Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Application of interference cancellation to third generation partnership project wireless systems Olawale, Kassim Olabode 2002

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

Item Metadata

Download

Media
831-ubc_2003-0061.pdf [ 3.01MB ]
Metadata
JSON: 831-1.0065594.json
JSON-LD: 831-1.0065594-ld.json
RDF/XML (Pretty): 831-1.0065594-rdf.xml
RDF/JSON: 831-1.0065594-rdf.json
Turtle: 831-1.0065594-turtle.txt
N-Triples: 831-1.0065594-rdf-ntriples.txt
Original Record: 831-1.0065594-source.json
Full Text
831-1.0065594-fulltext.txt
Citation
831-1.0065594.ris

Full Text

APPLICATION OF INTERFERENCE C A N C E L L A T I O N T O THIRD G E N E R A T I O N PARTNERSHIP PROJECT WIRELESS SYSTEMS by KASSIM O L A B O D E O L A W A L E B.Sc, Obafemi Awolowo University, 1992 A THESIS SUBMITTED IN PARTIAL F U L F I L M E N T OF T H E REQUIREMENTS FOR T H E D E G R E E OF M A S T E R OF APPLIED SCIENCE in T H E F A C U L T Y OF G R A D U A T E STUDIES (Department of Electrical and Computer Engineering) We accept this thesis as conforming to the required standard T H E UNIVERSITY OF BRITISH C O L U M B I A November 2002 © Kassim Olabode Olawale, 2002 In p r e s e n t i n g t h i s t h e s i s i n p a r t i a l f u l f i l m e n t of the requirements f o r an advanced degree at the U n i v e r s i t y of B r i t i s h Columbia, I agree that the L i b r a r y s h a l l make i t f r e e l y a v a i l a b l e f o r reference and study. I f u r t h e r agree that permission f o r extensive copying of t h i s t h e s i s f o r s c h o l a r l y purposes may be granted by the head of my department or by h i s or her r e p r e s e n t a t i v e s . I t i s understood that copying or p u b l i c a t i o n of t h i s t h e s i s f o r f i n a n c i a l gain s h a l l not be allowed without my w r i t t e n permission. Department of £Uctnca l Commute*- £vY<jw\ee.irwCj The U n i v e r s i t y of B r i t i s h Columbia Vancouver, Canada Date blecemWr 3 . 0 1 , 3 0 0 1 Abstract • % ^ j * . i . ' Direct Sequence Code Division Multiple Access (DS-CDMA) is one of the technologies available for present day as well as future wireless systems. A major limitation on the use of C D M A is Multiple Access Interference (MAI). This is caused by imperfect separation of users communicating within the same physical frequency and time domain. Eliminating M A I enhances detection of the desired user's signal in a D S - C D M A system. Multiuser detection considers the interference caused by other users in the same communication system while detecting the* desired user. Our work involves critical study and evaluation of some of the many multiuser detection algorithms for Third Generation Partnership Project (3GPP) wireless systems. We selected and compared performance of the Parallel Interference Cancellation (PIC), Successive Interference Cancellation (SIC) and Group-wise Serial Interference Cancellation (GSIC) methods in the uplink of a 3 GPP system. To evaluate the expected performance improvement, computer simulations were carried out using MATLAB® software. A U M T S (Universal Mobile Telecommunications System) base station with a conventional receiver was compared to another one employing PIC, SIC and GSIC receivers. Test parameters were set to values provided in the 3 GPP Standard. Power control was not implemented in our study. Results of these simulations demonstrate lower bit error rates in those receivers employing interference cancellation in most cases. We observe that the three receivers are equally suitable at low SIR, low data rates and small number of users. None of the three schemes is suitable for systems using high data rates (raw bit rate of 480 kbps or higher) or high number of users (more than 25 users per sector). The PIC receiver is better than the SIC and GSIC at about 0 dB SIR. With these observations, we conclude that any of these three schemes can be deployed to a 3 GPP system. Channel conditions under the 3GPP test scenarios are harsh and may give worst-case performance results with interference cancellation, especially without use of power control. In some environments such as wireless LAN's , low number of users and moderate data rates would prevail, in which case interference cancellation would improve performance significantly. n Table of Contents Abstract ii Table of Contents iii List of Figures vi List of Tables viii Acknowledgement ix Chapter 1 Overview and Summary 1 1.1 Introduction 1 1.2 Transition to Third Generation Systems 2 1.3 The Need for Multiuser Detection 2 1.4 Application of Multiuser Detection to Third Generation Systems 3 1.5 Summary of Thesis 3 Chapter 2 Review of Literature 5 2.1 Third Generation Personal Communication Systems 5 2.1.1 System Architecture 5 2.1.2 Physical Layer Basics 6 2.1.3 Physical Layer Services 7 2.1.4 Physical Layer Functions 8 2.1.5 System Capacity 11 2.2 Multiuser Detection 12 2.2.1 Conventional Detector 13 2.2.2 Detection in the presence of Multiple Access Interference (MAI) 15 2.2.3 Inter-Symbol Interference versus Multiple Access Interference 17 2.3 Possible Receiver Structures 18 2.3.1 Optimum Receiver 18 2.3.2 Successive Interference Cancellation 18 2.3.3 Parallel Interference Cancellation 20 2.3.4 Group-wise Serial Interference Cancellation 21 2.3.5 Interference Cancellation on the Uplink 22 2.4 Multiuser Detection and 3GPP Systems: System Capacity 22 Chapter 3 Channel Models 25 3.1 Model 1: Synchronous BPSK in A W G N 25 iii 3.2 Model 2: Asynchronous BPSK in A W G N 26 3.3 Model 3: Single-user in Multipath Propagation and A W G N 27 3.4 Model 4: Multiple Asynchronous Users in Multipath Propagation and A W G N 28 3.5 QPSK Signals 28 Chapter 4 Simulation Process 30 4.1 The Communication System 30 4.1.1 Generating Data 30 4.1.2 Spreading 31 4.1.3 Scrambling 32 4.1.4 Modulation 32 4.1.5 Channel 33 4.2 Conventional Receiver 35 4.3 Parallel Interference Cancellation 36 4.3.1 Regeneration 36 4.3.2 Interference Cancellation 37 4.3.3 Multiple Cancellation Stages 37 4.3.4 PIC Receiver Outputs 38 4.4 Group-wise Serial Interference Cancellation 38 4.5 Successive Interference Cancellation 39 4.6 Calculation of Signal to Interference Ratio (SIR) 39 4.6.1 Obtaining multiple SIR points 40 Chapter 5 Results 41 5.1 Parallel Interference Cancellation 41 5.1.1 Different Channel Situations 41 5.1.2 Number of Users in the Communication System 43 5.1.3 Scrambling Code Generation Method 45 5.1.4 Different Spreading Factors 46 5.1.5 Multiple Interference Cancellation Stages 48 5.1.6 Comparison with SIC and GSIC 49 5.2 Successive Interference Cancellation 50 5.2.1 Different Channel Situations 50 5.2.2 Number of Users in the Communication System 51 5.3 Group-wise Serial Interference Cancellation 52 5.3.1 Different Channel Situations 52 5.3.2 Number of Users in the Communication System 53 Chapter 6 Conclusions 54 iv 6.1 Similarities between the different receivers 54 6.1.1 Data rates 54 6.1.2 Number of users 55 6.2 Differences between the different receivers 55 6.3 Practical Applications 55 6.4 Areas of further studies 56 6.4.1 Effect of Power Control 56 References 58 Appendix A Abbreviations 62 Appendix B Simulation Software Test Tools 64 v List of Figures Fig. 2.1: U M T S Architecture (simplified) 6 Fig. 2.2: Frame structure for uplink DPDCH/DPCCH [8] 8 Fig. 2.3: Spreading of uplink dedicated channels [9] 9 Fig. 2.4: OVSF code tree. Top left shows the tree construction principle [3, 9] 10 Fig. 2.5: Modulation 10 Fig. 2.6: The Conventional C D M A Detector (the dotted box labeled as 'Bank of Matched Filters') 13 Fig. 2.7: The Decorrelating Detector (compare with Fig. 2.6) 16 Fig. 2.8: Successive Interference Cancellation 19 Fig. 2.9: Parallel Interference Cancellation 20 Fig. 2.10: Group-wise Serial Interference Cancellation 21 Fig. 3.1: Bit streams in an asynchronous channel 26 Fig. 3.2: Bit streams in a channel with multipath propagations 27 Fig. 4.1: Simulation block diagram 30 Fig. 4.2: Spreading of uplink dedicated channels with only one D P D C H 31 Fig. 4.3: Multi-path fading model 33 Fig. 4.4: Rake receiver 35 Fig. 4.5: Regeneration of received message signals 36 Fig. 4.6: Parallel Interference Cancellation 37 Fig. 4.7: Multiple Parallel Interference Cancellation Stages 37 Fig. 5.1: Performance in different channel situations (with 2 users) 42 Fig. 5.2: Performance in different