1 Physical Layer DSP Design of a Wireless Gigabit/s Indoor LAN by Eladio Clemente Arvelo Submitted to the Department of Electrical Engineering and Computer Science in Partial Fulfillment of the Requirements for the Degree of Master of Engineering in Electrical Engineering and Computer Science at the Massachusetts Institute of Technology jointly with QUALCOMM Incorporated May 2000 Copyright 2000 Eladio C. Arvelo. All rights reserved. The author hereby grants to M.I.T. permission to reproduce and distribute publicly paper and electronic copies of this thesis and to grant others the right to do so. Author Department of Electrical Engineering and Computer Science May 22, 2000 Certified by Charles G. Sodini MIT Thesis Supervisor Certified by Robert P. Gilmore QUALCOMM Thesis Supervisor Accepted by Arthur C. Smith Chairman, Department Committee on Graduate These
3 Physical Layer DSP Design of a Wireless Gigabit/s Indoor LAN by Eladio Clemente Arvelo Submitted to the Department of Electrical Engineering and Computer Science May 22, 2000 In Partial Fulfillment of the Requirements for the Degree of Master of Engineering in Electrical Engineering and Computer Science at the Massachusetts Institute of Technology jointly with QUALCOMM Incorporated Abstract The Wireless Gigabit/s Local-Area Network (WGLAN) project is aimed at providing highspeed data transmission between the Next Generation Internet and end-use devices within the home or office environment. The design of the digital signal processing (DSP) required at the physical layer of the network is the focus of this thesis. In particular, this thesis models the indoor radio channel environment at the 5.x GHz Unlicensed National Information Infrastructure (U-NII) frequency band, and proposes a multipath-resistant transceiver design based on Orthogonal Frequency Division Multiplexing (OFDM) with adaptive multilevel Quadrature Amplitude Modulation (M-QAM). The proposed network design allows two-way communication through a Time Division Duplexing (TDD) scheme and provides multiuser support through a series of algorithms that establish session links and allocate subchannels among devices in an optimal way. Finally, a custom-written software simulation is used to estimate the bit error rate (BER) network performance under different channel conditions. MIT Thesis Supervisor: Charles G. Sodini Title: Professor, Electrical Engineering and Computer Science
5 Acknowledgments My heartful appreciation goes to my thesis supervisors, Prof. Charles Sodini at MIT who provided me with the opportunity to work on the design of such an interesting project as the WGLAN system, and Rob Gilmore at Qualcomm who provided the technical expertise to shape the focus of this thesis as well as verifying the final design. I would like to thank Qualcomm Incorporated and the MIT VI-A Internship Program for sponsoring this thesis. In addition, I would like to acknowledge the helpful comments provided by the engineering team at Qualcomm, in particular by Rajiv Vijayan, who provided assistance on software simulation issues, and Prof. Elvino Sousa, who found time to review my thought process regarding OFDM system design. For providing the brain power to discuss technical issues related to this thesis, I would like to thank Beng-Teck Lim and Durodami Lisk for their valuable time. I would also like to thank my fellow MIT Co-Op interns at Qualcomm for making the summer an enjoyable time. Finally, I would like to dedicate this thesis to my parents, sister, and extended family who have always supported me in pursuing my own dreams wherever they may lead.
6 Table of Contents CHAPTER 1 : INTRODUCTION SYSTEM SPECIFICATIONS THESIS FOCUS THESIS OUTLINE CHAPTER 2 : WLAN CONCEPTS & STANDARDS WLAN TECHNOLOGIES Narrowband Technology Spread Spectrum Technology Infrared Technology Efficiency Considerations WLAN CONFIGURATIONS RADIO SPECTRUM REGULATIONS IEEE WLAN STANDARD Network Topologies Physical Layer Architecture FHSS Physical Layer DSSS Physical Layer HIPERLAN STANDARD Channel Access Physical Layer BLUETOOTH SPECIFICATION CHAPTER 3 : CHANNEL ANALYSIS CHARACTERIZING INDOOR RADIO CHANNELS Power Delay Profile Path Loss Delay Spread Coherence Bandwidth Doppler Spread Small-Scale Flat Fading CHANNEL PARAMETERS Propagation at 5.2 GHz Propagation at 5.8 GHz Selected Parameters CHANNEL MODELS LINK BUDGET Frequency Band Allocations Controller s Link Budget Adapter s Link Budget CHANNEL CAPACITY CHAPTER 4 : MODULATION AND DEMODULATION THE OFDM CONCEPT Conventional FDM Technology Orthogonal FDM Technology TRANSMITTER BLOCK DIAGRAM
