used to represent real-life situations. However, certain assumptions need to be made to reduce the amount of computations. In this section, a simulation model of the proposed network using OPNET will be presented. Then, the design of the IEEE 802.11 M A C state machines will be described. 3.3.1 T h e O P N E T S i m u l a t i o n E n v i r o n m e n t OPNET is a comprehensive engineering tool supporting the modeling of communication networks and distributed systems. The OPNET environment can be used for analyzing the behavior and performance of the modeled systems by a simulation study, including model design, simulation, data collection, and data analysis. There are four model-specification editors, called network, node, process, and parameter. The network editor depends on elements specified in the node editor, and the process editor defines the models in the nodes. The parameter editor is used to define parameter models, such as packet format, antenna pattern and so on. Chapter 3 Distributed WLAN Architecture and the Functional Evaluation of the IEEE 802.11 MAC Protocol 25 3.3 .2 T h e N o d e M o d e l Figure 3.4 shows the node model for the station (WT) in the simulator. In the node model, there are a packet generator, a sink, a M A G queueing processor, a radio transmitter and a radio receiver. The solid lines interconnect each block representing the packet streams and the dash lines represent the statistic information streams. Individual data frames are generated by the packet generator according to the Poisson random process with exponential interarrival times. Each station has its own packet generator which is statistically independent of others. When the packet is generated, it is passed to the M A C processor and then follows the M A C protocol to be sent out via the transmitter attached to the M A C processor. The receiver uses carrier sensing to determine the state of channel. If the received signal power is above the carrier sensing threshold (CST), the receiver senses the channel busy. Otherwise, the channel is considered free. Moreover, the received signal power of a packet has to be above the receiver sensitivity (RS) in order to be received by the receiver. The sink receives valid data frame from the M A C and then with the statistics from the data frame calculates the throughput and delay. o sjLnk a greon. E  Sir i II i II — r i x _ 0 m.aC rfcx_0 Figure 3.4 The node model for the WT. Chapter 3 Distributed WLAN Architecture and the Functional Evaluation of the IEEE 802.11 MAC Protocol 26 3.4 T h e C h a n n e l M o d e l s This section describes the channel models used in the simulation. Indoor radio propaga-tion is dominated by the same mechanisms as outdoor: reflection, diffraction, and scattering [28]. In general, indoor channels may be classified either as line-of-sight (LOS) or Obstructed (OBS), with varying degrees of clutter. Due to multiple reflections from various objects, the electromag-netic waves travel along different paths of varying lengths, and the strengths of the waves decrease as the distance between the transmitter and receiver increases. Propagation models predict the average received signal strength at a given distance from the transmitter. 3.4.1 P a t h L o s s M o d e l The average received signal power decreases with some power of distance. The average large-scale path loss for an arbitrary transmitter-receiver separation is expressed as a function of distance by using a path loss exponent, y. PL(dB) = PL(d0) + l O y l o g ^ j (3.1) PL(dQ) = -lOlog X2 .(4rc)24j (3.2) where y indicates the rate increases with distance, dQ is the reference distance determined from measurements close to the transmitter, d is the distance between the transmitter and the receiver, and X is the wavelength in meters. The reference distance is chosen as 1 meter and the path loss exponent, y is 3 [28]. The received signal power is: Chapter 3 Distributed WLAN Architecture and the Functional Evaluation of the IEEE 802.11 MAC Protocol 27 Pr(dBm) = Pt(dBm) + Gt(dB) + Gr(dB) - PL(dB) (3.3) where Pr is the received power, which is a function of the transmitter and receiver separation, Pt is the transmitted power, Gt is the transmitter antenna gain, and Gr is the receiver antenna gain. 3.4.2 R a y l e i g h F a d i n g In addition to the path loss model, in a mobile radio environment, the short-term fluctua-tions caused by multipath propagation is called small-scale fading. Small-scale fading is caused by wave interference when two or more multipath components arrive at the receiver over a short period of time. It is generally classified as either flat or frequency selective. If the mobile radio channel has a constant gain and a linear phase response over a bandwidth that is greater than that of the transmitted signal, then the received signal will undergo flat fading. The most common amplitude distribution of the instantaneous gain of flat fading channels is the Rayleigh distribu-tion. The Rayleigh flat fading channel model assumes that the channel induces an amplitude which varies in time according to the Rayleigh distribution. The Rayleigh distribution has a probability density function (pdf) given by: p(r) = -2cxp a 0 ( 2 \^ r V 2a j ( 0 < r < ° ° ) (r<0) (3.4) 2 where 2a is the average power of the received signal before envelope detection. The probability that the envelope of the received signal does not exceed a specified value R is given by the corresponding cumulative distribution function (CDF): Chapter 3 Distributed WLAN Architecture and the Functional Evaluation of the IEEE 802.11 MAC Protocol 28 P(R) = Pr(r<R) = jp(r)dr = 1 - exp v 2a2 j (3.5) In simulation, an exponential distributions generator is used to generate a series of numeric values with mean received power for each packet received. In Figure 3.5, the theoretical Rayleigh CDF is used to compare with the simulation results. The mean received power is 3.956 nW. The simula-tion results closely match the theoretical line. ' A X X A X > ! ! < X ) O O O O O O O O O O 0 | ; Simulation X Theoretical 10 15 20 Received Signal Power (nW) 25 30 Figure 3.5 Rayleigh CDF vs. Simulation results. 3.4.3 L o g - n o r m a l S h a d o w i n g Log-normal shadowing is the random shadowing effects which occur over a large number of measurement locations which have the same transmitter-receiver separation, but have different levels of clutter on the propagation path. The shadowing effect will be the same throughout each Chapter 3 Distributed WLAN Architecture and the Functional Evaluation of the IEEE 802.11 MAC Protocol 29 simulation, since all the stations (RBs and WTs) are assumed fixed. Besides, the shadowing effect between each transmitter and receiver pair is also fixed. With the effect of log-normal shadowing, the received power Pr L(dBm) will be: Pr L(dBm) - Pr(dBm) + La(dB) (3.6) where L c is a zero-mean Gaussian distributed random variable (in dB) with standard deviation o" dB. La is calculated for each transmitter and receiver pair before the simulation takes place. For indoor environment [28], a value of 6 dB is used for a. 3.5 S t a t i o n L o c a t i o n All the WTs and RBs are placed along a straight line. This simple one dimensional model can be used to represented a long corridor[26]. The WTs are placed 10.35 meters apart for 15 WTs and 5 meters for 30 WTs in the network. The RBs are evenly distributed throughout the network. Figure 3.6 shows the different arrangements for different numbers of RBs. Chapter 3 Distributed WLAN Architecture and the Functional Evaluation of the IEEE802.il MAC Protocol R B 1 dfe 2 A 1 2 3 4 5 6 W T 7 8 9 10 11 12 13 14 15 R B 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 W T R B 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 W T Figure 3.6 Four sets of network arrangement Chapter 3 Distributed WLAN Architecture and the Functional Evaluation of the IEEE 802.11 MAC Protocol 31 3 . 6 D e s i g n o f t h e I E E E 8 0 2 . 1 1 M A C P r o c e s s The process model of the M A C is the core of the simulator. The way to implement the 802.11 M A C protocol is to use a finite state machine which defines the transition conditions for changing states. In order to avoid complicated implementation, some assumptions are made. The RBs and WTs are treated as individual stations, so that each WT can generate packets. Besides, only one channel is used for both uplink and downlink traffic. Therefore, a station cannot be transmitting and receiving at the same time. This type of transmission is referred as half-duplex. There are two different methods that the RBs will use to return ACKs. The first method is that only one of the RBs will return A C K upon successful packet reception. With this method, the A C K might be lost due to collision, fading or shadowing. Therefore, the second method is introduced. This method requires all the RBs which successfully receive the packets to return ACKs. The first RB will return A C K after the SIFS, and the second RB will return A C K after the SIFS plus the transmis-sion time of the ACK, which is 28|J,5 + 1 \2\is. This is because the RB does not sense the channel before sending the ACK. If two RBs return ACKs after SIFS, both ACKS will collide. The third RB will wait two times the transmission time of the A C K plus the SIFS and the fourth RB will wait three times the transmission time of the A C K plus the SIFS before sending the A C K . This can make sure that all the ACKs will not collide with each other. This method provides multiple ACKs. Thus, even though one of the ACKs is lost because of fading or shadowing, the additional A C K can still be received in order to prevent retransmission of the packet. When more than one frame is received at a station, all frames collide and are assumed to be lost without the capture effect. Since all the WTs might receive packets from more than one RBs at the same time, the Chapter 3 Distributed WLAN Architecture and the Functional Evaluation of the IEEE 802.11 MAC Protocol 32 capture method described in 5.2 can improve the throughput of the system. 