channel situations (with 4 users) 43 Fig. 5.3: Performance as number of users change 44 Fig. 5.4: Performance improvement at BER of 10"4 for different number of users 44 Fig. 5.5: Performance difference due to scrambling code generation method (with SF = 64) 45 Fig. 5.6: Performance difference due to scrambling code generation method (with SF = 16) 46 Fig. 5.7: Performance at different spreading factors 47 Fig. 5.8: Performance improvement at BER of 10"4 for different data rates 47 Fig. 5.9: Effect of multiple interference cancellation stages 48 Fig. 5.10: Performance of the different interference cancellation schemes 49 Fig. 5.11: Performance in different channel situations 50 Fig. 5.12: Performance as number of users change 51 Fig. 5.13: Performance in different channel situations 52 Fig. 5.14: Performance as number of users change 53 Fig. B . l : Level crossing rate and cumulative distribution results 65 vi Fig. B.2: Sample processes for a channel with 3 paths 66 Fig. B.3: Sample processes for a channel with 4 paths 67 Fig. B.4: Comparison of a four-user system with theoretical results (SF = 64) 68 Fig. B.5: Comparison of a four-user system with theoretical results (SF = 8) 69 Fig. B.6: Comparison of a two-user system in a three-path Rayleigh fading channel with theoretical results 70 vii Lis.t^bf Tables^ Table 2.1: Basic Parameters of the Physical Layer 6 Table 2.2: Estimated capacity of a 3GPP system (chip rate: 4.096 Mcps) 12 Table 4.1: Channelization codes (single DPDCH) 31 Table 4.2: Channelization codes (multiple DPDCH) 32 Table 4.3: Propagation Conditions for Multipath Fading Environments 34 Table 5.1: Spreading factors (SF) versus data rate (chip rate: 3.84 * 106) 46 viii Acknowledgement I would like to thank my supervisors, Dr. Eduardo Casas and Dr. Robert Donaldson for their immeasurable guidance and support in the course of this work. I owe the successful completion of the thesis to them both. This work was supported by research grants from PMC-Sierra, Inc., Cadence Design Systems, Inc., and by the Canadian Natural Sciences and Engineering Research Council (NSERC). I appreciate the financial support received from them in the course of my research. I also want to thank my wife for her patience and support through this work. Finally, I am grateful to my parents and my brothers and sisters who provided the building foundations for me to get thus far. ix Chapter 1 Overview and Summary 1.1 Introduction Demand for wireless Personal Communications Systems (PCS) has continued to grow worldwide. This is partly attributable to the wide variety of functionalities and services available to subscribers in existing systems. In general, similar quality of service provided to users on Public Switched Telephone Networks (PSTN) is now available to PCS users. That means PCS subscribers receive regular telephone service while in motion, in addition to other enhancements. Wireless networks have some basic disadvantages not present in wireline systems. One of these is the limited radio frequency spectrum through which users of the wireless network communicate. More users on a given network imply less spectrum resources per user or more interference, depending on the multiple access scheme. For some types of access schemes (TDMA or FDMA) , a given network can support only a fixed maximum number of users. Therefore, call-blocking results after the concurrent user capacity is exhausted. For a particular type of scheme though (CDMA), soft capacity is possible. Rather than block new calls, the quality of individual users' call decreases as the number of users increase. In practice, call-blocking is still used to keep quality of each call at a desired level. Interference in the radio spectrum is not caused by other users in the same wireless network alone. Radio propagation effects also introduce distortion and delay in transmitted signals. Signals from users in adjacent and co-located networks, which may or may not belong to the same wireless service providers, also constitute interference. There are numerous other natural and man-made sources of interference. The use of multiple access schemes mentioned above helps to significantly reduce interference. In addition, various regional regulatory bodies set and enforce conditions for use of the radio spectrum. These, along with other advances in engineering, make concurrent communication by numerous subscribers possible. The promulgation of standards by different regional regulatory bodies causes some difficulties. Most of the standards are incompatible across regions. Different countries and continents use different technologies such as the multiple access schemes (TDMA, F D M A , CDMA) , multiplexing schemes (FDD, TDD), transmitting frequencies, modulation (BPSK, QPSK, G M S K , etc.), bandwidth, and so on. Therefore, equipment produced for use in some countries cannot be used in some others. These differences 1 generally limit range of usage by PCS subscribers to market regions implementing the same standard. Alternatively, the subscribers require different equipment to access services in different regions. 1.2 Transition to Third Generation Systems The first generation of wireless communication standards was mainly analog. These were implemented in different regions of the world as the need arose. Digital communication was introduced in the second generation of standards. Presently however, significant differences exist between the standards in use in various regions of the world. The major regulatory bodies across the world are collaborating to produce a globally uniform standard. Standards so developed will be deployed in countries of the respective participating regulatory bodies. These constitute the Third Generation (3G) Standards for personal communication systems. The collaboration responsible for producing the standards is called the Third Generation Partnership Project (3 GPP). Technical Specifications for 3GPP wireless systems have been published. The specification for the physical layer include such details as the multiple access scheme: Wideband C D M A , duplexing: F D D and T D D , modulation: QPSK, operating frequency, and a host of others (Forward Error Control coding, Scrambling, power control, etc.). 1.3 The Need for Multiuser Detection The radio interface for 3GPP PCS is based on Wideband Direct Sequence Code Division Multiple Access (Wideband DS-CDMA). In this multiple access scheme, signals from users in the communication system overlap in frequency and time. Different users employ unique (and almost) orthogonal pseudorandom (or pseudonoise) (PN) sequences to separate signals. At the receiver, the desired signal is detected by correlating the received signal with the appropriate PN sequence. The D S - C D M A principle is based on use of orthogonal PN codes. Essentially, if the PN codes used are orthogonal and each radio transmission propagates over a single path, the cross-correlation between the desired users PN code and the other users' signals will be zero. However, because transmission is asynchronous and signals travel over multiple paths, cross-correlation is not always zero. The sum of the non-zero cross-correlations constitutes interference to the desired signal. 2 A conventional receiver for D S - C D M A treats the residual signals from the cross-correlation as noise. This will be effective in situations where the aggregate cross-correlation is very low, compared with the autocorrelation of the desired signal with its PN code, but not otherwise. Multiuser detection is a technique to reduce the effect on the desired user from other simultaneously active users in a given wireless network coverage region. Various techniques have been developed for different circumstances. The general idea is to treat interference caused by other users in D S - C D M A as Multiple Access Interference (MAI) rather than as noise. By removing the interference, the quality of the communication system is enhanced directly. 