7 4.3 RECEIVER BLOCK DIAGRAM CHOOSING SYSTEM PARAMETERS CHAPTER 5 : MULTIPLE ACCESS SCHEME NETWORK TOPOLOGY NEGOTIATING A COMMUNICATION LINK Controller-to-Adapter Link Request Adapter-to-Controller Link Request Adapter-to-Adapter Link Request CONTROL CHANNEL Transmitting bits through the Control Channel Receiving bits through the Control Channel DATA CHANNELS Preliminary Considerations Subchannel Allocation Algorithm Special Cases LINK MAINTENANCE CHAPTER 6 : SOFTWARE SIMULATION SIMULATION MODEL SOFTWARE ENVIRONMENT Parameters Header File Simulator Program File Data Abstraction Files SIMULATION SCENARIOS AND RESULTS Static Channel Scenarios Dynamic Channel Scenarios Interference Channel Scenarios Subchannel Capacity Estimation CHAPTER 7 : CONCLUSION CHANNEL CHARACTERISTICS MODULATION AND DEMODULATION MULTIPLE ACCESS SCHEME SIMULATION AND FUTURE WORK BIBLIOGRAPHY. 89 APPENDIX I: SIMULATION RESULTS.. 95 APPENDIX II: SIMULATION PROGRAM 113 7
8 List of Figures FIGURE 1. WGLAN TOPOLOGY FIGURE 2. DESIGN METHODOLOGY FIGURE 3. NARROWBAND DIVIDES THE SPECTRUM INTO SUBCHANNELS WHILE SPREAD SPECTRUM EXTENDS THE ENTIRE BAND FIGURE 4. FHSS SYSTEMS HOP AMONG FREQUENCY CARRIERS AT PRE-DETERMINED INTERVALS OF TIME FIGURE 5. CONSIDER A 6-BIT CHIP SEQUENCE. TO SPREAD A DIGITAL DATA STREAM, EACH 1 IN THE STREAM IS SUBSTITUTED WITH THE SPECIFIED CHIP SEQUENCE, WHILE EACH 0 IS SUBSTITUTED WITH THE NEGATED SEQUENCE FIGURE 6. IBSS VERSUS ESS NETWORKS FIGURE 7. THE STANDARD SPECIFIES THE MAC AND PHY LAYER. THE PHY LAYER, IN TURN, CONSISTS OF THE PLCP AND PMD SUBLAYERS FIGURE 8. PLCP FRAME CORRESPONDING TO AN FHSS PHYSICAL LAYER IMPLEMENTATION FIGURE 9. PLCP FRAME CORRESPONDING TO A DSSS PHYSICAL LAYER IMPLEMENTATION FIGURE 10. (A) TIME DISPERSION AND (B) AMPLITUDE FADING ON MULTIPATH CHANNELS FIGURE 11. DIFFERENCE BETWEEN THE PARAMETERS T M, τ M, AND τ RMS IN A PDP P(T) FIGURE 12. FDD SCHEME DIFFICULTIES DUE TO COMMUNICATION BETWEEN ADAPTERS FIGURE 13. CHANNEL CAPACITY AS A FUNCTION OF DISTANCE FIGURE 14. TRANSCEIVER ARCHITECTURE FOR CONVENTIONAL FDM FIGURE 15. TYPICAL SPECTRUM FOR CONVENTIONAL FDM FIGURE 16. GRAPHICAL INTERPRETATION OF THE OFDM CONCEPT FIGURE 17. TYPICAL SPECTRUM FOR ORTHOGONAL FDM FIGURE 18. BLOCK DIAGRAM OF THE OFDM TRANSMITTER FIGURE 19. BLOCK DIAGRAM OF THE OFDM RECEIVER FIGURE 20. STAR-LIKE TOPOLOGY OF THE WGLAN FIGURE 21. TDD SCHEME FIGURE 22. CONTROLLER-TO-ADAPTER LINK NEGOTIATION FIGURE 23. ADAPTER-TO-ADAPTER LINK NEGOTIATION FIGURE 24. DISCRETE-TIME SIMULATION MODEL FIGURE 25. MODULE DEPENDENCY DIAGRAM OF THE SIMULATION PROGRAM FIGURE 26. BER PERFORMANCE FOR IDEAL CHANNEL MODEL FIGURE 27. BER PERFORMANCE FOR 64-QAM STATIC CHANNEL MODELS FIGURE 28. BER PERFORMANCE FOR 64-QAM DYNAMIC CHANNEL MODELS FIGURE 29. BER PERFORMANCE FOR 64-QAM INTERFERENCE CHANNEL MODELS FIGURE 30. BER PERFORMANCE FOR EXPONENTIAL CHANNEL MODELS
9 List of Tables TABLE 1. FREQUENCY BANDS ALLOCATED FOR USE BY ISM EQUIPMENT. FCC CODE PART TABLE 2. MAXIMUM POWER RESTRICTIONS OF THE U-NII BAND TABLE 3. FCC RESTRICTIONS ON THE U-NII BAND TABLE 4. LINK BUDGET FOR NETWORK CONTROLLER TABLE 5. LINK BUDGET FOR NETWORK ADAPTERS TABLE 6. SYSTEM PARAMETER CONSTRAINTS TABLE 7. SYSTEM PARAMETER VALUES TABLE 8. OFDM LINK REQUEST SCENARIO TABLE 9. SUBCHANNEL ALLOCATION BY MAXIMIZING EACH ADAPTER S BANDWIDTH EFFICIENCY TABLE 10. SUBCHANNEL ALLOCATION BY MAXIMIZING OVERALL BANDWIDTH EFFICIENCY TABLE 11. SUBCHANNEL ALLOCATION BY EQUITATIVE MAXIMIZATION OF BANDWIDTH EFFICIENCY TABLE 12. OPTIMIZING SUBCHANNEL ALLOCATIONS BASED ON USER SATISFACTION TABLE 13. PROPORTIONAL SCALING OF OFDM TRANSCEIVER PARAMETERS TABLE 14. MINIMUM SNR NEEDED TO ACHIEVE A TARGET BER FOR DIFFERENT CONSTELLATION SIZES.83 TABLE X GHZ INDOOR CHANNEL PARAMETERS TABLE 16. OFDM SYSTEM PARAMETER TABLE 17. SNR THRESHOLDS FOR A GIVEN BER AND CONSTELLATION SIZE
11 Chapter 1 : Introduction The latest advances in digital technology and explosive growth of the Internet have revolutionized the ways we handle daily information. Today is a common form of communication, current music hits are sold via the Web, meeting schedules are managed with Personal Digital Assistants (PDAs), and video conferencing is enable by low-cost equipment. Following the current trend in multimedia innovations, new electronic devices and applications are being conceived to handle larger amounts of information at faster speeds. In the near future, complete music albums may be distributed over the Net directly into a home entertainment system; video archives of yesterday s headlines may be downloaded from a news service into a High-Definition Television (HDTV); and company employees may use telepresence to interact with fellow workers overseas. Overall, the projected load on the Internet in the near future is so large that a Next Generation Internet (NGI)  has been proposed to handle the information needs of these novel devices. As the NGI delivers high-speed data between geographically distant locations, there is a need to develop an indoor local-area network (LAN) within the home or office to continue high-speed data transmission to end-use devices. Moreover, based on the portability of these devices, the challenge is to provide a low-power wireless solution for this network. This thesis builds upon the anticipated NGI to propose the physical layer digital signal processing (DSP) design of an indoor Wireless Gigabit Local-Area Network (WGLAN) capable of approaching gigabit-per-second (Gbps) data transmission rates at its peak performance.