3.6.1 T h e F i n i t e S t a t e M a c h i n e o f t h e 802 .11 M A C P r o t o c o l There are nine states in the state machine and there is a function block written in C programming language providing additional functions for the entire process. The states and their transitions are described in Table 3.1 and Figure 3.7. Figure 3.7 The state machine of the 802.11 MAC protocol. Chapter 3 Distributed WLAN Architecture and the Functional Evaluation of the IEEE 802.11 MAC Protocol 33 In Figure 3.7, the grey circles indicate unforced states and all these states can receive packet from the upper layer (the generator) or from the lower layer (the radio receiver). The receiving functions are written in the function block. When a packet is received from the upper layer, it will be inserted into the transmit queue. When a packet is received from the lower layer, it will be tested and corresponding action will be initiated. In Table 3.1, all the conditions for transi-tions are described for all the states. Table 3.1 Summary of all states and transitions of the 802.11 M A C protocol. State Next State Transitions Idle Frm_ready When a station is idle, no frame has been received from the generator or the radio receiver. If it receives a frame from the generator, it will go to the Frm_ready state. SIFS When a station is idle, a valid frame has been received and a replying frame has been prepared, it will go to the SIFS, and then wait for the SIFS times. Frm_ready DIFS This transition occurs, when the station senses the channel and it is free. Backoff This transition occurs when the station senses the channel and it is busy. DIFS Backoff This transition occurs when the channel is free for less than DIFS time. Transmit This transition occurs when the channel is free for DIFS time. A RTS or data frame is being sent to the radio transmitter. SIFS When a valid frame has been received and a reply-ing frame has been prepared. It will go to the SIFS, and then wait for the SIFS times. Chapter 3 Distributed WLAN Architecture and the Functional Evaluation of the IEEE 802.11 MAC Protocol State Next State Transitions Backoff Transmit This transition occurs when the backoff timer has expired. A RTS or data frame is being sent to the radio transmitter. Idle This transition occurs when the maximum number of retransmission is reached. The frame will be dis-carded. SIFS This occurs, when a valid frame has been received and a replying frame has been prepared. Transmit Tx_end This transition occurs when the packet has been sent to the radio transmitter. Tx_end ACK_wait This transition occurs after a station has transmit-ted a data frame and is waiting for an ACK frame. CTS_wait This transition occurs after a station has transmit-ted a RTS frame and is waiting for a CTS frame. Idle This transition occurs after a station has transmit-ted a CTS or an ACK frame. ACK wait Backoff This transition occurs when an ACK frame is not received in time, or an ACK frame has been received within the time required but a frame is waiting in the queue. Idle This transition occurs after a station has received an ACK frame within the time required, and no other frame is waiting or in the queue. CTS_wait SIFS This transition occurs after a station has received a CTS frame. Then, the data frame will be sent after the SIFS time interval. Backoff This transition occurs when a CTS frame is not received in time. SIFS Transmit This transition occurs when a station has waited for SIFS time interval. According to the specification, carrier sensing is done only when a CTS frame will be sent. Idle This transition occurs when a station senses the channel before sending a CTS frame and the chan-nel is busy, and then the CTS frame will be dis-carded. Chapter 3 Distributed WLAN Architecture and the Functional Evaluation of the IEEE 802.11 MAC Protocol 35 3.6 .2 T h e S i m u l a t i o n P a r a m e t e r s The simulation parameters can be set during the model specification. This is done via the OPNET Editors. The other method is to promote the attributes to the top level of the system model and then they will be specified at the start of the system's simulation. The first method is appropriate for those parameters that are constant during the simulation run, and the second method is used when the values need to be changed for each simulation run. Table 3.2 lists the simulation parameters. Chapter 3 Distributed WLAN Architecture and the Functional Evaluation of the IEEE 802.11 MAC Protocol 36 Table 3.2 The simulation parameters. Parameter Description Default ShortRetryLimit Maximum number of transmission attempts of a frame, the length of which is less than or equal to a RTSThreshold. 7 LongRetry Limit Maximum number of transmission attempts of a frame, the length of which is greater than a RTSThreshold. 4 RTSThreshold The number of Octets in an MPDU. An RTS/CTS hand-shake shall be performed for all frames where the length of the MPSU is greater than this threshold. 3000 bits (RTS/CTS not used in simulations unless specified.) DIFS Distributed Coordination Function Inter-frame Space. 128 ( I S SIFS Short In-frame Space. 28 ( I S Slot_Time Slot Time as defined in Equation (2.2). 50 | I S ACK_Timeout The maximum waiting time for the ACK frame. 300 \ls CW_min Minimum Contention Window. 15 CW_max Maximum Contention Window. 1023 RTS_frm RTS frame length. 160 bits CTS_frm CTS frame length. 112 bits ACK_frm ACK frame length. 112 bits DATA_frm (AO DATA frame length. 8000 bits QueueSize Size of the FIFO Queue. 10 frames Bit_rate Data rate. 1 Mbit/s Frequency Carrier Frequency. 2.4 GHz Pt(Tx) Transmitter Power. 100 mW RS Receiver Sensitivity. -85 dBm CST Carrier Sensing Threshold. -85 dBm c Capture Ratio. 10 dB T Path Loss Exponent. 3 o The Standard deviation of the log-normal shadowing. 6dB G, Transmitter antenna gain. OdB Gr Receiver antenna gain. OdB Chapter 3 Distributed WLAN Architecture and the Functional Evaluation of the IEEE 802.11 MAC Protocol 37 3.6 .3 D a t a C o l l e c t i o n The objective of most modeling efforts is to obtain measures of a system's performance or to make observations concerning a system's behavior. Both scalar and vector statistics can be computed and recorded automatically for a set of predefined statistics. Predefined statistics in OPNET are related to values that can be measured at specific objects within the model. Custom statistics can be declared by process models and OPNET provides support for both local statistics, which are maintained separately for each processor or queue, and global statistics, which are shared and contributed to by many entities in the model, such as the performance of the system. The following Table 3.3 summarizes the.statistics generated by the M A C process model. Moreover, OPNET provides a large number of built-in local statistics, such as transmission delay, propagation delay, received power, signal to noise ratio and bit error rate. Chapter 3 Distributed WLAN Architecture and the Functional Evaluation of the IEEE 802.11 MAC Protocol 38 Table 3.3 The Statistics collected by the process. Type Name Actual Load (bits) Actual Load (bits/sec) Actual Load (data packets) Actual Load (data packets/sec) Number of Collisions Local Access Delay (sec) Throughput (bits) Throughput (bits/sec) Throughput (data packets) Throughput (data packets/sec) Queueing Delay (sec) Actual Load (bits) Actual Load (bits/sec) Actual Load (data packets) Actual Load (data packets/sec) Number of Collisions Global Access Delay (sec) Throughput (bits) Throughput (bits/sec) Throughput (data packets) Throughput (data packets/sec) Queueing Delay (sec) Chapter 4 Simulation Results and Discussion - Part 1 In this chapter, the performance of the proposed distributed W L A N s using I E E E 802.11 M A C protocol in different channel environments w i l l be studied by computer simulation. N o capture is assumed in this chapter. First, the parameters and the assumptions used in the simula-tion are described and then a model verification is provided. The simulation results obtained in additive white Gaussian noise ( A W G N ) , Rayleigh fading and log-normal shadowing channels w i l l be presented and discussed. The effects of various parameters, such as number of W T s , number of R B s , «, transmitted power levels, Tx, and standard deviation, a , of the log-normal shadowing wi l l be examined. For all the simulation results, a 95% confidence interval within ± 5 % of the average values is shown and the worst case is within ±10 %. 4.1 S i m u l a t i o n D e s c r i p t i o n s 4 .1 .1 P e r f o r m a n c e M e a s u r e m e n t The performance is measured in terms of normalized throughput, r | , and the access delay, T , versus the normalized offered load, p . The p is defined as the average number of bits per second generated at the source stations (WTs) and normalized by data rate. The r| is defined as the average number of bits per second successfully received by the destination (RBs), and normal-ized by the data rate. The x is calculated when the (potential) packet is dequeued at the W T until the first bit of the packet is successfully received at the RBs . 4.1.2 N e t w o r k A r r a n g e m e n t A linear (one-dimensional) network arrangement with a length of 145 meters is assumed 39 Chapter 4 Simulation Results and Discussion - Part 1 40 in the simulation. There are two different network sizes, with 15 W T s and 30 W T s . The W T s are evenly placed in the system. The number of R B , n, used in the network is from one to 4. Two different methods (A and B) wi l l be used for the R B s to return the A C K . Method A is that only one of the R B s which successfully receives the data packets w i l l return the A C K (in the figure legend, it is noted as n RBs) . Method B is that all the R B s that successfully receive data packets return A C K frames (in the figure legend, it is noted as n R B s (all)). For the simulation in this and Chapter 5, the default Tx for the W T s / R B s is set to be 100 m W and the default value of the o for log-normal shadowing is 6 dB. 4 .1 .3 E r r o r - F r e e C h a n n e l s The bit-error rate ( B E R ) is obtained as a function of signal-to-noise power ratio. The signal power is computed based on path loss equations (3.1), (3.2) and (3.3) in Chapter 3. The receiver noise floor is obtained based on the ambient noise temperature at the receiver antenna and noise figure, F. The effective noise temperature, Te is calculated as Te = (F-\)T0 (4.1) where TQ is ambient room temperature (290 K to 300 K ) . In the simulation, the values of F and Te used are 1.07 (0.3 dB) and 20° C. The receiver noise floor Pn, is given by Pn = k(Te + Tb)BW (4.2) -23 where k is Bol tzmann ' s constant given by 1.38 x 10 Jou les /Ke lv in , Tb is the effective background temperature, 290 K , and BW is the equivalent bandwidth of the receiver, 1000 kHz . Chapter 4 Simulation Results and Discussion - Part 1 41 Thus, the receiver noise floor is -113.7 dBm. Since the RS is -85 dBm, i.e., the minimum signal-to-noise ratio (SNR) is 28.7 dB, there is no packets error at the RBs or WTs. 4 .2 S i m u l a t i o n M o d e l V a l i d a t i o n Figure 4.1 Normalized throughput versus normalized offered load for 5, 10 and 20 stations with CWmin = 32 and CWmax = 256 for adhoc and infrastructure networks. The throughput performance and access delay obtain from the ideal channel (no fading or no transmission error) model are used to compared with those in [13]. The results from [13] are obtained from an adhoc network, in which all the stations can communicate with each other such that any station can be a destination. In the (infrastructure) model, all the stations send packets to an AP as destination. Figure 4.1 shows the throughput for 5, 10 and 20 stations with CWmin = 32 and CWmax = 256. It can be seen that the throughput of the infrastructure model is slightly lower than that in [13] for p > 0.7. In the infrastructure model, there is an extra AP which is not in Chapter 4 Simulation Results and Discussion - Part 1 42 adhoc model. Since the AP sends the A C K to the stations, the amount of contention is increased and the performance is degraded slightly. Figure 4.2 shows the access delay for different values of CWmin and CWmax with 5 stations. It can be seen that the results are close to those in [13]. 