1.4 Application of Multiuser Detection to Third Generation Systems Researchers worldwide have published results from various detailed works on multiuser detection in D S - C D M A communication systems. Usually, studies are based on mathematical models, computer simulations or both. In order to limit the complexity of the analysis, simplifying assumptions are made. The results of such exercise, though valid, may not be immediately applicable. In the thesis work described herein, a real life problem is used to define a practical solution. In other words, the multiuser detectors discussed here can be deployed in a PCS environment that meets 3GPP technical specifications. This is similar to what would be done by companies designing a 3GPP-compliant system. To do this, the multiuser detector must take into consideration, many practical implementation issues. Specifically, it must give adequate consideration to multipath propagation, channel attenuation and delay. To ensure the research is completed in good time, our work made use of the extensive publications available on multiuser detection. Proposed solutions were computer simulated and compared in performance to the conventional receiver (without multiuser detection). 1.5 Summary of Thesis The rest of this thesis report is laid out as follows. A summary review of 3GPP PCS, multiuser detection and the application of multiuser detection to 3GPP systems are given in Chapter 2. The 3GPP air interface model is described in Chapter 3. Tools and processes used for computer simulations of the model are presented in Chapter 4, along with the simulation 3 parameters and assumptions. Chapter 5 contains simulation results. Conclusions and further areas of research are presented in Chapter 6. Abbreviations used in this document are listed in Appendix A. Test tools used to measure the quality of the simulation software can be found in Appendix B. 4 C h a p t e r 2 Rev iew of Literature In this chapter, the basic background on which the work is built is presented. The required parts of the third generation wireless personal communication system are presented first. This is followed by a general discussion on multiuser detection. 2.1 Third Generation Personal Communication Systems Third Generation (3G) wireless personal communication systems (PCS) evolved from existing standards. Several regional communication regulation bodies submitted proposals towards the formulation of the new global standard [1, 2]. Dahlman et. al. gave a detailed description of W - C D M A in [3]. This article is a very useful introduction to 3GPP systems, since the physical layer of the 3GPP specification did not change much from the W - C D M A Standard. The book by Holma and Toskala [4] covers the 3GPP Standard in detail. In the rest of this section, a discussion of the technical specification published by the Third Generation Partnership Project (3GPP) will be presented. The area of interest is the Physical Layer (Layer 1). However, a general overview of the architecture will be discussed first. 2.1.1 System Architecture The Third Generation system is designed such that most communication equipment (telephone, fax, computer or laptop, etc.) can be connected to the network seamlessly. It is planned to be adopted widely throughout the world. The communication system is called a Universal Mobile Telecommunication System (UMTS). Fig. 2.1 shows the basic elements of the U M T S architecture. The structure consists of the following [5, 6]: User Equipment The User Equipment (UE) domain represents the facility used by the subscriber to access services provided by the network. It consists of the applications selected by the user. It also provides radio connection of these applications to the network. UMTS Terrestrial Radio Access Network The U M T S Terrestrial Radio Access Network (UTRAN) provides facilities required for the user equipment to access the core network. It would normally 5 consist of numerous base stations and services related to these. The radio interface (Uu) between user equipment and U T R A N is the reference point for services between them. ! i 1 ! 1 1 I I < UE 1 Radio UTRAN I Core Networ (Uu) i U Fig. 2.1: U M T S Architecture (simplified) Core Network The Core Network (CN) provides the basic telecommunications services of the system. Functions covered here normally include management of user location information, control of network features and services, the transfer (switching and transmission) mechanisms for signaling, and other similar services. The radio interface (between the user equipment and UTRAN) is the area of interest in this work. 2.1.2 Physical Layer Basics The Physical Layer (PHY), also known as Layer 1, is the lowest layer of the OSI Reference Model. It provides data transport services to higher layers. These services include channel coding and decoding, interleaving, multiplexing and demultiplexing, rate matching, modulation and demodulation, spreading and despreading, power control and radio transmissions. The multiple access scheme employed in U M T S is Wideband Direct Sequence Code Division Multiple Access (Wideband DS-CDMA). It uses one of two duplex modes; Frequency Division Duplexing (FDD) and Time Division Duplexing (TDD). In this work, only the FDD mode is discussed. Table 2.1 below shows the basic parameters of the Physical layer in U M T S [8]. Table 2.1: Basic Parameters of the Physical Layer Description Parameter Multiple Access Wideband D S - C D M A Duplex modes FDD, T D D Channel Bandwidth 5 MHz or multiples of. Chip rate 3.84 Mcps Multi-rate concept Orthogonal Variable Spreading Factor (OVSF) Frame Length 10 ms or multiples of. Coherent detection Pilot symbols transmitted Spreading Modulation QPSK Data Modulation QPSK Roll-off factor 0.22 Handover Soft and interfrequency handover 6 A summary of the services and functions offered at the Physical layer (for FDD) is provided below. 2.1.3 Physical Layer Services The Physical layer receives data (in the form of Transport Channels: TrCH) from the Medium Access Control (MAC) layer and maps them into Physical channels [7, 8, 9]. Physical channels can be broadly categorised into two: Dedicated Channels and Common Channels. Both categories exist on the downlink (from U T R A N to user equipment) and on the uplink (from user equipment to UTRAN) . Dedicated channels in the uplink are discussed in the following subsections. The interference cancellation schemes studied in this work are best suited to the uplink (please see section 2.3.5). Common Channels are used mainly for synchronisation, random access and transmission of general control signals. They are not discussed in this document. 2.1.3.1 Dedicated Channels There are two types of Dedicated Channels (DCH): Dedicated Physical Data Channels (DPDCH) and Dedicated Physical Control Channels (DPCCH). Data channels carry payload signals. Control channels carry Physical layer control information such as pilot and power control symbols. On the uplink1 (UL), D P D C H are multiplexed with D P C C H using variable spreading, as well as, in-phase and quadrature (I/Q) modulations. The spreading process is further discussed later. 2.1.3.2 Frame Structure The structure of the radio frame for the uplink dedicated channel (DCH) is shown in Fig. 2.2 below. It consists of the data channel (DPDCH) and the control channel (DPCCH). Each radio frame is 10 ms in length. It is sub-divided into 15 slots of 2560 chips each. Thus, there are 3.84 x 104 chips per 10 ms frame. More than one transport channel could be transmitted at a time. How the transport channels are combined is specified in the optional Transport Format Control Indicator (TFCI). This is controlled from higher protocol layers. Blind detection of transport format combination is used when TFCI is omitted. Pilot symbols are used for coherent detection. In this thesis, only one transport channel was considered. 1 On the downlink (DL), they are time-multiplexed. 7 Therefore, further consideration of TFCI or blind detection of transport format combination will not be necessary. DPDCH DPCCH Data Ndata bits Tslot — 2560 chips • Pilot Transport Format Control Indicator Feedback Information Transmit Power Control ^— Tslot — 2560 chips , 10 bits Slot #0 Slot #1 Slot #i Slot #14 •4 1 radio frame: T f = 10 ms Fig. 2.2: Frame structure for uplink D P D C H / D P C C H [9] 2.1.4 Physical Layer Functions In transmitting the various physical channels, the transport channel (from the M A C layer) is processed through a number of functions [7, 8]. These operations are reversed in the physical layer of the receiver to recover the transmitted data. The major functions of interest in this thesis are spreading, scrambling and modulation. These are presented below (from section 2.1.4.1 onwards). Other Physical layer basic functions include: • Cyclic Redundancy Coding (CRC) • Channel Coding: convolutional coding, turbo coding or no coding. • Interleaving • Radio Frame Segmentation • Rate Matching • Transport Channel Multiplexing • Physical Channel Segmentation • Physical Channel Mapping 8 These functions are not discussed further. They all occur after interference cancellation has been concluded. Therefore, they are not relevant to the comparison of different interference cancellation schemes. 