12 1.1 System Specifications As shown in Figure 1, the WGLAN topology consists of a single network controller that is the gateway between the NGI and the local-area network, and multiple network adapters connected to end-use devices. The network controller is an advanced workstation that estimates channel properties and allocates system resources among competing adapters. The network adapters are peripheral devices to end-use appliances that consist of a digital signal processor, which interfaces with the appliance; a baseband analog processor, which maps blocks of bits onto their respective analog signals and viceversa; and an RF transceiver, which modulates and demodulates the analog signals onto a specific carrier frequency. WGLAN Adapter WGLAN Adapter Appliance Network Controller w/ channel processor WGLAN Adapter Appliance NGI RF Transceiver Baseband Analog Processor Digital Signal Processor Appliance Figure 1. WGLAN topology The WGLAN should support bidirectional communication for any controller-adapter or adapter-adapter pairing. In addition, all communication links should provide real-time data transmission and symmetric data throughputs in both link directions to support those interactive multimedia applications that have intensive audio and video streaming requirements. 12
13 In order to maximize system capacity, the network should employ modulation techniques that are bandwidth efficient. Thus, there is a special interest to dynamically adjust the bit rate of a communication link to the maximum possible M-level Quadrature Amplitude Modulation (M-QAM) based on the signal-to-noise ratio (SNR) and signal-to-interference ratio (SNI) of the channel. In addition, the use of Orthogonal Frequency Division Multiplexing (OFDM) should be considered as an efficient way of partitioning the available channel. Finally, the allocation of system resources should be optimal enough to satisfy the simultaneous requests of multiple users. Since different users will typically request varying degrees of data throughput and quality of service, the network should be flexible enough to meet different user requirements based on the channel conditions for each user. 1.2 Thesis Focus The focus of this thesis is to the design the functionality of the digital signal processor of the network controller and adapters. Such functionality includes dividing the radio spectrum into subchannels, implementing bandwidth-efficient modulation and demodulation techniques, and enforcing custom algorithms for the allocation of network resources among users. Since the design of the DSP component requires an evaluation of design tradeoffs at a system-wide level, this thesis uses an incremental methodology based on the communications model shown in Figure 2. Starting at the center of this model, the first step is to analyze the channel conditions in the indoor environment, then propose a suitable modulation and demodulation technique, and finally suggest protocols for supporting multiple users. User 1 User 1 User 2 modulation channel demodulation User 2 User N User N Figure 2. Design Methodology 13
14 1.3 Thesis Outline This thesis is organized in six chapters, which are presented in the same chronological order as the different research stages of the project. Chapter 2 summarizes the current technology on wireless LANs and introduces general terminology. Chapter 3 analyzes the indoor channel environment and proposes simple models based on parameters obtained from public literature. Chapter 4 develops the transceiver architecture and analyzes the design tradeoffs involved in choosing specific parameter values. Chapter 5 introduces several network protocols to support multiuser communication. Chapter 6 details how the software simulation was put together and discusses the result of simulating the network performance under different channel conditions. Chapter 7 concludes the thesis by summarizing achievements and presenting leads for further research. 14
15 Chapter 2 : WLAN Concepts & Standards The wireless LAN market has grown rapidly in the last five years . Today, WLANs are used in hospitals, to record patient information at bedside; in car rental companies, to input car-return information; in warehouses and retail shops, to keep inventories; and in restaurants, to place orders. Student volunteers participating in the MIT China Educational Technology Initiative (MIT-CETI) program have also made use of wireless networking products to bridge Chinese high schools to the Internet service provided by nearby universities. Although the first experimental WLANs were conceived as replacement for wires, most WLANs nowadays are employed to extend rather than replace existing wired networks. Indeed, wireless LANs are ubiquitous among industries that require mobile computing resources or that face extreme difficulties in deploying new physical media. In terms of advantages, wireless networks are easy to install since they require no additional cables, reduce long-term costs in dynamic organizations that are characterized by frequent moves and changes, and increase productivity in environments that require real-time mobile access to information. 2.1 WLAN Technologies Wireless LANs use electromagnetic airwaves (radio or light) to transmit information from one point to another without relying on any physical connection. Normally, the data being transmitted is superimposed or modulated onto a specific radio wave or carrier, which then delivers electromagnetic energy to a remote receiver. Once data is modulated onto a carrier, the resulting radio signal occupies more than a single frequency because the frequency or bit rate of the modulating information adds to that of the carrier .