40 30 20 C W = 512,4096 (sim) C W = 128,1024 (sim) C W = 32,256 (sim) C W = 512,4096 (ref) C W = 128,1024 (ref) C W = 32,256 (ref) O X 10 . . - X - — - ' 0.2 0.3 0.4 Normalized Offered Load 0.5 0.6 Figure 4.2 Access delay versus normalized offered load with different values of CW(min/max) for 5 stations for adhoc and infrastructure networks. 4.3 AWGN Channel The Tx of the WTs/RBs is set to be 100 mW so that the received power of the packets at the WTs/RBs is above the CST/RS. Thus, all the packets can be received1 at the RBs or sensed at the WTs. Since the channel is assumed to be error-free, all the received packets at the RBs contain no error. Therefore, the packets are discarded only when multiple packets collide at the RB(s). Since the received power of the packets at the WTs is always above the CST, the WTs are able to A packet is assumed to be received at the RB if the received power of that packet is above the RS. Chapter 4 Simulation Results and Discussion - Part J 43 sense the channel status correctly in order to avoid the occurrence of hidden terminals in the network. Thus, the probability of more than one WTs sending packets within a short period of time is very low due to the use of carrier sensing. 4.3.1 Simulation Results In this section, the simulation results in AWGN channel are presented. The normalized throughput for 15 WTs using Method A and B are shown in Figure 4.3(a) and (b) respectively. In Figure 4.3(a), the saturated throughput, S, is quite high, about 0.9, because the probability of collision is very low. For p < 0.9 the throughput increases linearly with p since there is no packets lost due to collision. For p > 0.9, the probability of more than one WTs sending the packets increases and thus r| is kept at about 0.9. The reason why r\ is below one is because some channel time is used for returning ACKs, for DIFS and for SIFS. It can be seen that r\ is indepen-i dent of n. Because all the packets sent from the WTs can be received by one RB. Hence, adding more RBs does not increase the chance of receiving more packets. In Figure 4.3(b), r| decreases with n. In Method B, the potential packets at the WTs cannot be sent until all the ACKs returned from the RBs have been transmitted, because the channel is sensed as busy by other WTs when the RBs are transmitting the ACKs, thus, the multiple A C K frames increase the waiting time for the WTs to send their potential packets and the number of packets transmitted from the WTs is decreased. Similar observation can be obtained in Figure 4.4 for 30 WTs. The r\ for 15 and 30 WTs are about the same. Chapter 4 Simulation Results and Discussion - Part 1 Figure 4.3 Normalized throughput versus normalized offered load for 15 WTs in AWGN channel using (a) Method A and (b) Method B. Figure 4.4 Normalized throughput versus normalized offered load for 30 WTs in AWGN channel using (a) Method A and (b) Method B. Chapter 4 Simulation Results and Discussion - Part 1 45 The access delay for 15 WTs using Method A and B are shown in Figure 4.5(a) and (b) respectively. In Figure 4.5(a), the access delay is the same for any n. Because only one of the RBs returns A C K , it does not increase the waiting time of sending the potential packets at the WT. Since the A C K is always received correctly at the desired WT, the packets are not required to retransmit. In Figure 4.5(b), the access delay increases with n due to the additional A C K returned to increase the waiting time. The access delay for 30 WTs using Method A and B are also shown in Figure 4.6(a) and (b) respectively. Although r| for 15 and 30 WTs are about the same, the access delay for 30 WTs is much longer than that for 15 WTs. This is because more WTs are sending packets in the network and thus the waiting time for sending a packet becomes longer. Figure 4.5 Access delay versus normalized offered load for 15 WTs in AWGN channel using (a) Method A and (b) Method B. Chapter 4 Simulation Results and Discussion - Part 1 46 Figure 4.6 Access delay versus normalized offered load for 30 WTs in AWGN channel using (a) Method A and (b) Method B. 4.4 R a y l e i g h F a d i n g C h a n n e l In addition to the receiver noise floor and the path loss in AWGN channel, we assume a very slow Rayleigh fading channel in this section. Thus, the signal strength is assumed to be constant over a packet. If the packet sent from W T A arrives at W T B and its received power is below the CST due to fading effect, W T B cannot correctly sense the channel status as a busy channel. Then, W T A is hidden from W T B . So, a collision might occur if W T B sends a packet. In Rayleigh fading channel, a packet will not be received when its received power at the RB is below RS, and it will be discarded if it collides with other packets at the RB. 4.4.1 S i m u l a t i o n R e s u l t s The throughput curves for 15 WTs using Method A and B in a Rayleigh fading channel are shown in Figure 4.7(a) and (b), respectively. Compared to results for the AWGN channel in Chapter 4 Simulation Results and Discussion - Part 1 47 Figure 4.3 where no hidden terminal is assumed, the saturated throughput, S, in a Rayleigh fading channel is much lower. With no hidden terminal, the WTs can sense the channel status (idle or busy) accurately in order to avoid collision. However, in a Rayleigh fading channel, there is a possibility of some terminals hidden from the WTs so that the WTs cannot sense the channel busy when transmitting their packets. Thus, it increases the chance of collisions and degrades the performance. In a Rayleigh fading channel, a packet may not be received by the RB if the packet is faded2 and its received power is below RS. By using more RBs, the probability of the packet being faded, at all the RBs is decreased. Thus, it can increase the chance of a packet to be received by, at least, one of the RBs in order to increase the throughput. It can be seen that S increases with n. In Figure 4.7(a), S for 2 RBs is increased by 18% compared to that for one RB. For 4 RBs, S can be increased by 38%. In Figure 4.7(b), the performance is not increased by much as n increases. The S for 2 and 4 RBs is increased by 14% and 21% respectively compared to that for one RB. With 3 or 4 RBs, S with Method B is lower than that with Method A. In Method A, a single A C K is returned from one of the RBs even though 2 or more RBs may receive the packet. If the A C K is faded at the desired WT, the packet needs to be retransmitted. The fundamental idea of multiple ACKs in Method B is to increase the chance of A C K reception at the WTs in order to avoid the retransmission of the packets. However, the probability of an A C K being faded at the desired WT is not high. In other words, the desired WT can always receive one of the first two ACKs. Because if a packet sent from the WT can be received (not faded) at the RB, then the A C K sent from that RB can also be received (not faded) at the desired WT with a fairly high probabil-ity. Therefore, more ACKs sent from the RBs may not be necessary. Even though more RBs can increase the chance of packets reception, in Method B, the waiting time of the WTs to send their A packet is assumed to be faded at a WT (RB) if the received power of that packet at the WT (RB) is below the CST (RS). Chapter 4 Simulation Results and Discussion - Part 1 48 potential packets is increased due to multiple ACKs. Therefore, more RBs or multiple ACKs increases the waiting time of the potential packet to be sent and the throughput decreases with n in Method B. In Figure 4.8(a), the access delay for 15 WTs using Method A decreases with n. The x for 2, 3 and 4 RBs can be reduced by about 10%, 20% and 30% respectively compared to one RB. This is because the throughput increases with n as shown in Figure 4.7(a). However, x in Method B does not decrease very much as n increases as shown in Figure 4.8(b). The i for 4 RBs is only reduced by 10% compared to one RB. Because the WTs cannot send their potential packets until all the RBs have returned their ACKs which increase the waiting time (access delay) for the WTs to send packets. Figure 4.7 Normalized throughput versus normalized offered load for 15 WTs in Rayleigh fading channel using (a) Method A and (b) Method B. Chapter 4 Simulation Results and Discussion - Part 1 Figure 4.9 (a) Normalized throughput and (b) access delay versus normalized offered load for 30 WTs in Rayleigh fading channel using Method A. Chapter 4 Simulation Results and Discussion - Part 1 50 The normalized throughput and access delay for 30 WTs using Method A are depicted in Figure 4.9(a) and (b) respectively. It shows that for the same number of RB, r\ and x for 30 WTs are worse than those for 15 WTs. This is because there are more hidden terminals and hence collisions occur more frequently. With one RB, S for 30 WTs is about 50% less than that for 15 WTs. With 4 RBs, S for 30 WTs is about 35% less than that for 15 WTs. It can be seen that S for 2, 3 or 4 RBs can be increased by about 50, 85 and 110% respectively compared to one RB. The performance improvement for 30 WTs is greater than that for 15WTs. Therefore, adding more RBs is more efficient for a system with a large number of WTs. In Figure 4.9(b), x for 2, 3 and 4 RBs can be decreased by 14%, 28% and 38%, respectively. In the Rayleigh fading channel, for one RB, x with 30 WTs is about 2.5 times longer than that with 15 WTs. In contrast, over the AWGN channel, x with 30 WTs is about 60% longer than that with 15 WTs. Figure 4.10 Saturated throughput versus transmitted power in Rayleigh fading channel using Method A for (a) 15 WTs and (b) 30 WTs. Chapter 4 Simulation Results and Discussion - Part 1 51 The saturated throughput with 15 WTs and 30 WTs using Method A for transmitted power from 50 m W to 150mW are shown in Figure 4.10(a) and (b). It can be seen that S increases with Tx. For one R B at low Tx, the probability of packets being faded (not received) at the R B is high, especially for the packets sent from the WTs further away from the R B . Thus, the R B might not be able to cover all the WTs as shown in Figure 4.11. R B Figure 4.11 The