2.1.4.1 Spreading and Scrambling Fig. 2.3 shows how spreading and scrambling are implemented for uplink2 dedicated channels (DPDCH and DPCCH). One D P C C H is always required for every dedicated channel radio frame. Zero to six D P D C H could be multiplexed with one DPCCH. The physical channels are separated from each other using channelization codes Q for DPDCH and cc for DPCCH. Different physical channels are spread using different spreading factors (SF). Cd,l DPDCH, cd,3 DPDCH, Cd,5 DPDCH, DPDCH, DPDCH, DPDCH6 Cd,2 © cd,4 O Cd,6 O DPCCH Pd i + J Q , Fig. 2.3: Spreading of uplink dedicated channels [10] Spreading applied to uplink and downlink channels are slightly different. 9 Channelization codes are Orthogonal Variable Spreading Factor (OVSF) codes generated using the code tree of Fig. 2.4. (c, c) (c, -c) C,.„ = (l) C4.0=(l,1,1,1) c2.„=(U) c4.,=(i, 1,-1,-1) C4.2=(l,-l,l,-l) c2,,=(i,-i) c4.,=(i,-i,-i,i) Fig. 2.4: OVSF code tree. Top left shows the tree construction principle [3, 10]. pd and pc are gain factors used to weigh all the DPDCH and D P C C H respectively. Possible values of both PJ and fic are 0, 1/15, 2/15, 14/15, 1. Either of them should be 1 at any particular time. All D P D C H use the same value of pd chosen. The branches labeled I and Q are summed to form a complex signal, I + jQ. This complex signal is then scrambled with S<|c|, (as shown in Fig. 2.3 above). Scrambling codes are a segment of Gold sequences combined into complex codes of length 2 2 5 - 1 (long codes) or 255 (short codes). 2.1.4.2 Modulation Scrambled data is modulated for transmission using QPSK as shown in Fig. 2.5. cos(cot) I Complex-valued chip sequence from spreading operations s Split real & imag. parts Re{S} Pulse-shaping * l lm{S} Pulse-shaping -sin(cot) Fig. 2.5: Modulation 2.1.4.3 Power Control Closed loop power control is performed at the physical layer using Transmit Power Control (TPC). Higher protocol layers exercise open loop power control. Power control was not used in this work. 10 Closed loop power control consists of two parts: inner loop power control [4, 7] and outer loop power control [4, 6]. Inner loop power control is exercised at the Physical layer. This generally involves ordering the transmitter to increase or decrease transmit power in order to achieve a target signal to interference ratio (SIR) at the receiver. The Radio Resource Control (RRC) layer (Layer 3) controls setting of the target SIR used in inner loop power control. This constitutes the outer loop power control. The target SIR is such that the received signal meets required quality, usually bit error rate (BER). The required quality in turn depends on information data rate and quality of service (QoS) expected. Target SIR is also affected if the user equipment is in a soft handover position. If a communication system with TV concurrent users employs closed loop power control, the target SIR is always specified for each of the N users at the respective R R C layer. Thus, the target SIR will vary for each user. It will also change for a given user if circumstances change significantly during the duration of a connection (such as if the user moved into a handover state in the course of the call). 2.1.5 System Capacity Interference cancellation increases system capacity for a given performance level. Thus, a discussion of the number of concurrently active users that a 3GPP compliant communication system can support will help in assessing application of the interference cancellation schemes. Wang et. al. studied capacity estimation for a communication system that is based on the 3 GPP Standard [11]. The authors considered both voice only and voice plus data systems. They did not implement any multiuser detection scheme in the receiver. They found that the maximum capacity of a sector is about 100 users for a voice only system. The bandwidth used in their simulation was 5 M H z and the raw Physical layer data rate of voice users' signals was 32 kbps (SF of 128 on a chip rate of 4.096 Mcps). As would be expected, the capacity drops as each high-speed data user joins the system. The maximum capacity for high-speed data users is 5, which is obtained when there is no voice user in the system. The data users were transmitting at 256 kbps. Table 2.2 below is produced as a linear extension of these figures to other data rates. The capacity figures are provided per sector. A cell with three sectors will therefore have three times the figures shown. 11 Table 2.2: Estimated capacity of a 3GPP system (chip rate: 4.096 Mcps) SF Raw Data Number of Rate (kbps) Users per Sector 128 32 ~ 100 64 64 - 2 0 - 5 0 16 256 ~5 8 512 - 2 - 3 A voice activity factor of 3/8 was applied to all voice signals in the aforementioned simulation. Therefore, the system is likely to have a much lower capacity than just the ratio of data rates if users have higher data rates (of 64 kbps for instance). The capacity is therefore represented as a range in that case. 2.2 Multiuser Detection In Direct Sequence Code Division Multiple Access (DS-CDMA), all users share the same channel frequency and time slot. The signal to be transmitted is spread with a pseudo-random (pseudo-noise: PN) sequence. The PN sequence is unique for each user. The receiver despreads the message sent by correlating the received signal with the PN sequence for the user. In general, the conditions under which the receiver will successfully despread the message are: • the PN sequences used by all the users in the communication system are completely orthogonal • there is no multipath propagation • transmissions are synchronous • signal to noise ratio is not too low In practice however, the wireless medium naturally consists of multipath propagation. This introduces inter-symbol interference (ISI) into the transmitted message. In addition, due to asynchronous transmission, it is practically difficult to receive completely orthogonal PN sequences. Therefore, the cross-correlation between the PN sequences for different users (at the receiver) are not always zero. The sum of these cross-correlation terms produces multiple access interference (MAI). The conventional receiver, which is usually the matched filter receiver, considers M A I as Gaussian noise. Thus, any information from the desired user that is contained in the MAI is lost. No attempt is made to recover this information. Effectively, MAI raises the noise floor. As the number of users in the communication system increases, so does the MAI. The presence of relatively stronger 12 individual users in the communication system also further increases MAI. As M A I (and thus, the noise floor) increases, the quality of the decoded signal gets worse. In the limit, some established calls will be dropped and new calls cannot be established. This is a limitation on the capacity of the system. Multiuser Detection is the use of signal processing techniques to recover desired information from MAI. Several algorithms have been proposed in literature. They all exploit various known properties of signals in a C D M A system. A brief exposition of multiuser detection is given next to ease into the discussion of the thesis. It starts with an introduction to the conventional receiver. A mathematical model of the communication system is presented. The major multiuser detection algorithms are then introduced. 2.2.1 Conventional Detector The conventional C D M A detector used in today's receivers employs matched filters. That is, the message for each user is retrieved from the received signal by passing it (the received signal) through a correlator matched to the unique PN sequence for that user. Consider a channel with a total of K active users as shown in Fig. 2.6 [12, 13]. The received signal, r(t) consists of the modulated and encoded message signals, xk(t) for all the K users. For each message bit, this can be expressed as: User 1 *,(0 Channel 1 Transmitter x2(t) User 2 Channel 2 Transmitter **« User K Channe l K Transmitter Channel T, dt s2(t) 1 (Tb t =iTh ^ 0 Detect ion Devices Bank o f Matched Filters Fig. 2.6: The Conventional C D M A Detector (the dotted box labeled as 'Bank of Matched Filters'). 