16 Most commercial WLANs are based on radio or infrared technologies . Radio technology is normally subdivided into narrowband and spread spectrum. Each technology comes with its own set of advantages and limitations, as outlined in the next few sections Narrowband Technology In a narrowband network, the available radio spectrum is divided into frequency channels so that different network users may transmit and receive data on specific frequencies. The bandwidth of each channel is kept as narrow as possible, and is usually determined by the bit rate of the data being sent. At the receiver end, any particular user then filters out all the radio signals except those on its designated frequency. Narrowband is a bandwidth efficient technology because it packs all the information content in the minimum possible bandwidth. Unfortunately, such efficiency also makes narrowband technology very susceptible to errors due to interference on any radio channel. For this reason, customers employing narrowband technologies must normally obtain an FCC license to guarantee an interference-free environment on their frequency channels Spread Spectrum Technology An alternative technology is to spread the data bits across the entire available radio spectrum according to user-specific parameters. In this case, the transmitted signal is less susceptible to narrowband interference since it uses more bandwidth than its actual information content. In fact, the transmitted signal resembles background noise to all users but the intended recipient. Frequency hopping and direct sequence are the most common spreading techniques. narrowbands Power spread Figure 3. Narrowband divides the spectrum into subchannels while spread spectrum extends the entire band. Frequency 16
17 Frequency-Hopping Spread Spectrum Technology Frequency-hopping spread spectrum (FHSS) systems spread the signal energy over a wide band by jumping among narrowband carriers at specific time intervals. When the jumps follow a pattern known to both transmitter and receiver, the system can be synchronized to maintain a single logical channel, otherwise the signal resembles short-duration impulse noise. FHSS minimizes interference by limiting the time spent at each carrier, thus lowering the probability that any two transmitters will use the same frequency at any point in time. Carriers f 5 f 4 f 3 f 2 f 1 t1 t2 t3 t4 t5 t6 t7 t8 t9 t10 Time Figure 4. FHSS systems hop among frequency carriers at predetermined intervals of time. Direct-Sequence Spread Spectrum Technology In direct-sequence spread spectrum (DSSS) systems, the signal energy is spread across a wide band by replacing each data bit with multiple sub-bits or chips that occupy the same time interval as the original data bit. The bandwidth ratio between chips and data bits is called processing gain. The higher the processing gain, the greater the probability that the data can be recovered even if several chips are corrupted during transmission. To an unintended receiver, DSSS signals resemble low-power wideband noise that may actually fall below thermal noise. Chip sequence: Original data stream: Spread data stream: Figure 5. Consider a 6-bit chip sequence. To spread a digital data stream, each 1 in the stream is substituted with the specified chip sequence, while each 0 is substituted with the negated sequence. 17
18 2.1.3 Infrared Technology Some wireless networks are also based on infrared (IR) radiation, typically between 800 and 900 nanometers, just below visible light in the electromagnetic spectrum. Since infrared light does not penetrate opaque objects, IR links are either direct, when highly focused beams transmit signals over the shortest path between sender and receiver; or diffused, when reflective surfaces are used to flood a room with infrared energy. Due to their limited range (about 3 feet) and point-to-point nature, infrared systems are less suitable for multiuser networks and so they will not be taken into account in this thesis Efficiency Considerations Naturally, network designers would rather use the most efficient technology, if only they could agree on a single definition for efficiency. Some designers maximize bandwidth efficiency, or the number of bits-per-second transmitted per unit of spectrum bandwidth (bps/hz). Other designers optimize power efficiency, or the amount of power dissipated at different amplifier stages. Yet other designers maximize multiuser efficiency, or the number of multiple users that can communicate simultaneously through the network. The different interpretations for efficiency often suggest mutually exclusive design tradeoffs. For instance, networks that are very bandwidth efficient tend to use large signal constellations with multiple amplitude levels, even though their implementation requires linear power amplifiers that are inherently very power inefficient. The disagreement on how to measure network efficiency is epitomized by the contending cellular network technologies. FDMA cellular networks make use of bandwidth efficient narrowband technology, even though their constraints on frequency reuse limit multiuser efficiency. On the other hand, CDMA cellular networks make use of multiuser efficient spread spectrum technology, even though their actual bandwidth efficiency is much lower than 1 bps/hz. In order to avoid such controversial standpoints, this thesis will focus first on maximizing bandwidth efficiency, and then on optimizing multiuser efficiency. Power efficiency concerns are postponed until the actual implementation study of the WGLAN. 18
19 2.2 WLAN Configurations Regardless of their underlying technology, wireless LANs always assume one of two basic configurations, ad hoc or infrastructure. In an ad hoc WLAN, two computers equipped with wireless adapter cards can set up a peer-to-peer connection whenever they are within range of one another. Such an improvised configuration requires no central controller but only gives access to the resources available within the two networked computers. In infrastructure networks, computers equipped with wireless adapters can communicate with each other, as well as with a wired LAN through access points scattered throughout a building. Since the access points are connected to the wired network, each wireless client can also have access to server resources. Some networks allow mobile clients to roam, that is, to move seamlessly among a cluster of access points. In those networks, clients are handed off from one access point to another in a way that is invisible to the client, ensuring unbroken connectivity during the same communication session. The WGLAN project constitutes an infrastructure network. In our case, the wireless adapters used by mobile clients are called network adapters, while the access points connected to the wired LAN are called network controllers. Since the WGLAN is intended for indoor use within the home or office, each setup consists of a single network controller, which eliminates the need to include roaming capabilities into the entire system. 2.3 Radio Spectrum Regulations In the United States, the Federal Communications Commission (FCC) regulates all radio emissions within the frequencies of 3 khz to 300 GHz, which includes most WLAN technologies except those based on infrared light. According to FCC regulations, manufacturers of radio-based WLANs employing narrowband technology must obtain a license to use specific radio frequencies at every site where a wireless network is to be deployed. In general, licenses increase costs and reduce system flexibility since each wireless network must be custom designed to make use of frequencies that are unique to each site. 19