coverage area of one RB for 15 WTs at low Tx. The WTs beyond the coverage area 3 means that the probability of packets sent from those W T s being received at the R B is quite low due to fading. In addition, the number of hidden terminals for the W T s is also very high because the probability of packets being faded (below C S T ) at the W T s is high for low Tx. It increases collisions and reduces the chance of packet reception at R B . Therefore, S is quite low at low Tx. As Tx increases, the coverage area increases and the number of hidden terminals is reduced. Thus, more packets can be received at the R B . In Figure 4.10, it can be seen that S increases with n because the chance of a packet being received by RBs increases with n. However, the rate of performance increases as a function of Tx decreases The coverage area means that there is a high probability of packet reception at the RB if the WTs are located within the coverage area of the RB. Chapter 4 Simulation Results and Discussion - Part 1 52 with n because increasing Tx does not increase the packet reception probability by much for high n, i.e., n = 4. This is because all the WTs can be covered by at least one RB as shown in Figure 4.12. Figure 4.12 The coverage area for the 4 RBs with 15 WTS at low Tx. The relative performance improvement for using more RBs with 30 WTs is greater than that with 15 WTs for all the evaluation range. However, the actual performance of 30 WTs is still worse than that of 15 WTs due to frequent collisions and longer waiting time for the potential packet transmission. In both figures, the performance improvement for using more RBs compared to one RB decreases as Tx increases. In Figure 4.10(a), S of one, 2, 3 and 4 RBs are very close together. This is because as Tx increases, more WTs can be covered by the RBs and the number of hidden terminals decreases. Eventually, for a very high Tx, the performance will be similar to that in AWGN channel in previous section in which the performance becomes the same for any n. Figure 4.13 shows the access delay with different Tx for 15 and 30 WTs. The x decreases with Tx and the decreasing rate decreases with n. The reasons are the same as the observation for S in Figure 4.10. The x for 30 WTs is longer than that for 15 WTs because of more collisions due to Chapter 4 Simulation Results and Discussion - Part 1 53 more hidden terminals and the waiting time of transmitting the potential packets at the WTs. 160 50 60 70 80 90 100 110 120 130 140 150 50 60 70 80 90 100 110 120 130 140 150 TX Power (mW) TX Power (mW) (a) (b) Figure 4.13 Access delay versus transmitted power in Rayleigh fading channel using Method A for (a) 15 WTs and (b) 30 WTs. The distribution of the saturated throughput on each WT for one RB with 15 WTs at Tx = 50 mW and 150 mW are shown in Figure 4.14(a) and (b) respectively. It can be seen that for Tx = 50 mW, S is mostly concentrated on the WTs near the RB which are WT 6, 7, 8, 9 and 10. Thus, one RB at Tx = 50 mW can cover few WTs as illustrated in Figure 4.11. For Tx = 150 mW, the distribution looks similar to Tx = 50 mW. The S is similarly concentrated on WT 5, 6 7, 8, 9, 10. Thus, the coverage area does not increase with increasing Tx= 150 mW. This is just because the Tx= 150 mW is not high enough to increase the coverage area significantly. But it can be seen that S on each W T for Tx = 150 mW is higher than that for Tx = 50 mW due to the increasing of the packet reception and reducing the number of hidden terminals. The distribution of the saturated throughput on each WT for 4 RBs with 15 WTs at Tx = 50 mW and 150 mW are shown in Figure Chapter 4 Simulation Results and Discussion - Part 1 54 4.14(c) and (d) respectively. For Tx = 50 mW, S is more or less evenly distributed in each WT because each RB can evenly cover few WTs in the network as shown in Figure 4.12. However, for Tx = 150 mW, S is more concentrated on the middle WTs. Because the middle WTs can be covered by more RBs as the coverage area of each RB increases. It can be expected that the distri-bution will become more even when increasing Tx. 0.025 r I 0.02 f D) I 0.0151-J> 0.01| w 0.005r l i l i 1 2 3 4 5 6 7 8 9 101112131415 Wireless Terminal (a) 0.03 r - 0.0251 D. j? 0.021 o i_ i~ 0.0151 T3 3 ra W 0.01 0.005 1 2 3 4 5 6 7 8 9 101112131415 Wireless Terminal (c) 0.07 3 a £° l o r-|0 . 5 2 0 0 0.05 3 0.04 a r : D) 2 0.03 H S 0.02 3 lo w 0.01 i l 1 2 3 4 5 6 7 8 9 101112131415 Wireless Terminal (b) 1 2 3 4 5 6 7 8 9 101112131415 Wireless Terminal (d) Figure 4.14 The saturated throughput distribution on each of the 15 WT for (a) one RB with Tx = 50 mW, (b) one RB with Tx = 150 mW, (c) 4 RBs with Tx = 50 mW and (d) 4 RBs with Tx = 150 mW Chapter 4 Simulation Results and Discussion - Part 1 55 4.5 Log-Normal Shadowing Channel In this section, the simulation results are presented for a log-normal shadowing channel which has a log-distance path loss in addition to receiver noise floor. The parameter considered is the standard deviation, a (in dB), of the log-normal shadowing. The values of a used in the simulation are from 4 dB to 14 dB. The received power of a packet will be calculated based on equation (3.6) in Chapter 3. 4.5.1 Simulation Results The saturated throughput versus a in a log-normal shadowing channel for 15 WTs using Method A and B are shown in Figure 4.15. Figure 4.15 Saturated throughput versus O" in log-normal shadowing channel for 15 WTs using Method A and Method B. Chapter 4 Simulation Results and Discussion - Part 1 56 Again, S of Method B is lower than that for Method A because of longer waiting time for the multiple ACKs. In Figure 4.15, it can be seen that for Method A, S decreases with a. Because for a large value of a, the coverage area of each RB is smaller and more terminals will be hidden from each other. Therefore, there will be more unsuccessful received packets or collided packets at the RBs. At low a, S of any n approaches about 0.7 whereas S in AWGN channel is about 0.9. This is because as a decreases, the effect of log-normal shadowing diminishes and hence, the performance will approach that in AWGN channel. The reasons are similar to the high Tx for Rayleigh fading channel in previous section where the probability of packet reception increases and the number of hidden terminal decreases. The performance improvement of using more RBs compared to one RB is increased with a. For a = 6 dB, the performance of using 2, 3 and 4 RBs is improved by 4%, 22% and 28% respectively compared to one RB. For a = 12 dB, the perfor-mance of using 2, 3 and 4 RBs can be improved by 55%, 110% and 130%. Since there are more hidden terminals for a high value of a, multiple RBs can improve the performance significantly. Figure 4.16 shows the access delay for 15 WTs using Method A. For a < 8 dB, x for one RB is better (less) than that for multiple RBs, but S of one RB is worse (lower) than that for multiple RBs for any values of a as in Figure 4.15. This is because one RB might not be able to receive the packets from a few (one or two) WTs due to shadowing. Using more RBs, it always increases the chance of receiving a packet sent from the WTs received by at least one of the RBs. Therefore, S of one RB is always lower than that of multiple RBs. But this is different from x because x of a packet is calculated if and only if the packet is successfully received at the RB. Since the total number of WTs being covered by one RB is lower than that by 4 RBs, there are Chapter 4 Simulation Results and Discussion - Part 1 57 fewer WTs which packets can be successful received by the RB. Therefore, there are fewer ACKs for each WT to wait to send its potential packet. As o increases, x for one RB becomes higher than that for multiple RBs. In this case, the probability of a packet not being received by one RB is increased due to more hidden terminals and faded packets. Thus, more retransmissions are required. For more RBs, the probability of a packet being received by any one of the RBs is higher and the number of retransmissions can be reduced. The saturated throughput and access delay for 30 WTs using Method A are shown in Figure 4.17(a) and (b). Both figures look similar to the results for 15 WTs, but the overall performance is lower. 120 4 RBs 3 RBs 4 6 8 10 12 14 Shadow (SD in dB) Figure 4.16 Access delay versus O " in log-normal shadowing channel for 15 WTs using Method A. Chapter 4 Simulation Results and Discussion - Part 1 58 Figure 4.17 (a) Saturated Throughput and (b) access delay versus CJ using Method A for 30 WTs. Chapter 5 Simulation Results and Discussion - Part 2 In this chapter, the performance of the proposed distributed WLANs using IEEE 802.11 M A C protocol will be studied with and without capture effect. A channel model with receiver noise floor, log-distance path loss, Rayleigh fading and log-normal shadowing will be used. Since the performance with Method A is always better than that with Method B as shown in previous chapter, only Method A will be considered in the simulation. The effects of changing the number, n, of RBs, transmitted power levels, the standard deviation of the log-normal shadowing on the performance will be investigated. In order to get insight into the study of distributed networks, a two-dimensional network model will be used as comparison for one-dimensional model. In addition, the length of packet size and the ACK_Timeout will also be studied. The Tx= 100 mW and a = 6 will be used as default values. 5.1 L o g - n o r m a l S h a d o w i n g P l u s R a y l e i g h F a d i n g C h a n n e l W i t h N o C a p t u r e In this section, the simulation results with no capture in a log-normal shadowing plus Rayleigh fading channel will be presented. Figure 5.1(a) and (b) show the normalized throughput for 15 and 30 WTs respectively. It can be seen that T) for 30 WTs is lower than that for 15 WTs because of more collisions and longer waiting time. Compared to the Rayleigh fading channel in Figure 4.7(a), r\ with one or 2 RBs is lower in this channel. This is because the probability of a packet being received successfully in log-normal shadowing plus Rayleigh fading channel is generally lower than that in a Rayleigh fading channel due to the additional degradation from log-normal shadowing. The rj with 3 or 4 RBs is about the same in both channels, because the probability of a packet being successfully received increases with n, alleviating the effect of the 59 Chapter 5 Simulation Results and Discussion - Part 2 60 log-normal shadowing. For 30 WTs, T| for n > 2 is about the same as that in a Rayleigh fading channel as shown in Figure 4.9(a). The S with 4 RBs for 15 WTs and 30 WTs can be increased by about 100% and 200% respectively compared to that for one RB. 0.2 0.4 0.6 0.8 Normalized Offered Load (a) 0.2 0.4 0.6 0.8 Normalized Offered Load (b) Figure 5.1 Normalized throughput versus normalized offered load in log-normal shadowing plus Rayleigh fading channel for (a) 15 WTs and (b) 30 WTs. The access delay for 15 and 30 WTs are shown in Figure 5.2(a) and (b) respectively. For 15 WTs, the access delay of one and 2 RBs is higher than that in a Rayleigh fading channel. For 30 WTs, the access delay for any n is about the same in both channels. Thus, the performance, r\ and x, for n>2 is about the same as in both the Rayleigh fading channel and log-normal shadow-ing plus Rayleigh fading channel, with a = 6 dB and Tx= 100 mW. However, compared to the results for the log-normal shadowing channel in Section 4.5, presented in Figure 4.16 and Figure 4.17, T (5) in log-normal shadowing plus Rayleigh fading channel is doubled (40% lower). Hence, the performance in a log-normal shadowing plus Rayleigh fading channel is much worse Chapter 5 Simulation Results and Discussion - Part 2 than that in a channel with log-normal shadowing only for any n with a = 6 dB and Tx = mW. This is because the effect of Rayleigh fading degrades the performance significantly. 