13 KO = Yixk (0 + n(t) with xk (t) = Akbksk (0 A: where Ak, bk, sk(t) represent the amplitude, modulation signal and the PN sequence waveform respectively for the M i user, and n(t) is the additive white Gaussian noise. For simplicity, the modulation scheme discussed below is BPSK. So, bk = ± 1 . Extension to QPSK (used in 3GPP) is discussed subsequently. In this analysis, the channel is considered to be additive white Gaussian noise only. No other channel degradations are featured. In the matched filter detector, r(t) is correlated with the respective PN sequence waveform of the desired user. Over a bit of duration Tb, this gives: OT yk = S T~ f A J b J s j + i ~ f "CK ^ d t E i - 2 - 1 j=i *b +b Define (normalized) correlation coefficient as: p • k = — f s;. (t)sk (t)dt j,k = 1,.. ., K Tb Pjk = 1 for all j = k (since sk(t) = ± 1). The above equation can be reduced to [14]: K yk = A A + Z PM A j b j + ZK E c i - 2 2 7=1 yk = A A +MAik+zk where MAIk = ] T P j kAjbj 7=1 and z t The desired message is contained in Akbk. By design, pJJ( should be zero when j * k. This occurs when the PN codes are completely orthogonal. In such cases, MAIk (multiple access interference) will be 14 zero. In practice however, the codes used are not completely orthogonal. Therefore, pJtk always falls between 0 and 1. 2.2.2 Detection in the presence of Multiple A c c e s s Interference (MAI) Several efforts have been made to enhance the performance of the conventional detector [14, 15]. These are directed towards better preparation of the signals at source to limit the magnitude of M A I generated. Examples include the design of codes with very low cross-correlation (pjik « 0); use of power control (to reduce the magnitude of MAI) and the use of Forward Error Correction coding (FEC). A lot of work has also been done in the area of multiuser detection, since Verdu proposed the optimal multiuser detector [16]. This proposal employs the maximum-likelihood sequence (MLS) detector. The major drawback is the complexity of its implementation. It requires data matrix-vector of the order of 2NK, where K is the total number of active users and /V is the average number of message bits sent per user. For practical values of K and N, this receiver is impracticable. This has led researchers to seek sub-optimum detectors, which perform better than the conventional detector while reducing the level of implementation complexity required. Two classes of sub-optimum detectors have been investigated [14]. These are: • Linear Detectors • Subtractive Interference Cancellation Linear detectors are used to estimate and remove the correlation coefficients pJJc from the output of the conventional detector, yk. Examples include Decorrelating Detector and the Linear Minimum Mean-Square Error detector. Detectors grouped under the Subtractive Interference Cancellation class usually estimate the multiple access-interference (MAI) in the decoded signal yk, and subtract it from the received signal r{i). The new r\t) is then detected using the conventional detector, hopefully without the effect of MAI. Iterative processes could be employed to obtain even better results. Examples of detectors in this class are: Successive Interference Cancellation (SIC), Parallel Interference Cancellation (PIC) and Group-wise Serial Interference Cancellation. These receivers are discussed later in this chapter. 15 There are dozens of other variations of these detectors which have been proposed and simulated by researchers. They all differ in performance, complexity, practicality and disadvantages. 2.2.2.1 Decorrelating Detector In the decorrelating detector, matched filters are used to separate various users' signals. The output of the bank of matched filters is multiplied by the inverse of cross-correlation matrix as shown in Fig. 2.7. Earlier, Fig. 2.6 shows that the output of the conventional detector is a matrix, y = [y,, y2,...,y^'• This could be translated from Eqs. 2.1 and 2.2 above as [15, 12]: Channel r(t) Receiver Bank o f Correlators (Matched Filter) yt yK Mat r i x F i l ter /f"1 Dec i s i on yi Dec i s i on • • yt Dec i s i on • Dec i s i on Fig. 2.7: The Decorrelating Detector (compare with Fig. 2.6) y = RAb + z Eq. 2.3 where R is a K * K matrix of cross-correlation coefficients p]k, j, k = 1, . . . , K, A is a K * K diagonal matrix of Ak,k, k = 1,. . . , K and b is a K-vector of bk, k = 1 , . . . , K. The Decorrelating Detector is obtained by placing a filter that implements the inverse of matrix R (that is: R"1) immediately after the conventional detector (Fig. 2.7). This will result in the following (making use of Eq. 2.3): y' = R"'y = Ab + R"'z Eq. 2.4 y' consists of the desired message vector, Ab and an enhanced noise vector, R"'z. The decorrelating detector is flawed mainly in its enhancement of noise. As the inverse of cross-correlation matrix is non-zero, and the output of the matched filter still contains additive noise, the multiplication to remove MAI also increases noise. In addition, the cross-correlation matrix, to be inverted in real time, is a K by K matrix for a synchronous system with K active users. When the effect of multipath propagations is added, this dimension grows considerably. Consequently, implementation of the detector 16 becomes more complex for practical values of K. One solution to the complexity is the use of cascaded parallel interference cancellation. This is discussed later in section 2.3.3. 2.2.2.2 Linear Minimum Mean Square Error The Linear Minimum Mean Square Error (MMSE) is similar to the decorrelating detector [15]. It de-correlates output of the conventional detector. Before inverting correlation matrix R, it is modified by adding inverse of the signal to noise ratio: RMMSE = R + A 2 (No/2) Thus, multiple access interference is removed and noise enhancement (normally found in the decorrelating detector) is avoided. The basic advantage of the Linear M M S E is that it out-performs the decorrelating detector in additive white Gaussian noise (AWGN). In very little or no noise, the Linear M M S E becomes similar to the decorrelating detector. It also adapts to changing channel characteristics such as when users set-up new calls or terminate existing ones. Signal strength of all received users is required to calculate the modified cross-correlation matrix. Acquiring this, and the propagation delays constitute the major disadvantages of the Linear M M S E . 2.2.3 Inter-Symbol Interference versus Multiple A c c e s s Interference Multipath propagation also causes inter-symbol interference (ISI). This is due to the arrival at the receiver of several versions of the same message waveform, each with a time delay. The waveforms for the respective message bits overlap. ISI is usually modeled along similar lines as demonstrated above for MAI. In other words, assuming the PN sequence used are completely orthogonal, the output of the detector for the M i user (in the presence of multipath fading) will be [13]: yk = Akbk + ISIk + zk Equalizers have been built to mitigate the effect of ISI. With the similarity in the model for ISI and MAI, researchers have found that detectors designed for MAI are very similar to equalizers designed for elimination of ISI. Klein et. al. [17] presented a few equalizers that can remove both. 17 2.3 Possible Receiver Structures The optimum receiver is the Maximum Likelihood Sequence detector. This is considered too complex for practical implementation. Therefore, a number of sub-optimum multiuser detection algorithms have been proposed in published literature. The focus of these is to improve performance over the matched filter receiver, while reducing complexity. A review of the algorithms considered in this work is done here. 2.3.1 Optimum Receiver MAI occurs when the signals reaching the receiver are no longer orthogonal. The cross-correlation between desired user's signal and signals of other users in the channel is no longer zero. In such a communication system, the optimum receiver is the Maximum Likelihood Sequence (MLS) detector [12, 16]. At the receiver, a