20 The commercial development of radio-based WLANs started in 1985 when the FCC allowed the license-free use of a set of frequencies called the Industrial, Scientific, and Medical (ISM) bands  listed in Table 1. Nowadays, most WLANs make use of the ISM bands centered at 915 MHz, 2.45 GHz, and 5.8 GHz since they are the lowest frequency allocations that gather several megahertz of spectrum. Higher operating frequencies generally have a shorter propagation range and incur higher manufacturing costs, which explains, for instance, why the 915 MHz band is the most crowded ISM band. The other two bands offer additional benefits of their own: the 2.45 GHz ISM band is the only unlicensed allocation acceptable worldwide, and the 5.8 GHz band has a total bandwidth that spans 150 MHz of spectrum. Table 1. Frequency bands allocated for use by ISM equipment. FCC Code Part Low Bandwidth Bands Popular Bands High Bandwidth Bands 6.78 MHz ± 15.0 khz 915 MHz ± 13.0 MHz GHz ± MHz MHz ± 7.0 khz GHz ± 50.0 MHz GHz ± MHz MHz ± khz 5.8 GHz ± 75.0 MHz GHz ± MHz MHz ± 20.0 khz GHz ± 1.0 GHz In order to support multiuser access to ISM bands, Part 15 of the FCC code stipulates that operation on those bands be limited to frequency hopping and direct sequence spread spectrum radiators with maximum peak output of 1 watt. Section also specifies that frequency hopping systems in the 2.4 and 5.8 GHz bands shall use at least 75 hopping frequencies, with hopping channels not wider than 1 MHz, and average time of occupancy on any frequency not greater than 0.4 seconds within a 30 second period. For direct sequence systems, the minimum 6 db bandwidth shall be at least 500 khz, with no maximum specified. Current FHSS-based wireless networks using ISM bands achieve peak data transmission rates of up to 2.0 Mbps. Such data rates are insufficiently low for some high-speed applications, yet little can be done to increase these rates given that the hopping channels of FHSS systems can be at most 1 MHz wide, with a typical spectral efficiency currently below 2 bps/hz. In 1997, the FCC recognized that the technical restrictions on ISM bands hindered the development of high-speed WLANs, and decided to open up an additional part of the spectrum, which named the Unlicensed National Information Infrastructure (U-NII) band. 20
21 According to the FCC, U-NII devices will provide short-range, high-speed wireless digital communications on an unlicensed basis to facilitate access to the National Information Infrastructure, or Next Generation Internet . Following this vision, the FCC provided the U- NII band with 300 MHz of non-contiguous spectrum located at GHz and GHz. Use of this band is regulated by Part 15-E of the FCC code, which allows significant flexibility in the design of U-NII devices by adopting the minimum technical rules needed to prevent interference to other services and to ensure efficient use of the spectrum. Table 2. Maximum power restrictions of the U-NII band. Frequency Band (in GHz) Peak Transmit Power Peak Power Spectral Density mw 2.5 mw/mhz mw 12.5 mw/mhz mw 50 mw/mhz As opposed to the rigid regulations on ISM bands, devices operating in the U-NII band must simply comply with the radiation power restrictions summarized in Table 2. Note that the GHz section of the U-NII band overlaps with part of the ISM bands centered at 5.8 GHz. Even though both bands are restricted to 1 watt of radiation power, the U-NII 5.8 GHz band is not limited to spread spectrum technologies. All factors considered, the U-NII band provides the bandwidth and design freedom needed to support the objectives of the WGLAN project. Before proceeding to model the indoor channel characteristics of the U-NII band, we first review the three major wireless LAN standards currently in existence that make use of ISM frequencies. 2.4 IEEE WLAN Standard In the early days of wireless networking, WLAN products implemented by different manufacturers were based on proprietary technology that was mutually incompatible. As a consequence, the first WLAN customers had no alternative but to assume the high costs of depending on a single manufacturer to provide all of their networking needs. The Institute of Electrical and Electronic Engineers (IEEE) recognized the need for a WLAN standard in 1992, 21
22 and by June of 1997 it had approved a radio (and infrared) standard for the 2.4-GHz ISM band to provide interoperability among wireless networks from different manufacturers. The Standard, officially named the IEEE Standard for Wireless LAN Medium Access (MAC) and Physical Layer (PHY) Specifications, defines the protocols needed to provide wireless connectivity of fixed, portable, and mobile stations moving at pedestrian and vehicular speed within a local area. Specific features of the standard include: Data rates of 1 or 2 Mbps, using FHSS, DSSS, or infrared modulation. Carrier-sense multiple access with collision avoidance (CSMA/CA) Data fragmentation to support asynchronous and time-bounded delivery service. Error control at the frame level and acknowledgement of each packet received. Rules for power management, authentication, and addressing. Although gives specifications for both the MAC and PHY layers, the following overview will focus on the radio-based physical layers that are of direct interest to this thesis Network Topologies The IEEE Standard supports two types of topologies: Independent Basic Service Set (IBSS) networks and Extended Service Set (ESS) networks. Both networks utilize a basic building block called the BSS, which provides a coverage area whereby stations of the BSS remain fully connected. A station is free to move within the BSS, but it can no longer communicate directly with other stations if it leaves the area of coverage of the BSS. An IBSS network is a stand-alone BSS that has no backbone infrastructure and consists of at least two wireless stations. This type of network is also known as an ad hoc network and satisfies most needs of users occupying a small, improvised area. On the other hand, an ESS network is connected to a wired infrastructure that may include additional BSS networks. 22