61 100 o 100 0.2 0.4 0.6 0.8 Normalized Offered Load (a) 0.2 0.4 0.6 Normalized Offered Load (b) 0.8 Figure 5.2 Access delay versus normalized offered load in log-normal shadowing plus Rayleigh fading channel for (a) 15 WTs and (b) 30 WTs. The saturated throughput for 15 and 30 WTs with different transmitted power are shown in Figure 5.3(a) and (b), respectively. It can be seen that S increases with Tx but the rate of increase decreases as n increases. The explanation is the same as that for the Rayleigh fading channel for different Tx in Figure 4.10 in Section 4.4. The performance improvement, on average, of 4 RBs over one RB for 15 WTs and 30 WTs are 120% and 220%, respectively, as shown in Figure 5.3. Thus, it is more efficient to use more RBs in a network with a larger number of WTs. Compared to the Rayleigh fading channel in Figure 4.10, it can be seen that S with any n in both channels is about the same at low Tx = 50 mW. For Tx > 60 mW, S in a log-normal shadowing plus Rayleigh fading channel becomes lower and the difference increases as Tx increases. Chapter 5 Simulation Results and Discussion - Part 2 62 100 110 120 130 140 150 TX Power (mW) (a) 80 90 100 110 120 130 140 150 TX Power (mW) (b) Figure 5.3 Saturated throughput as a function of transmitted power with a = 6 dB in log-normal plus Rayleigh fading channel for (a) 15 WTs and (b) 30 WTs. For low Tx, the probability of a packet being non-faded at RB/WTs with Rayleigh fading is already low. Thus, the additional effect of log-normal shadowing of a = 6 dB does not increase the non-faded probability by much. Hence, S in both channels is about the same. As Tx increases, the non-faded probability with Rayleigh fading increases. Then, the additional effect of log-normal shadowing can significantly decrease the non-faded probability. Therefore, S is lower in a log-normal plus Rayleigh fading channel and thus the difference in two channels increases with Tx. For 3 or 4 RBs, the performance difference in the two channels is small. This is because the degradation of performance due to log-normal shadowing can be alleviated by adding more RBs. The performance improvement of 4 RBs to one RB for 15 WTs at low and high Tx in a log-normal shadowing plus Rayleigh fading channel is 200% and 60%, respectively, whereas the performance improvement at low Tx and high Tx in a Rayleigh fading channel is 150% and 20%, respectively. Therefore, the performance improvement degrades rapidly in a Rayleigh fading channel. The Chapter 5 Simulation Results and Discussion - Part 2 63 access delay for 15 and 30 WTs are shown in Figure 5.4(a) and (b), respectively. For one or 2 RBs, T for Tx = 50 mW in a log-normal shadowing plus Rayleigh fading channel is about the same as in a Rayleigh fading channel as shown in Figure 4.13. Similarly, x for Tx > 60 mW in a log-normal shadowing plus Rayleigh fading channel is higher than that in a Rayleigh fading channel and the difference increases with Tx. It can be seen that the performance improvement by using more RBs is greater in a log-normal shadowing plus Rayleigh fading channel than in a Rayleigh fading channel. 180 50 60 70 80 90 100 110 120 130 140 150 50 60 70 80 90 100 110 120 130 140 150 TX Power (mW) TX Power (mW) (a) (b) Figure 5.4 Access delay as a function of transmitted power with a = 6 dB for (a) 15 WTs and (b) 30 WTs. The effect of a on the performance of throughput and access delay for 30 WTs is presented in Figure 5.5(a) and (b) respectively. In Figure 5.5(a), it can be seen that S does not Chapter 5 Simulation Results and Discussion - Part 2 64 M 0.15 8 o 330 300 270 240 210 Shadow (SD in dB) (a) Shadow (SD in dB) (b) Figure 5.5 (a) Saturated throughput and (b) access delay as a function rj for 30 WTs. change with c by much whereas in Figure 4.17(a) in a log-normal shadowing channel, S decreases with a significantly. The performance improvement of 2, 3 and 4 RBs over one RB is about 67%, 160% and 225% respectively for any values of a . Compared to a log-normal shadow-ing channel, S for small values of a in this channel is much lower. The difference becomes smaller as a increases. For small values of a, the performance is degraded mainly due to Rayleigh fading, because the effect of log-normal shadowing is relatively small and does not degrade the performance very much at small a. This can be observed in Figure 4.17(a) where S is quite high, about 0.65, at small a = 4 dB. As a increases, the effect of Rayleigh fading diminishes relatively compared to log-normal shadowing. Thus, at c = 14 dB, the effect of Rayleigh fading becomes negligible and S in a log-normal shadowing plus Rayleigh fading channel is about the same as in a log-normal shadowing channel. The x in this channel is also Chapter 5 Simulation Results and Discussion - Part 2 65 worse (higher) than that in a log-normal shadowing channel as seen in Figure 4.17(b). As a increases, the difference gets smaller. For example, at a = 4, T for one RB in this channel can be 3 times more than that in a log-normal shadowing channel. For one RB, x decreases with a. However, the throughput does not increase with a. This is because when there is only one RB and 30 WTs, there will be more collisions and more WTs shadowed from the RB as a increases. Therefore, the number of WTs that can reach the RB is reduced, and that can reduce the waiting time of the packet. Since the access delay is only calculated for the packets which can be received correctly at the RB, the number of correctly received packets is reduced as c increases. For n > 2 , x does not vary much with o~. In conclusions, the performance in a log-normal shadowing plus Rayleigh fading channel of n > 2 is not affected by Tx or o" significantly. 5.2 L o g - n o r m a l S h a d o w i n g P l u s R a y l e i g h F a d i n g C h a n n e l W i t h C a p t u r e 5.2.1 C a p t u r e M o d e l For the simulation results of throughput and access delay obtained in the previous sections, no capture is assumed to be used. Hence, if two or more packets arrive at the receiver with any amount of overlap in time, all the packets are unusable and discarded due to collision. Since frequency modulation exhibits a so-called capture effect characteristic [28], the perfor-mance can be improved by considering that the receiver can successfully capture a packet with the largest received power amount several packets that arrive at the receiver at about the same time. It is assumed that the received signal powers are more or less constant over a packet duration. In general, two common conditions for a receiver to capture packet A , A e {1,2, ...,m} where m is the number of packets arriving at the receiver simultaneously, can Chapter 5 Simulation Results and Discussion - Part 2 be used [2][3], i.e., 66 max {T-} >c, i = 1, 2, m, A (5.1) and SIR = > c (5.2) m i = 1, i *• A where T j, j=\,2,..., ra, is the received power of packet j and c is the capture ratio. The value of c will depend on the particular modulation and coding technique used in the system. In (5.2), SIR is defined as the signal-to-total interference power ratio. In the simulation, the second capture model is employed because the first capture model is somewhat unrealistic and optimistic. A packet with the largest received power can be captured by the receiver if and only if the SIR is greater than c for the duration of the packet duration. For example, if two packets, A and B, arrive at the receiver at about the same time (within a packet duration) with overlap as shown in Figure 5.6, there will be three possible outcomes, i.e., (i) packet A will be captured if — > c, (iii) otherwise, both packets will be discarded. Chapter 5 Simulation Results and Discussion - Part 2 67 Packet A Packet B Figure 5.6 The capture model. 5.2 .2 S i m u l a t i o n R e s u l t s In this section, the simulation results are obtained for a log-normal shadowing plus Rayleigh fading channel with capture. The default value of c is chosen to be 10 dB. The through-put performance for 15 WTs in a log-normal shadowing plus Rayleigh fading channel with capture using Method A is presented in Figure 5.7(a). It can been seen that r| with capture is much higher, about 2-3 times more, than that with no capture as in Figure 5.1(a). Because capture can increase the chance of packet reception, especially for those WTs near a RB. The r\ can be greater than 1 for n > 3 because there is a higher possibility that two or more packets can be received in the RBs at about the same time. For example, three packets, A, B and C, are sent from three WTs within a packet duration. If the received power of Packet A at RBj is higher than the sum of the received power Packet B and Packet C, then Packet A will be captured. Similarly, Packet B may be captured at R B 2 if the received power of Packet B is greater than the sum of ime ^ O v e r l a p Chapter 5 Simulation Results and Discussion - Part 2 68 other two packets. In this case, both packets A and B can be received by the RBs within a packet duration. The performance improvement of 2, 3 and 4 RBs over one RB is 66%, 120% and 160% which the performance improvement is greater than that in no capture cases as in Figure 5.1(a). Because capture with more RBs can receive multiple packets. The access delay is shown in Figure 5.7(b). 