matched filter is employed for every user transmitting in the system. This bank of matched filters precedes Viterbi decoding algorithm. To detect the signal for the desired user, the receiver searches all possible combinations of the arriving sequence before selecting the one with the shortest Euclidean distance. That is the most likely sequence in a channel with additive white Gaussian noise, given that the message bits transmitted are equiprobable. This receiver structure does not treat (unintentional) interference from other users in the C D M A channel as noise. Rather, it de-spreads all such signals using the bank of matched filters. It therefore exploits information available in the channel to enhance detection of desired user. The major setback of the M L S E is the complexity of implementation. To search through a sequence of message with length N bits sent by K users for instance is potentially a search through 2NK possibilities. For practical values of message lengths and number of users, this is rather too large. The promise of higher capacity brought by the M L S E , which is hindered by its complexity, has led to research work in sub-optimum detectors. The goal is to design receivers with (nearly) as good performance as the M L S E but with much lower complexity. 2.3.2 Success ive Interference Cancellation The decorrelating detector and the Linear M M S E are linear multiuser detectors. Another common group of multiuser detectors is those that do interference cancellation. In general, they subtract multiple access interference (MAI) from the received signal to obtain the desired signal. 18 In Successive Interference Cancellation (SIC) receivers [14, 15, 18], received signals are ranked in descending order of signal power. The signal with the strongest power (say for User 1) is sent through the conventional (matched filter) detector and the message sent is decoded. The decoded message is regenerated and subtracted from the original received signal, r(t), which is delayed to allow time for processes required before subtraction. The residual signal, r\t) does not contain signal for User 1, and consequently, it does not contain M A I due to User 1. The receiver selects the next strong signal and repeats the process as for User 1. This is shown in Fig. 2.8. The process continues until all the users have been detected. Alternatively, the process could continue until sufficient number of powerful signals have been detected and cancelled. The residual signals can then be detected using conventional detector as usual. r(t) Matched Filter r J Matched Filter r J Matched Filter y>SS) =F Fig. 2.8: Successive Interference Cancellation SIC introduces lengthy delays in detection. The user with least received signal power can only be detected after all stronger users have been detected. In addition, this system relies on the normal imbalance in received signal power. The first signal detected (for User 1) is done using the conventional detector in the presence of MAI. If this detection is wrong, then the subtracted signal is also wrong. Rather than canceling interference, it would be increased. SIC actually performs worse than the conventional receiver when the received signal power is balanced. 19 3GPP systems employ closed power control. Received power is likely to be almost equal for all users. Therefore, the SIC will only be a suitable receiver for 3GPP systems, when power control is not in use, or when power control has not successfully closed up the received signal strength of the various users. 2.3.3 Parallel Interference Cancellation Parallel Interference Cancellation (PIC) is similar to SIC in principle [14]. It subtracts (cancels out) interference from the received signal before detecting the desired users. The difference in PIC as compared to SIC is that cancellation of interference is done at once (in parallel). Fig. 2.9 shows the PIC detector. The received signal, r(t) is sent through a first stage of detection. This is usually the conventional detector or a linear multiuser detector like the decorrelating detector. Output of this first stage is regenerated to obtain the received signal, ?,(/) for each user. Thereafter, interference signal is summed up and subtracted from r(t). r(t) is delayed to allow time for processes required before subtraction of interference. The remaining signal is assumed to be free of multiple access interference. Final detection can therefore be done with the conventional detector. Ordering of signals in received signal power is no longer required. Delay in detection is also greatly reduced. However, knowledge of channel attenuation and delay is required to enable the system correctly estimate the signals to subtract. r(t) Regenerate Bank of Correlators (Bank of Matched Filters) ?i.(0 A. A ' i A A r2 ML j*2 2 o J*K Bank of Correlators (Bank of Matched Filters) y'jt) i f 3 Sum up Interference Fig. 2.9: Parallel Interference Cancellation The Parallel Interference Cancellation receiver can be cascaded into multiple iterative stages [19]. This iteration results in convergence ultimately to a decorrelating detector if the system load is not greater than 17% [20]. Convergence can be improved by using graduated weights to limit the interference 20 cancelled out at each stage [21]. The weight applied increases as the confidence in estimated interference improves in the iteration process. 2.3.4 Group-wise Serial Interference Cancellation Successive Interference Cancellation (SIC) can be made much faster. First group received signals by some factor. A n example is the grouping together of signals with the same spreading factor (SF) in a multirate system. Each group is then detected as a single stage in the SIC process. That is, all the signals in the first group are detected, regenerated and subtracted from the received signal (at once). Then, the next group is selected and the process repeated. Fig. 2.10 depicts this process. Detection of each group can use any of the multiuser detection schemes already discussed above: decorrelating detector, linear M M S E or PIC. This scheme is called Group-wise Serial Interference Cancellation (GSIC) [22, 23, 24]. Use of SIC on each group in the GSIC gives similar performance with the SIC. K0 Group signals wi th some cri ter ia: by received signal strength or by bit rate rg(t) Detect Group 1 B i t streams for users in Group 1 Regenerate Group 1 Detect Group 2 Bi t streams for users in Group 2 Reg enerate Group 2 + Detect Last Group B i t streams for users in Last G roup Fig. 2.10: Group-wise Serial Interference Cancellation GSIC finds a balance between the SIC and the scheme used internally on the groups. For instance, GSIC with internal decorrelator reduces dimension of the matrix to be inverted in the decorrelator, reduces 21 delay suffered due to successive cancellation and reduces chances of incorrect decision at the first stage of successive cancellation. 2.3.5 Interference Cancellation on the Uplink The uplink of the communication system was the focus of this thesis. The uplink is usually the capacity limiting link, compared to the downlink. In addition, base stations have the capacity to support the computational intensity required for interference cancellation. The base station will normally receive and decode signals from all users in the cell. To accomplish that, the base station will have information about all users, such as scrambling and spreading codes; and it will track and acquire other related information, such as signal delay and phase. The same information is required for regeneration of received signals in interference cancellation schemes. Base stations also have access to continuous power supply sources to support computational demand. The user equipment, especially if it is a mobile phone, is usually small, operates on a battery and tracks only information related to a single user (normally). So, it has lower capacity for intensive computations. Thus, the uplink is better suited to implementation of interference cancellation. 