23 Station A BSS 1 access point Wired LAN access point Station B BSS 2 IBSS ESS Figure 6. IBSS versus ESS networks. The standard assumes three types of mobility. In no-transition, stations are static or move within a local BSS. In BSS-transition, stations move between different BSS within the same ESS. And in ESS-transition, stations move between BSS that belong to different ESS. The standard supports the no-transition and BSS-transition mobility types but cannot guarantee continued connectivity during an ESS-transition Physical Layer Architecture The physical layer architecture of each compliant station is divided into the Physical Layer Convergence Procedure (PLCP) sublayer and the Physical Medium Dependent (PMD) sublayer. The PLCP sublayer minimizes the dependence of the MAC Layer with the wireless medium by appending fields to the MAC protocol data units (MPDUs) that contain information needed by the PMD sublayer transmitters and receivers. This composite frame is known as a PLCP protocol data unit (PPDU). Meanwhile, the PMD sublayer interfaces directly with the wireless medium and implements carrier sense algorithms as well as modulation and demodulation of frames during transmission and reception modes, respectively. MAC Layer PHY PLCP Sublayer PMD Sublayer Figure 7. The standard specifies the MAC and PHY layer. The PHY layer, in turn, consists of the PLCP and PMD sublayers. 23
24 2.4.3 FHSS Physical Layer Frequency-hopping is one of two radio-based spread spectrum physical layers proposed in the wireless LAN standard. The FHSS physical layer has the lowest power consumption, lowest potential data rates from individual physical layers, highest aggregate capacity using multiple physical layers, and less range than direct-sequence . FHSS Frame Format A single FHSS PLCP frame consists of a preamble, to enable the receiver to synchronize its clocking functions; a header, to provide information about the frame; and a whitened PSDU, or PLCP Service Data Unit, which is the MPDU the station sends. 80 bits 16 bits 12 bits 4 bits 16 bits Variable size SYNC Start frame delimiter PLW PSF Header error check Whitened PSDU PLCP preamble PLCP header PLCP service data unit Figure 8. PLCP frame corresponding to an FHSS physical layer implementation. The PLW, or PSDU Length Word, field specifies the length of the PSDU (0-4,095 octets). The PSF, or PLCP Signaling Field, specifies the data rate of the whitened PSDU portion of the frame. The Header Error Check contains the result of applying the CRC-16 error detection algorithm to the PLW and PSF fields. The physical layer does not determine whether errors are present within the PSDU, that is a function of the MAC Layer. CRC-16 detects all single- and double-bit errors and ensures detection of % of all possible errors. Finally, the Whitened PSDU minimizes the dc bias by stuffing special symbols after every four octets of the data signal. The PSDU whitening process uses a length-127 frame synchronous scrambler and a 32/33 bias-suppression encoding algorithm to randomize the data. 24
25 Frequency Hopping Functions The standard defines a set of hopping channels that are 1 MHz wide and are evenly spaced across the 2.4 GHz ISM band. The number of channels varies geographically, North America and most of Europe has 79, while Japan has 23 for instance. The available hopping frequencies are segregated into three distinct hopping sets. The FHSS-based PMD sublayer transmits PPDUs by hopping from channel-to-channel according to a particular pseudo-random sequence that distributes the data signal uniformly across the frequencies in a specific hopping set. After a hopping sequence is selected in an access point, all stations automatically synchronize to the correct hopping sequence. Frequency Modulation Functions The FHSS PMD sublayer transmits symbols using two-level or four-level Gaussian frequency shift keying (GFSK) modulation depending on whether the data rate is 1 or 2 Mbps, respectively. GFSK is a variant of frequency shift keying where the signal spectrum is shaped by a Gaussian filter characterized by its baseband bandwidth B and transmission rate 1/T. When transmitting data streams at 1 Mbps, the input to the GFSK modulator is either a 0 or 1 coming from the PLCP sublayer. The modulator transmits the binary data by shifting the center operating frequency F C for each hop by a slight deviation f d, where f d must be greater than 110 khz. Consequently, the symbol 1 is encoded with a transmit frequency of (F C + f d ), while the symbol 0 is encoded with a transmit frequency of (F C f d ). When transmitting data streams at 2 Mbps, the input to the GFSK modulator is combinations of 2 bits (00, 01, 10 or 11) coming from the PLCP sublayer. Each of these 2-bit symbols is sent at 1 Msymbol/s, meaning that the bit rate is 2 Mbps. Consequently, the symbol 10 is encoded with a transmit frequency of (F C + 3f d ), symbol 11 is encoded with (F C + f d ), symbol 01 is encoded (F C f d ), and symbol 00 is encoded (F C 3f d ). The nominal value for f d in both two-level and four-level GFSK modulation is 160 khz. 25
26 2.4.4 DSSS Physical Layer Direct-sequence is the second radio-based spread spectrum physical layer proposed in the wireless LAN standard. The DSSS physical layer has the highest power consumption, highest potential data rates from individual physical layers, lowest aggregate capacity using multiple physical layers, and more range than frequency hopping . DSSS Frame Format The format of a DSSS PLCP frame consists of a preamble, to enable the receiver to synchronize its clocking functions; a header, to provide information about the frame; and a whitened PSDU, or PLCP Service Data Unit, which is the MPDU the station sends. 128 bits 16 bits 8 bits 8 bits 16 bits 8 bits Variable size SYNC Start frame delimiter Signal Service Length Frame check sequence MPDU PLCP preamble PLCP header PLCP service data unit Figure 9. PLCP frame corresponding to a DSSS physical layer implementation. The Signal field specifies the type of modulation that the receiver must use to demodulate the signal. The Service field is reserved for future use. The Length field is an unsigned 16-bit integer indicating the number of microseconds to transmit the MPDU. The receiver uses this information to determine the end of the frame. The Frame Check Sequence field contains the result of applying the CRC-16 error detection algorithm to the PLW and PSF fields, similar to the FHSS physical layer. Finally, the PSDU field is the same as the MPDU being sent by the MAC Layer, which can range from 0 bits to a maximum size set by the PMD sublayer. 26