1.4 0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4 Normalized Offered Load Normalized Offered Load (a) (b) Figure 5.7 (a) Normalized throughput and (b) access delay versus normalized offered load in log-normal shadowing and Rayleigh fading channel with capture for 15 WTs using Method A. Compared to no capture in Figure 5.2(a), x with capture is much lower because the number of retransmissions due to collisions is reduced. For one RB, x goes up when the traffic load is low and drops when the traffic load increases (p > 1.0). This is because when the traffic load is low, the probability of collision is low and the packets sent from most of the WTs can be received at the RB with several retransmissions. However, as load increases, the probability of collision increases, only the packets sent from those WTs close to the RB can be received due to Chapter 5 Simulation Results and Discussion - Part 2 69 power capture. In other words, those packets sent from the WTs which are far away from the RB can never be received by the RB. Therefore, the access delay is calculated based on those packets sent from few WTs that are close to the RB. The distribution of S at each WT for n = 1, 2, 3 and 4 with no capture and capture is shown in Figure 5.8 and Figure 5.9, respectively. In Figure 5.8(a) for one RB, it can be seen that S is more concentrated on WTs 5 to 11 which are more or less under the coverage area of the RB. On other WTs, S is comparatively smaller. As n increases, S at each WT is distributed more evenly. In Figure 5.9(a) for one RB, S is highly concentrated at WTs 7, 8 and 9, especially for WT 8 at which S is about 40% of the total S in the system. Because a packet is captured based on the largest received power which is inversely proportional to the third power of the distance between the transmitter and receiver, packets from a WT which is close to the RB can be captured more frequently. Since WTs 7, 8 and 9 are the closet WTs to the RB as shown in Figure 3.6, their packets can always be captured by the RB. The S at other WTs is significantly lower, which is about the same as no capture cases shown in Figure 5.8(a). Since their packets cannot be captured due to the larger distance, the performance of these WTs is similar to no capture cases. As n increases as shown in Figure 5.9, it can be seen that S is mostly concentrated on the WTs near the RBs. The performance is improved only for the WTs near the RBs. In conclusion, capture takes advantage of the near-far effect which causes an unfair traffic performance for each WTs even though the total system performance is increased. Chapter 5 Simulation Results and Discussion - Part 2 70 0.03 r •3 0.025 o 0.02 0.015 0.01 z 0.005 r 0.03 r 0.025 f 0.02 ^ 0.015 « 0.01 | z 0.005 f 11 1 2 3 4 5 6 7 8 9 101112131415 WTno. (a) 1 2 3 4 5 6 7 8 9 101112131415 WT no. (c) 0.03 •3 0.025 CL 0.02 0.015 6 7 8 9 101112131415 WT no. (b) 0.03 0.025 0.02 0.015 2 0.01 0.005 1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 1 3 1 4 1 5 WTno. (d) Figure 5.8 The throughput distribution of 15 WTs with no capture for (a) one RB, (b) 2 RBs, (c) 3 RBs and (d) 4 RBs. Chapter 5 Simulation Results and Discussion - Part 2 71 0.2 a> 3 Q H •D CO N "ro "0.15 0.1 0.05 0.2 10.15 O) 3 O • ized Th o CO 1 0.05 0 1 2 3 4 5 6 7 8 9 101112131415 WTno. (a) 0.2 3 O •0.15 0.1 0.05 J i i i ll 1 2 3 4 5 6 7 8 9 101112131415 WTno. (C) J L * l Li I , 1 2 3 4 5 6 7 8 9 101112131415 WTno. (b) 0.2 "0.15 0.1 ro I 0.05 1 2 3 4 5 6 7 8 9 101112131415 WT no. (d) Figure 5.9 The throughput distribution of 15 WTs with capture for (a) one RB, (b) 2 RBs, (c) 3 RBs and (d) 4 RBs. The saturated throughput and access delay for 15 WTs in a log-normal shadowing plus Rayleigh fading channel using Method A with capture for a = 6 dB with different transmitted power are shown in Figure 5.10(a) and (b) respectively. It can be seen that S is not affected by Tx very much. This observation is different from the results for no capture in which S increases with Tx as shown in Figure 5.3(a). The reason is as follows. The received packet powers at the RB are higher for the WTs near it. The total number of WTs close to the RB is not changed whether Tx is high or low. Chapter 5 Simulation Results and Discussion - Part 2 72 1.3 <hr 1.2 1.1 1 £ 0.9 2 0 8 * 0.7 0.6 0.5 4 0.4 50 75 100 TX Power (mW) (a) 125 150 75 100 125 TX Power (mW) (b) 150 Figure 5.10 (a) The saturated throughput and (b) access delay as a function of transmitted power in log-normal plus Rayleigh fading channel at o = 6 dB for 15 WTs using Method A with capture. As an example of one RB case, the received packet powers from WTs 7, 8 and 9 are always largest at the RB no matter if the Tx of WTs is high or low. Besides, for high or low Tx, the probability of a packet sent from a close WT being received (non-faded) at the RB is quite high because if is close to the RB. Thus, S is about the same at any Tx for the evaluation range because the packets from the closest WTs can always be captured and the reception (non-faded) probabil-ity of the packets from these WTs are quite high at any Tx. In Figure 5.10(b), x with one RB increases with Tx because the non-faded (above CST) probability of packets at WTs decreases with Tx and the time of busy channel sensed by WTs becomes longer. Thus a longer waiting time is required for WTs to send packets. For n > 2, x is not changed with Tx by much. Although the waiting time for WTs to send packets increases with Tx, the chance of a packet being successfully received at RBs is increased as well due to multiple RBs. Thus, x can be kept at about the same. Chapter 5 Simulation Results and Discussion - Part 2 73 The throughput and access delay for 15 WTs in a log-normal shadowing plus Rayleigh fading channel using Method A with capture at Tx = 100 mW with different values of a are shown in Figure 5.11(a) and (b) respectively. The S for n > 2 decreases slightly with a because the probability of WTs being shadowed at the RBs increases. Therefore, the chance of the WTs being covered by multiple RBs is reduced. Since there is no multiple RBs' effect for one RB case, S is not affected by a . Similarly x with one RB decreases with a because the probability of packets being non-shadowed or detected at the WTs is decreased with c. Since the shadowing effect increases, the number of WTs shadowed from the RB increases. Therefore, the time for fewer stations completing for accessing the channel become less. For n > 2, x is not changed much with a because packets can be received by multiple RBs. & Shadow (SD in dB) (a) Shadow (SD in dB) (b) Figure 5.11 (a) The saturated throughput and (b) access delay as a function of a in log-normal plus Rayleigh fading channel at Tx= 100 mW for 15 WTs using Method A with capture. Chapter 5 Simulation Results and Discussion - Part 2 74 5.3 O t h e r S y s t e m P a r a m e t e r s In this section, the simulation results will be obtained for different system configurations or parameters. A two-dimensional model will be used to compare the performance obtained from one-dimensional. The effect of c, packet length and ACK_Timeout will be considered. 5.3.1 T w o - d i m e n s i o n a l M o d e l The previous results are obtained from one-dimensional model. This section will present the simulation results for a two-dimensional model for comparison. The two-dimensional model contains 15 WTs located in three different rows as shown in Figure 5.12. Figure 5.12 The station location for 15WTs and different number of RBs in a two-dimensional model. Chapter 5 Simulation Results and Discussion - Part 2 75 Each row has 5 WTs and the distance between the WTs is 36.25 meter. Therefore, the network range is still about 145 meter. The distance between each row is 5 meter. The RB(s) are placed in the middle row and spaced evenly. The network arrangement for the system with n = 1, 2, 3 and 4 is shown in Figure 5.12(a) to (d). 0.5 1 1.5 Normalized Offered Load (a) 0.5 1 1.5 Normalized Offered Load (b) Figure 5.13 (a) Normalized throughput and (b) access delay versus normalized offered load in log-normal shadowing plus Rayleigh fading channel for no capture in two-dimensional model. The T) and x for 15 WTs in a log-normal shadowing plus Rayleigh fading channel with no capture using Method A in a two-dimensional model are presented in (a) and (b), respectively. The r\ and x are slightly worse than those in one-dimensional model in Figure 5.1(a) and Figure 5.2(a). The distribution of S on each WT in a two-dimensional model is shown in Figure 5.14. Compared to one-dimensional model in Figure 5.8, it can be seen that S on each WT is slightly lower in two-dimensional model. The first reason is that there are more packets being faded at the RB in a two-dimensional model because there are more WTs located far away from RB. An example of a single RB in one-dimensional model, there are two WTs 1 and 15 located farthest Chapter 5 Simulation Results and Discussion - Part 2 76 away from RB. But, in a two-dimensional model, there are more WTs 1, 5, 6, 10, 11 and 15 located farthest away from RB. The second reason is that there are more possible hidden terminals for each WT. For example, two terminals are hidden from each other if the distance between them is larger than about 70 m. Then, for one-dimensional model, WT 1 has only one hidden terminal, WT 15. However, for a two-dimensional model, WT 1 has three hidden terminals, WTs 5, 10 and 15. Thus, more possible hidden terminals increase the occurrence of collisions. Fi gure 5.14 The saturated throughput distribution for 15 WTs on each WT with no capture in a two-dimensional model for (a) one RB, (b) 2 RBs, (c) 3 RBs and (d) 4 RBs. Chapter 5 Simulation Results and Discussion - Part 2 77 Normalized Offered Load Normalized Offered Load (a) (b) Figure 5.15 (a) Normalized throughput and (b) access delay versus normalized offered load in log-normal shadowing, a = 6, and Rayleigh fading channel at Tx - 100 mW with capture for 15 WTs using Method A in a two-dimensional model. The T| and % for 15 WTs in a log-normal shadowing plus Rayleigh fading channel with capture using Method A in a two-dimensional model are presented in Figure 5.15(a) and (b), respectively. The performance with capture is much better than that with no capture. Compared to one-dimensional model in Figure 5.7, the performance in a two-dimensional model for n = 1 is about the same, but the performance is worse for n > 2. In a two-dimensional model, the perfor-mance of 3 RBs is better than that for 4 RBs whereas in previous results, the performance of 3 RBs is worse than that with 4 RBs. This can be explained by looking at the distribution of S on each WT as shown in Figure 5.16. With capture, most of the packets being received or captured are sent from the WTs near the RB. For example with one RB as shown in Figure 5.16(a), most of the packets captured by the RB are sent from WTs 3, 8 and 13 which are close to the RB. The S for other WTs is similar to that of the no capture cases as seen in Figure 5.14(a). For three RBs' Chapter 5 Simulation Results and Discussion - Part 2 78 case, each RB is located in columns 2, 3 and 4. Similarly, most of the packets being captured are from the WTs located in columns 2, 3 and 4 as seen in Figure 5.16(c). For four