2.4 Multiuser Detection and 3GPP Systems: System Capacity For a given level of performance, multiuser detection can be used to increase capacity of a C D M A system relative to the conventional receiver. Most published technical documents do not provide an indication of the increase in capacity expected. Results obtained or demonstrated through simulation often use a manageable number of users. Published papers that give analysis of this issue are reviewed in this section. The focus was limited to papers related to this thesis: 3GPP (UMTS) systems (or close to it) and interference cancellation. Not all the papers discussed relate directly to 3GPP systems. Hottinen et. al. discussed use of multiuser detection to increase capacity of a C D M A communication system [25]. The authors concluded that at SNR = 10 dB, the system can accommodate about 30 users with BER of 1 x 10"3. This is obtained after two stages of interference cancellation and application of a 1/2 rate convolutional decoding. Simulations carried out were not entirely based on 3GPP systems. For example, variable data rate considerations were omitted. BPSK was used for spreading and modulation. The bandwidth was 2MHz. The spreading codes used to separate users were Gold codes. The 22 paper does however establish that performance improvement of the multiuser receiver over the conventional receiver degrades as more users are added to the communication system. The authors continued this work by testing the same multiuser receiver in the same communication system that now enables multiple data rates [26]. They concluded that performance improvement of the multiuser detection receiver is better when there are few users with high data rates compared with a large number of low data rate users. They did not give an opinion on what happens if the system had a constant number of users with the same data rate which was varied from low to high. The authors further considered the effect of the multiuser detection receiver at the system level, using the results they obtained from the link level [27]. They observed that, at 5% outage probability, the capacity of the system is about 4.5 users/MHz/Cell. This almost doubles the capacity of the conventional receiver, which is 2.3 users/MHz/Cell. 65% interference cancellation was performed. The data rate was 74.3 kbps and the required B E R was 1 x IO"3. It should be noted that, from this analysis, the capacity of the system (employing the multiuser detection receiver) for a cell operating with 5 M H z bandwidth is 22.5 users. This includes users in a soft handover state. In addition, the authors cautioned that increasing the number of users will significantly degrade the system and the 5% outage probability will quickly become unrealistic. Besides, other interference not accounted for in the cancellation scheme will add to system degradation. In a paper by Mozaffaripour and Tafazolli, although the authors did not intend to determine cell capacity, their results were obtained only for very low data rates and the number of users in the system was set to 20 [28]. The simulations done used U M T S parameters published in 3GPP Standard documents. Similarly, the contributions by Del Re et. al. featured simulations that used no more than 17 users, out of which 16 transmitted at low (raw physical) data rates of 60 kbps [29]. The authors specified a new interference cancellation scheme. They did not intend to demonstrate capacity of the system. Ammar et. al. discussed the capacity that could be obtained for a U M T S system in [30]. The paper shows performance based on the number of users in the system. The best results showed that a U M T S communication system with 30 users at a spreading factor of 32 for each user would be a B E R of 8 x 10"2 when desired user has an 8 dB SNR. That is after multiple interference cancellation stages. The 23 paper demonstrates that the performance of the SIC receiver reduces as the number of users in the system increased. This is not mitigated by the multiple cancellation stages. Pirinen and Glisic also discussed maximum capacity of a W C D M A communication system with consideration for numerous factors [31]. The authors concluded that, for a spreading factor of 256, the conventional receiver will yield a maximum capacity of about 50 users. Capacity improvement due to multiuser detection more than doubles that capacity to 120. In their simulations, multiple data rates were not considered and BPSK modulation was used. With due considerations to the difference in data rate (or spreading factor), this is similar to the capacity indicated by Amar et. al. (see preceding paragraph). 24 Chapter 3 Channel Models A simple channel model was used to explain multiuser detection in Chapter 2. This is to simplify the concept and make it easy to follow. In practice, the channel is subjected to various degradations that have to be considered. A few of such degradations are shown in this chapter. Mathematical models commonly used in literature are reviewed. It is easier to start with the description of the ideal channel, free of fading and multipath, and introduce more realistic models gradually. While the following discussions used a BPSK signal, the models apply equally to a QPSK signal (cf. section 3.5). 3.1 Model 1: Synchronous BPSK in AWGN Assume a communication system with the followings: K synchronous users (transmitting at the same time) Binary Antipodal Modulation (Binary Phase Shift Keying) in Additive White Gaussian Noise without multipath propagation without fading This is the system assumed in the previous chapter. The receiver will observe the following signal (cf. Fig. 2.6). K KO = I > * (0 + n{t) with xk (t) = Akbksk (t) k=\ K r(t) = ^ Akbksk(t) + n(t) k=\ In general, multiuser detection is not required in this system. If the PN codes used in transmission are orthogonal, they will remain orthogonal in the received signal r(t). That is, the cross-correlation pjtk between the P N codes of any two users j and k will always be zero. Therefore, there will be no multiple access interference (MAI). The conventional receiver is optimum in these conditions. 25 3.2 Model 2: Asynchronous BPSK in A WGN If the communication system of model 1 consists of asynchronous users, then system conditions will be as follows: K asynchronous users Binary Antipodal Modulation (Binary Phase Shift Keying) in Additive White Gaussian Noise without multipath propagation without fading The received signal consists of overlapping bits due to the uncoordinated signaling from users. An example of the delay pattern is shown in Fig. 3.1. Let xk represent the offset (or delay) of the users relative to each other. Then, the spreading code and message bits for the M i user and /th bit are, respectively, sk(t - iTb - xk) and bk[i\ Hence, the received signal for each user, xk(t), is given by (cf. with the expression for xk(i) above): xk = ^Akbk[i]sk(t-iTb -rk) where 2M+ 1 is the total number of message bits. K M ^ = Z Z AKbK (' - tTb ~ ^ ) + k=] i=-M foxK users. Bit streams for User 1 (&,) User 2 (b2) User 3 (63) bit /'=-1 -T, bit i=0 bit i=\ Z Z L 71 2TU 3Th b *b l 3 b -"b T,=0 Fig. 3.1: Bit streams in an asynchronous channel time, t 26 3.3 Model 3: Single-user in Multipath Propagation and AWGN Consider a communication system with a single user. Due to multiple reflection of the user's signal, L propagation paths exist between the transmitter and receiver. System conditions can be summarized as: Single user Binary Antipodal Modulation (Binary Phase Shift Keying) in Additive White Gaussian Noise L propagation paths The channel is similar to Model 2, except that K is replaced with L. That is, the receiver sees L users transmitting with the same spreading factor. The /th transmitted message (after spreading) is: b[i] s(t-iTb) and the received message from each path is: b[i] s(t - iTh - ii) where x, is the relative delay of each path to the one received first (that is, x0 = 0). The transmitted and received message (after spreading) is shown in Fig. 3.2 below. Bit streams transmitted bit / = -1 bit i = 0 bit i = 1 received through: path 1 path 2 path 3 i—r T,=0 2Th 37), time, t Fig. 3.2: Bit streams in a channel with multipath propagations The expression for r(t), given in Model 2, then becomes: L M KO = X Z A ib\}Kt~iT b -z,) + n(t) /=1 i=-M 27 The received amplitude, At from each propagation path is given by A] = ai(f) A where a/(t) is the attenuation on path /. A is transmit amplitude and is the same for all the different paths. 