27 DSSS Spreading Sequence The DSSS physical layer digitally spreads the original PPDU before modulating it onto one of the 14 available frequencies as specified in the standard. Different modulating frequencies are provided to allow the concurrent operation of multiple users. The original standard specifies an 11-chip Barker sequence as the spreading code for data payloads of 1 Mbps and 2 Mbps. Higher data rates are supported by the b extension, which specifies an 8-chip complementary code keying (CCK) for payloads of 5.5 Mbps and 11 Mbps. The modulation rate depends on the data payload, and it is set to 1, 2, 4, and 8 chips/symbol respectively so that the chipping rate is always equal to 11 MHz. Frequency Modulation Functions The spread PLCP frame is modulated by shifting the phase of the transmit carrier frequency. For the regular data rates as specified in the original standard, if the initial frame had a 1 Mbps data rate, then DSSS PMD sublayer transmits the spread frame using differential binary phase shift keying (DBPSK) modulation. Otherwise, the 2 Mbps data rate uses differential quadrature phase shift keying (DQPSK) modulation. In phase shift keying, the phase of the carrier frequency is changed to represent different data symbols. When transmitting data streams at 1 Mbps, the input to the modulator is either a 0 or 1 coming from the PLCP sublayer. The modulator transmits this data by shifting the phase of the carrier frequency in increments of 180 degrees. Consequently, the symbol 1 is encoded with a phase shift of 180 degrees, while the symbol 0 is encoded with a phase shift of 0 degrees. When transmitting data streams at 2 Mbps, the input to the modulator is combinations of 2 bits (00, 01, 10 or 11) coming from the PLCP sublayer. Each of these 2-bit symbols is sent at 1 Msymbol/s, meaning that the bit rate is 2 Mbps. Consequently, the symbol 00 is encoded with a phase shift of 0 degrees, symbol 01 is encoded with 90 degrees, symbol 11 is encoded with 180 degrees, and symbol 10 is encoded with 270 or negative 90 degrees. 27
28 2.5 HIPERLAN Standard In 1998, the European Telecommunications Standards Institute (ETSI) adopted its High Performance Radio Local Area Network (HIPERLAN) standard. This standard defines the Physical and Medium Access Control layers for a high data rate wireless network operating in the GHz frequency band. Systems that are HIPERLAN-compliant can be deployed in an ad-hoc or pre-arranged fashion, where nodes can move as fast as 360 degrees per second or 1.4 meters per second. HIPERLAN systems also provide coverage beyond the radio range limitation of a single node, support asynchronous and time-bound communication, and attempt to conserve power by arranging the times when mobile clients need to be active for signal reception Channel Access The nominal frequency band of HIPERLAN is the GHz band, which supports three channels, each MHz wide. This is a pan-european frequency allocation on a secondary, non-interference basis. Within Europe, the GHz band can also be allocated on demand to provide two additional channels. In order to increase multiuser capacity, HIPERLAN equipment is required to operate on all five channels. The channel access protocol used in HIPERLAN is by means of a listen before talk scheme that is termed Elimination-Yield Non-pre-emptive Priority Multiple Access (EY- NPMA). A node alternately transmits a short burst and listens to the channel to determine if another node is transmitting. If a node can hear another transmission then it yields to that node and tries to gain access to the channel when it becomes free again. Five different levels of traffic priority are supported by allowing a node with higher priority to start contending for the channel before nodes with lower priority. The access protocol also includes a contention resolution mechanism that copes with the possibility of multiple nodes trying to access the channel at the exact same time. 28
29 2.5.2 Physical Layer The tasks of the physical layer include the following: (1) Modulating and demodulating radio carriers to create RF links. (2) Acquiring bit and burst synchronization. (3) Transmitting or receiving a defined number of bits at a requested time and on a particular carrier frequency. (4) Encoding and decoding the Forward Error Correction scheme. (5) Deciding whether a channel is idle or busy, for the purposes of deferral during channel access attempts. HIPERLAN transmits information using data bursts. There are two types of data bursts: LBR data bursts encode low bit-rate data streams, while LBR-HBR data bursts encode high bitrate data streams preceded by a LBR segment. The two admissible signaling rates are Mbps for the low bit rate, and Mbps for the high bit rate. The maximum permissible transmit power is 1 watt, giving a maximum operating range of about 50 meters. User data is sent in blocks of 416 bits (52 octets). This data is encoded for error correction/detection purposes using a (31,26) BCH code to produce blocks of 496 bits (62 octets). The encoded data is then interleaved over one of these blocks to randomize bit errors. The maximum packet length in HIPERLAN is 47 blocks of 62 octets (around bits), which lasts for about 1ms. This is the maximum time over which the indoor radio channel is assumed to be reasonably static. In addition, each packet is protected with a CRC field. The modulation technique employed depends on the data bit rate. Gaussian Minimum Shift Keying (GMSK) with BT=0.3 is used to modulate high bit rate transmissions, while Frequency Shift Keying (FSK) is used to modulate low bit rate transmissions. GMSK is a special form of binary continuous-phase FSK based on the minimum frequency separation that is necessary to ensure the orthogonality of the modulation signals from the response of a Gaussian filter to a rectangular pulse interval of length T . Implementation details about this modulation technique can be found in the HIPERLAN standard . In FSK, signaling is accomplished by adding small deviations to the center carrier frequency. Thus, the symbol 0 is encoded with the frequency (F C f d ), while the symbol 1 is encoded with the frequency (F C + f d ), where f d is set to 368 khz. 29