RBs' case, each RB is located between columns. For example, if a RB is located between columns 1 and 2, the received power from the WTs in these two columns at the RB will be more or less the same. Hence, if two packets are sent from these two columns, the RB may not effectively capture one of the packets. The throughput is low in column 3 than that in columns 2 and 4 is because the interference level for column 3 is higher. For example, a RB is located between columns 2 and 3. If a packet is sent from a W T in column 2, the interference power comes from the WTs in columns 1, 3, 4, and 5. If a packet is sent from a WT in column 3, the interference power comes from the WTs in columns 1,2, 4, and 5. In addition, the received power of both packets is the same at the RB. Obviously, the total interference power of columns 1, 3, 4, and 5 is lower than that of columns 1, 2, 4, and 5. The Throughput is low in columns 1 and 5, because there is only one RB close by but there are at least two RBs located between columns 2, 3, and 4. Chapter 5 Simulation Results and Discussion - Part 2 7 9 Figure 5.16 The saturated throughput distribution for 15 WTs on each WT with capture in a two-dimensional model for (a) one RB, (b) 2 RBs, (c) 3 RBs and (d) 4 RBs. The saturated throughput and access delay for 15 WTs a in log-normal shadowing plus Rayleigh fading channel using Method A with capture for o = 6 dB with different transmitted power are shown in Figure 5.17(a) and (b) respectively. Similar to a one-dimensional model in Figure 5.10, S or x for any n is not affected by Tx very much except that x for one RB increases with Tx. Chapter 5 Simulation Results and Discussion - Part 2 80 J3 75 100 125 TX Power (mW) (a) 150 75 100 125 TX Power (mW) (b) 150 Figure 5.17 (a) Saturated throughput and (b) access delay as a function of transmitted power in log-normal plus Rayleigh fading channel at o = 6 dB for 15 WTs using Method A with capture in a two-dimensional model. 5.3 .2 P a c k e t L e n g t h The packet length used in the previous simulations is 8000 bits. In this section, the effect of the packet size, N, on the performance will be examined. The saturated throughput in a log-normal shadowing plus Rayleigh fading channel for 15 WTs with capture for different N is shown in Figure 5.18. It can be seen that S is more or less constant with /Vfor any n. For a given period of time, the number of packet received is smaller for a longer packet. But a longer packet contains more information bits. Hence, the throughput in terms of average number of bits per second successfully received will be the same for any N. Even though S is independent of N, x increases with N as shown in Figure 5.18(b). Since the transmission time is longer for larger TY, the channel will be sensed as busy for a longer time. Thus, the WTs need to wait for a longer time to send their packets. In conclusion, longer packet size does not improve the throughput, but it degrades the Chapter 5 Simulation Results and Discussion - Part 2 81 access delay. However, the length of packet size mainly depends on the types of message being delivered. Longer packets reduce the fraction of packet overhead. 1.4 0.6 4 RBs 3 RBs 2 RBs 1 RB 0.4 4000 8000 Packet Size (bits) (a) 12000 8000 Packet Size (bits) (b) 12000 Figure 5.18 (a) Saturated throughput and (b) access delay as a function of packet size in log-normal shadowing, a = 6, and Rayleigh fading channel at Tx - 100 mW with capture for 15 WTs using Method A. 5.3 .3 E f f e c t o f A C K _ T i m e o u t The default value of ACK_Timeout is 300 us in the previous results. The saturated throughput and access delay for 15 WTs in a log-normal shadowing plus Rayleigh fading channel with no capture and capture as a function of ACK_Timeout are shown in Figure 5.19 and Figure 5.20. The access delay with no capture and capture are shown in Figure 5.21 and Figure 5.22 respectively. It can be seen that the performance, S and x, is not changed with ACK_Timeout for both Method A and B. For Method A, since only a single A C K is returned from a RB, the desired WT after sending a packet is required to wait for the transmission time of the ACK, 112 p, s, SIFS, 28 ps, and the propagation time, < 1 ps, which is about 141 ps. If the A C K is lost, the desired Chapter 5 Simulation Results and Discussion - Part 2 82 g. 0.25 _ o _ + + J " - -4 R B s (a l l ) 3 R B s (a l l ) 2 R B s (a l l ) 4 R B s O + • -3 R B s _ 2 R B s _ 1 R B -1 1 1 200 300 400 500 600 ACKlTIMEOUT (US) Figure 5.19 Saturated throughput as a function of ACK_Timeout for 15 WTs with no capture in log-normal shadowing plus Rayleigh fading channel using Method A and Method B. WT will follow the backoff procedures to retransmit the packet only when the ACK_Timeout is expired. If the ACK_Timeout is longer, the desired WT needs to wait for a longer time to retrans-mit the packet and thus the performance is degraded. In all four figures, however, it can be seen that the performance does not degrade with increasing ACK_Timeout. This is because most of the A C K can be received correctly by the desired WT. The reason is that if the packet sent from the desired WT can be successfully received at the RB, the A C K returned from that RB can also be received at the desired WT with a equally high probability. With capture, the chance of receiving a A C K is higher. It can be seen that the performance of Method B is lower than that of Method A. This is because multiple ACKs increase the channel busy time. Chapter 5 Simulation Results and Discussion - Part 2 83 1.4 1.2 • n ' 0.8 0.6 4 R B s (a l l ) 3 R B s (a l l ) 2 R B s (a l l ) 4 R B s 3 R B s 2 R B s 1 R B O -t-• 0.4 200 300 400 ACK_TIMEOUT (us) 500 600 Figure 5.20 Saturated throughput as a function of ACK_Timeout for 15 WTs with capture in log-normal shadowing plus Rayleigh fading channel using Method A and Method B. Figure 5.21 Access delay as a function of ACK_Timeout for 15 WTs with no capture in log-normal shadowing plus Rayleigh fading channel using Method A (a) and Method B (b). Chapter 5 Simulation Results and Discussion - Part 2 84 S 40 50 200 300 400 500 ACK_TIMEOUT (us) (a) 600 30 4 R B s (a l l ) 3 R B s (a l l ) 2 R B s (a l l ) 1 R B 200 300 400 500 ACK_TIMEOUT (us) (b) 600 Figure 5.22 Access delay as a function of ACK_Timeout for 15 WTs with capture in log-normal shadowing plus Rayleigh fading channel using Method A (a) and Method B (b). 5.3.4 E f f e c t o f C a p t u r e R a t i o The capture ratio, c, used in previous sections is 10 dB. In this section, the effect of differ-ent values of c is investigated. Figure 5.23(a) and (b) show the saturated throughput and access delay with various values of c for 15 WTs in a log-normal shadowing plus Rayleigh fading channel respectively. As expected, the performance improves with decreasing c because a packet can be more easily to be captured for a smaller value of c. However, the value of c cannot be chosen arbitrary because it depends on the system characteristics such as modulation schemes and coding techniques. Figure 5.23 (a) Saturated throughput and (b) access delay as a function of c for 15 WTs with capture in log-normal shadowing plus Rayleigh fading channel. Chapter 6 Conclusions In this thesis, the performance, in terms of normalized throughput, T ) , and access delay, x, of the proposed W L A N with distributed RBs using the IEEE 802.11 M A C protocol is studied extensively by computer simulations. The simulation model is validated against previous results. The channel environments used in the simulation are A W G N , Rayleigh fading, log-normal shadowing and log-normal shadowing plus Rayleigh fading channel. The effect of various parameters, such as number of WTs, number of RBs, n, transmitted power level, Tx, and standard deviation, a, of the log-normal shadowing are examined. Two methods used for the RBs to return the ACKs are studied. It was found that Method A can provide better performance than Method B in all channel environments and ranges of parameters considered. The simulation results with no-capture assumption are obtained for all four channel models. In AWGN channel, it is found that T) and x do not change with n. The r| is the same for different number of WTs and about 0.9, but x with more WTs is higher. Thus, using more RBs is not desirable in A W G N channel. With the other three channel models, results show that the performance can always be improved using more RBs. The performance improvement of using multiple RBs to single RB depends on the system parameters, i.e., number of WTs, Tx, a and n. Use the default values of Tx = 100 mW and a = 6 dB as an example. In a Rayleigh fading channel, the saturated throughput (access delay), S fx), with 4 RBs for 15 WTs and 30 WTs can be improved by about 40% (30%) and 100% (40%) respectively. In a log-normal shadowing channel, S with 4 RBs for 15 WTs and 30 WTs can be improved by 28% and 38% respectively. The x is about the same. In a log-normal shadowing plus Rayleigh fading channel, S with 4 RBs 86 Chapter 6 Conclusions 87 for 15 WTs and 30 WTs can be improved by 100% and 200% respectively. The x can be improved by about 50% for both 15 and 30 WTs. Therefore, the results show that the distributed W L A N architecture employing multiple RBs is more effective in a system with a larger number of WTs in terms of performance improvements over a system with a single RB. In a Rayleigh fading channel, S with any n increases with Tx, but x is not affected very much by Tx. Because for low Tx each RB can only cover few WTs; therefore, each WT covered by a RB needs only to wait for few other WTs completing their transmissions in order to send its packet. The performance improvement of multiple RBs over single RB at low Tx is much greater, about 10 times from the evaluation range considered in figures, than that at high Tx. In a log-normal shadowing channel, S ( x ) decreases (increases) with a . The performance improvement at high a is much greater, about 15 times from the evaluation range considered in figures, than that at low a . In a log-normal shadowing plus Rayleigh fading channel, the performance is more or less independent of Tx and a . The S with 4 RBs for 30 WTs can be improved by at least 120% over S with a single RB for any values of Tx and a . The simulation results with capture are only obtained for a log-normal shadowing plus Rayleigh fading channel. With use of capture, the performance can be improved significantly with increasing n. The performance improvement of multiple RBs over one RBs is greater than that for no capture cases. For Tx = 100 mW and a = 6 dB, S with 4 RBs and 15 WTs can be improved about 160% over one RB. In general, the performance is not affected significantly by Tx or a except that the access delay with one RB increases (decreases) with Tx(a). Chapter 6 Conclusions 88 