30 2.6 Bluetooth Specification The Bluetooth Specification was released in December of 1999 as a de facto standard created by a Special Interest Group (SIG) that includes several telecommunication industry leaders. Bluetooth provides a short-range radio link, within a so-called Personal Area Network (PAN), intended to replace the cable(s) connecting portable and/or fixed electronic devices . Due to its limited 10-meter range, Bluetooth is not considered a wireless LAN by any means, yet an overview of its technical features may prove useful for future reference. Bluetooth-enabled devices can establish point-to-point and point-to-multipoint connections by forming an ad hoc network or piconet. Each piconet consists of one Bluetooth unit that acts as master, and several units that act as slaves. All devices in the same piconet are synchronized to the master s clock so that full-duplex transmission may be accomplished through a Time Division Duplex (TDD) scheme. Multiple piconets with overlapping coverage areas form a scatternet. Slave units can participate in different piconets on a Time-Division Multiplex basis, and master units in one piconet can act as slaves in another piconet. At the physical layer, Bluetooth radios communicate using frequency-hopping spread spectrum technology at the globally available 2.4 GHz ISM band. The frequency hopping channels are 1 MHz wide and the gross data rate is 1 Mbps. The modulation used is GFSK with BT = 0.5, where a binary one corresponds to a positive frequency deviation and a binary zero to a negative frequency deviation. The minimum deviation shall exceed 115 khz. In contrast to the and HIPERLAN standard, the Bluetooth Specification explains how to manage cryptic keys to provide for authentication and user privacy. The specification also includes instructions on how to interface to multiple communication protocols such IrDA, USB, RS232, and UART. Indeed, the versatility and technical feasibility of Bluetooth played a decisive role for this specification to become a de facto standard. 30
31 Chapter 3 : Channel Analysis The communications channel is the physical medium through which data-bearing signals propagate. Some examples include the twisted-wire-pair telephone line, coaxial cable, fiber optic lines, and the wireless radio frequency spectrum. Since energy dissipates differently in various media, the communications channel must be characterized to determine such network parameters as power requirements, maximum range, signaling rate, and data capacity. In this chapter, we first review the mathematical concepts used to characterize indoor radio channels of the type used by the WGLAN. Next, we refer to the public literature to extract key parameters that characterize the U-NII frequency bands. Finally, we use the collected information to propose several channel models and a network link budget. 3.1 Characterizing Indoor Radio Channels The indoor radio channel is an adverse communication channel where the transmitted signal arrives at the receiver via multiple propagation paths that differ in amplitude, phase, and delay time. Due to the existence of multipaths, the information signals that propagate through indoor radio channels are distorted by both time dispersion and amplitude fading . The phenomenon of time dispersion is illustrated in Figure 10a. Here, even though the transmitter sends an extremely short pulse, ideally an impulse, the channel response due to multiple scatterers causes the received signal to be spread in time. If, in addition, the channel experiences physical changes over time, then the response of any signal transmitted through it will change with time as well, hence the designation of time-varying channel .
32 Signal Transmitted Signal Received Frequency Spectrum of Received Signal power T 1 T 2 T 3 time time time (a) Figure 10. (a) Time dispersion and (b) amplitude fading on multipath channels power frequency (b) The phenomenon of amplitude fading is illustrated in Figure 10b. Since multiple propagation paths have different phase offsets, two or more multipath components may add destructively at times, causing the received signal to vanish, or fade, in certain locations. Fading is equivalent to a notch in the channel s magnitude frequency response. Mathematically, the baseband multipath channel impulse response is modeled as  Eq. 1 N jθk h( t) = β e δ ( t τ ) k = 1 k k Where for N total paths, k is the path index, β k is the real-positive path gain, θ k is the phase shift, and τ k is the time delay of the kth path. δ( ) is the Dirac delta function. Originally, the time delay of each propagation path is lower bounded by the speed of electromagnetic waves, i.e m/s. However, since the absolute delay of the channel is irrelevant, the first arriving path is taken as a time reference by setting τ 1 = 0. Because of the motion of people and equipment in and around the indoor environment, the parameters β k, θ k, and τ k are time-varying stochastic processes. Such variations in the structure of the multipath channel cause the received signals to be unpredictable to the user of the channel, and so we must characterize the time-variant multipath channel statistically. However, later in this chapter we will show that since the transmitter and receiver have a fixed position, the channel parameters change very slowly in comparison with the signaling rate, and thus the communications channel may be assumed to be quasi-static . 32
33 3.1.1 Power Delay Profile The power delay profile (PDP) , also called the multipath intensity profile or the delay power spectrum , is equivalent to the autocorrelation of the channel impulse response. The PDP P(t) gives the time distribution of the received signal power from a transmitted δ- pulse, and is defined as follows Eq. 2 * 2 P( t) h( t) h ( t) = h( t) = β δ ( t τ ) 2 N k = 1 k k In practical measurements, the transmitted pulses s(t) have finite width. When the pulse width of s(t) is less than the delay time differences between the paths, the PDP is given by Eq P( t) = s( t) h( t) = β s( t τ ) N k = 1 k k Path Loss Signals propagating through free space experience a power loss that is proportional to the distance, d, from the source, and inversely proportional to the wavelength, λ, of the carrier wave. If the carrier wave is centered at a frequency f, then its wavelength is given by the relationship λ = c / f, where c = m/s. Hence, in free space the power loss L s is given by Eq. 4 L s 2 4 π d = λ In reality, however, the signal travels through multiple paths not in free space. Hence, path loss measurements in indoor environments require a different model. The following simple model is frequently used to describe path loss (in db)  Eq. 5 Loss L L L + 10 n log10 ( d / d ) = o + d = s d 1m o o = Where n is the path loss exponent of the environment and d (in meters) is the distance between the transmitter and receiver, which together define the additional path loss component L d. Meanwhile, L o is the free space loss (in db) of a path of d o meters (often d o = 1m). 33