The impact of a two-dimensional model, packet size and ACK_Timeout on the perfor-mance of distributed RBs are investigated for no capture and capture in a log-normal shadowing plus Rayleigh fading channel. It is found that the configuration of the system (one- or two-dimensional) will affect the performance slightly because the distances between WTs or the distances between WTs and RBs may vary. Thus, it will affect the number of hidden terminals and the occurrence of collisions. The S is independent of packet size, but T degrades with packet size. Therefore, it is better to use a shorter packet size rather than a longer packet size. However, the packet size depends on the types of message. It is found that the ACK_Timeout does not affect the performance in both no capture and capture cases. As expected, the performance is increased with decreasing the capture ratio because small value of capture ratio allows more packets to be captured easily. However, this value to be chosen depends on the system configuration. In conclusions, distributed RBs are effective in a channel with fading or shadowing. The performance can always be improved with increasing n over a single RB, especially for a channel severely degraded by fading or shadowing. In a channel with fading or shadowing, a packet might not be received at the RB. Thus, there is a high chance that a packet is discarded at the RB. Multiple RBs can increase the packet reception probability at one of the RBs for fading cases because this is a relatively low probability that a packet is faded at all RBs. For shadowing cases, multiple RBs can be considered as a macrodiversity (multiple-base-station diversity) because if a WT is shadowed from one RB, there is a good chance that the WT might not be shadowed from another RB by adding more RBs. Packets might be faded or shadowed at the WTs so that some WTs may be hidden from each other. While two or more WTs, which are hidden from each other, send the packets at about the same time, those packets are collided and discard at the RB with no-capture. Therefore, for a fixed number of RBs the performance of a system with more WTs is Chapter 6 Conclusions 89 worse than that with less WTs. Because more WTs increase the number of hidden terminals and thus the number of collisions at RBs is increased as well. With capture, the performance improve-ment even better than that with no capture. However, it causes the near-far effect problem in which the performance of the WTs close to the RB can be improved greatly. Among the related topics for further study are the following: (i) To study the performance by theoretical analysis rather than computer simulation. (ii) To study the performance with mobile WTs. (iii) To use RTS/CTS to improve the performance. (iv) To use different channel for uplink and downlink. Glossary 90 X Access Delay. AP Access Point. A C K Acknowledgment. AWGN Additive White Gaussian Noise. TQ Ambient Room Temperature. BSA Basic Service Area. BSS Basic Service Set. BER Bit-error Rate. k Boltzmann's Constant. c Capture Ratio. CS Carrier Sense. C S M A / C A Carrier Sense Multiple Access with Collision Avoidance. CST Carrier Sensing Threshold. CFP Contention Free Period. CP Contention Period. CW Contention Window. DSSS Direct Sequency Spread Spectrum. d Distance between the transmitter and the receiver. DCF Distributed Coordination Function. DIFS Distributed Coordination Function Interframe Space. DS Distribution System. DSS Distribution System Services. Tfo Effective Background Temperature. T Effective Noise Temperature. BW Equivalent Bandwidth of the Receiver. ELFS Extended Interframe Space. ESA Extended Service Area. ESS Extended Service Set. FCS Frame Check Sequence. FHSS Frequency Hopping Spread Spectrum. IFS Interframe Space. LOS Line of sight. LANs Local Area Networks. M A C Medium Access Control. MPDU M A C Protocol Data Unit. MSDU M A C Service Data Unit. NAV Network Allocation Vector. F Noise Figure. p Normalized Offered Load. T\ Normalized Throughput. n Number of Radio Bridges. OBS Obstructed. OPNET OPtimized Network Engineering Tools. 92 /V Packet Size. y Path Loss Exponent. PHY Physical Layer. PCF Point Coordination Function. PIFS Point Coordination Function Interframe Space. RBs Radio Bridges. RTS/CTS Ready to Send/Clear to Send. P Received Power. G Receiver Antenna Gain. P Receiver Noise Floor. n RS Receiver Sensitivity. dg Reference Distance. S Saturated Throughput. SIFS Short Interframe Space. SNR Signal-to-noise Ratio. CJ Standard Deviation for the log-normal shadowing (dB). SS Station Service. TSF Timing Synchronization Function. P (T ) Transmitted Power. Gf Transmitter Antenna Gain. X Wavelength. WIN Wireless in Building Network. WLANs Wireless Local Area Networks. WTs Wireless Terminals. L„ Zero-mean Gaussian Distributed Random Variable Bibliography [1] V. C. M . Leung, "Diversity Interconnection of Wireless Terminals to Local Area Networks via Radio Bridges", Electronics Letters, vol. 28, no. 5, pp. 489-490, Feb. 1992. [2] C. T. Lau and C. Leung, "Capture Models for Mobile Packet Radio Networks", IEEE Transactions on Communications, vol. 40, no. 5, pp. 917-925, May 1992. [3] J. C. Arnbak and W. V. Blitterswijk, "Capacity of Slotted A L O H A in Rayleigh-Fading Channels", IEEE Journal on Selected Areas in Communications, vol. SAC-5, no. 2, pp. 261-269, Feb. 1987. [4] V. C. M . Leung and A. W. Y. Au, "A Wireless Local Area Network Employing Distributed Radio Bridges", ACM/Baltzer Wireless Network Journal, vol. 2, pp. 97-107, Oct 5, 1995. [5] V. C. M . Leung, "Internetworking Wireless Terminals to Local Area Networks via Radio Bridges", Proceedings of the IEEE International Conference on Selected Topics in Wireless Communications, Vancouver, B.C. pp. 126-129, June 1992. [6] IEEE, Piscataway, NJ, IEEE 802.11 - Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications, Dec. 1994. P802.il/Dl. [7] IEEE, Piscataway, NJ, IEEE P802.ll - Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications, May. 9, 1997. P802.11D6.1. [8] S. M . Nor and I. Ismail, "Performance of CSMA-CA M A C Protocol for distributed Radio Local Area Networks", IEEE International Symposium on Personal Indoor and Mobile Radio Communications, vol. 2, pp. 912-914, Toronto, Canada, 1995. [9] J. Weinmiller, H. Woesner and A. Wolisz, "Analyzing and Improving the IEEE 802.11 M A C Protocol for Wireless LANs", International Workshop of MASCOTS'96: Modeling Analysis, and Simulation, pp. 200-206, San Jose, CA, 1996. [10] J. Weinmiller, H. Woesner, J. P. Ebert and A. Wolisz, "Analyzing the RTS/CTS Mechanism in the D F W M A C Media Access Protocol for Wireless LANs", Proceedings of IFIP TC6 Workshop Personal Wireless Communications (Wireless Local Access), Prague, Czech Republic, pp. 117-130, April 1995. 94 Bibliography 95 [11] H. Woesner, J. Weinmiller, J. P. Ebert and A. Wolisz, "Modified Backoff Algorithms for DFWMAC's Distributed Coordination Function", Proceedings of 2 ITG Fachtagung Mobile Kommunikation '95, Neu-Ulm, Germany, pp. 363-370. Sept. 1995. [12] M . A. Visser and M . E. Zarki, "Voice and Data transmission over an 802.11 Wireless network", IEEE International Symposium on Personal, Indoor and Mobile Radio Communications, vol. 2, pp. 648-652, Toronto, Canada, 1995. [13] G. Bianchi, L. Fratta and M . Oliveri, "Performance Evaluation and Enhancement of the C S M A / C A M A C Protocol for 802.11 Wireless LANs", The Seventh IEEE International Symposium on Personal, Indoor and Mobile Radio Communications, vol. 2, pp. 392-396, Taipei, Taiwan, Oct 15-18, 1996. [14] G. Bianchi, "IEEE 802.1 l~Saturation Throughput Analysis", IEEE Communications letters, vol. 2, no. 12, pp. 318-320, Dec 1, 1998. [15] H. S. Chhaya, "Performance evaluation of the IEEE 802.11 M A C Protocol for Wireless LANs", M . S. thesis, Dep. of Elec. and Comp. Eng, Illinois Institute of Technology, Chicago, Illinois, May 1996. [16] H. S. Chhaya and S. Gupta, "Performance of Asynchronous Data Transfer methods of IEEE 802.11 M A C Protocol", IEEE Personal Communications, vol. 3, no. 5, pp. 8-15, Oct. 1996. [17] J. H. Kim and J. K. Lee, "Performance Analysis of M A C protocols for Wireless L A N in Rayleigh and Shadow Fading Channels", IEEE Global Telecommunications Conference, vol. 1, pp. 404-408, Phoenix, AZ, 1997. [18] A. Zahedi and K. Pahlavan, "Throughput of a Wireless L A N access point in presence of natural hidden terminals and capture effects", The 7th IEEE International Symposium on Personal, Indoor and Mobile Radio Communications, vol. 2, pp. 397-401, Taipei, Taiwan, Oct. 1996. [19] A. Zahedi and K. Pahlavan, "Natural Hidden Terminal and the Performance of the Wireless LANs", Annual International Conference on Universal Personal Communications Record. vol. 2, pp. 929-933, San Diego, CA, 1997. [20] D. Buchholz, P. Odlyzko, M . Taylor and R. White, "Wireless In-Building Network Architecture and Protocols", IEEE Network Magazine, vol. 5, no. 6, pp. 31-38, Nov. 1991. Bibliography 96 [21] W. K. W. Cheung, "Performance Study of the IEEE 802.11 Wireless Media Access Control Protocols", M . A. Sc. thesis, Dep. Elec. Eng, Univ. British Columbia, Vancouver, Canada, 1995. [22] B. P. Crow, I. Widjaja, J. G. Kim and P. Sakai, "Investigation of the IEEE 802.11 Medium Access Control (MAC) Sublayer Functions", Proceedings of IEEE INFOCOM '97, vol. 1, pp. 126-133, Kobe, Japan, April 1997. [23] B. P. Crow, I. Widjaja, J. G. Kim and P. T. Sakai, "IEEE 802.11 Wireless Local Area Networks", IEEE Communications Magazine, vol. 35, no. 9, pp. 116-126, Sept. 1997. [24] L. Tang, "Modeling and Performance Analysis of In-Building Distributed W L A N using C S M A with Spread Spectrum Signaling", M . A. Sc. thesis, Dep. Elec. Eng, Univ. British Columbia, Vancouver, Canada, 1995. [25] A. W. Y. Au, "Multi-Receiver Performance of Slotted A L O H A Multiple Access with Direct Sequence Spread-Spectrum signalling for Wireless In-Building networks", M . A. Sc. thesis, Dep. Elec. Eng, Univ. British Columbia, Vancouver, Canada, 1993. [26] C. Roobol, "On the packet delay in Wireless Local Area Networks with Access Port Diversity and Power Control", Proceedings of 6th International Symposium on Personal, Indoor and Mobile Radio Communications, vol. 2, pp. 917-920, Toronto, Canada, Sept. 1995. [27] Optimized Network Engineering Tools (OPNET) Tutorial Manual, Release 3.0, MIL 3 Inc. Washington, DC. August 1996. [28] T. S. Rappaport, Wireless Communications Principles and Practice. Prentice Hall, Upper Saddle River, NJ, 1996. ISBN 0-7803-1167-1. [29] T. S. Ho and K. C. Chen, "Performance Analysis of IEEE 802.11 C S M A / C A Medium Access Control Protocol", IEEE International symposium on Personal, Indoor and Mobile Radio Communications, vol. 2, pp. 407-411, Taipei, Taiwan, 1996. 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0065166/manifest

Comment

Related Items