3.4 Model 4: Multiple Asynchronous Users in Multipath Propagation and AWGN Channel conditions commonly found in practical communication systems will consist of multiple users transmitting asynchronously. The signals of each user will also propagate through multiple paths. These conditions are summarized below: K asynchronous users Binary Antipodal Modulation (Binary Phase Shift Keying) in Additive White Gaussian Noise L propagation paths Each of the K users' signals is assumed to propagate through L independent paths. The received signal, xk(t) for the M i user is similar to that provided in Model 3 above. Thus, L M where Akj is the received amplitude for user k through the /th path. xkJ ( = xk + x,) is the delay on the /th path of the M i user. The consolidated signal, r(i) observed by the receiver is given by K k=\ With K different users on L multiple propagation paths, there are K * L constituent signals in the received signal, r(t). 3.5 QPSK Signals 3GPP systems use QPSK in spreading (and in modulation) of signals (cf. Fig. 2.3 and Fig. 2.5). Analysis could still be done at the baseband. In such situations, signals modulated into QPSK will be separated into their BPSK representation before interference cancellation is applied. With respect to the previous models, the signals in the expressions will be complex signals. Otherwise, the models remain applicable. In Model 1 for example, received signal becomes: 28 K k=\ where a new font-type has been used to represent complex signals. 29 Chapter 4 Simulation Process The Parallel Interference Cancellation (PIC) scheme described in Chapter 2 was simulated for a 3GPP uplink air interface. For comparisons, the Successive Interference Cancellation (SIC) and Group-wise Serial Interference Cancellation (GSIC) schemes were also simulated. This chapter discusses how the simulations were done. Results are presented in the next chapter. 4.1 The Communication System The communication system simulated is based on the 3GPP model. A number of uplink users transmit to the same base station concurrently. The two-user case is demonstrated in Fig. 4.1 below. Users are separated by choosing different scrambling codes. Signals from the different users are combined in the channel. User l is considered the desired user. Processing of the received signal is with the effort to retrieve data transmitted by User l . The message signal retrieved is compared to the transmitted signal to determine the bit error rate (BER). generate data spread scramble modulate I cdl tscdl generate data spread scramble modulate cd2 scd2 decision despread descramble demodulate cd l I scdl Fig. 4.1: Simulation block diagram 4.1.1 Generating Data Data is generated as random bits (O's and 1 's) and then converted to real numbers (±1) . Each user generates payload data carried in the Dedicated Physical Data Channel (DPDCH) and control information carried in the Dedicated Physical Control Channel (DPCCH). Both are combined to form the Dedicated Channel (DCH) data using the frame structure described in Chapter 2. 30 4.1.2 Spreading Spreading uses Orthogonal Variable Spreading Factor (OVSF) codes (also called channelization codes). It allows for transmission of more than one physical data channel at a time to increase data rate. The physical data channel(s), DPDCH's and the physical control channel, D P C C H for each user are multiplexed together to form a single D C H . There are two possible cases. If the number of DPDCH's to be transmitted simultaneously is more than one, the multiple DPDCH's are separated by using different channelization codes for each. The D P C C H is also spread to the same chip rate to separate it from the accompanying DPDCH's. Fig. 2.3 in Chapter 2 shows this relationship. The channels are then combined as in-phase and quadrature components of a Quadrature Phase Shift Keying (QPSK) signal (cf. Fig. 2.3). If there is only one DPDCH, it would be separated from the D P C C H using QPSK as shown in Fig. 4.2 below. Spreading is still applied to both channels to increase bandwidth. Fig. 4.2: Spreading of uplink dedicated channels with only one D P D C H Generation of OVSF codes was discussed in Chapter 2. The channelization codes used for each spreading factor (SF) when there is only one D P D C H to be transmitted are listed in Table 4.1 below. Table 4.2 shows the codes used for multiple DPDCH's. Table 4.1: Channelization codes (single DPDCH) SF Channelization code generated 64 [ 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 -1 ] 16 [ 1 1 1 1 - 1 - 1 - 1 - 1 1 1 1 1 - 1 - 1 - 1 - 1 1 8 [ 1 1 - 1 - 1 1 1 - 1 - 1 1 4 [ 1 - 1 1-1 ] 31 Table 4.2: Channelization codes (multiple DPDCH) SF Channel Number Channelization code generated 4 1,3 [ 1 - 1 1-1 ] 2,4 [ 1 - 1 - 1 1 ] 5,6 [ 1 1 - 1 - 1 ] Al l DPCCH's are spread with the same OVSF code at a spreading factor of 256. The chips of the code are all set to 1. Al l users use the same code for their control channels. As evident from the foregoing discussions, channelization codes used for different users with different spreading factors will be different. For users with the same spreading factor and number of dedicated physical data channels, the same channelization codes are used. Different users are separated using scrambling codes. 4.1.3 Scrambling Each user is allotted a unique complex scrambling code (Sdch in Fig. 4.2 above) by U T R A N for the duration of calls that require a dedicated connection. The scrambling code is applied to the QPSK signal obtained after spreading. Scrambling serves to separate different users for multiple access transmitting and receiving. Two different algorithms are provided in the 3GPP Standard for generating scrambling codes [10], resulting in 'short' scrambling codes and 'long' scrambling codes. Short scrambling codes are of length 255. The code is repeated afterwards for message signals longer than that. Long scrambling codes are of length 2 2 5 - 1. They are truncated to a frame length of 38,400 chips. Both codes are complex. Scrambling is implemented as a complex multiplication of the signal and the scrambling code. Scrambling does not increase the bandwidth of the message signal any further than has been done during spreading. The message signal is already at the chip rate before being scrambled and it remains at the chip rate after scrambling. 4.1.4 Modulation A root raised cosine (RRC) filter is applied to the signal being transmitted for pulse shaping. The impulse response of the filter is defined as [32]. 32 \ 4at sin — ( l-a) H cos — (l + a) T J rp \ / K1c J m 1 -Uat) 2 " { Tc J Eq.4.1 where Tc( = \l 3.84 x 106 ) is the chip duration and a ( = 0.22 ) is the roll-off factor. 4.1.5 Channel The simplest channel situation is obtained when there is no fading. Signals transmitted by the different users in the communication system are simply added together in the channel. Multipath and Rayleigh fading are used as test conditions in some simulations. The signal transmitted by each user is sent through a delay, attenuated and then multiplied by a Rayleigh process (Fig. 4.3). The delay and attenuation, which are varied for each of the multiple paths, are relative to the first path. The Rayleigh process for each path is independent. Rayleigh process delay attenuation Rayleigh process delay attenuation Rayleigh process Fig. 4.3: Multi-path fading model 4.1.5.1 Multipath fading propagation conditions The standard for 3GPP systems provides performance test parameters for U E [32] and U T R A N [33] receivers in various propagation conditions. Specifications in these documents for multipath propagations are used in the simulations. Different cases of relative channel delay, x and average power, A were considered for each path in a channel with multipath propagation. These are listed in Table 4.3 below. 33 Table 4.3: Propagation Conditions for Multipath Fading Environments Case 1, Case 2, Case 3, Case 4, speed 3km/h speed 3 km/h speed 120 km/h speed 250 km/h Relative Average Relative Average Relative Average Relative Average Delay Power Delay Power Delay Power Delay Power [ns] [dB] [ns] [dB] [ns] [dB] [ns] [dB] 0 0 0 0 0 0 0 0 976 -10 976 0 260 -3 260 -3 20000 0 521 -6 521 -6 781 -9 781 -9 4.1.5.2 Rayleigh Process Rayleigh processes were generated using Jake's model [34]. The mathematical expressions used in generaton of processes are presented below. N Set N =34 Define Nn= — I 2 j Number of constituent waves in the received signal large enough for central limit theorem to apply and other conditions number of oscillators used in simulation, less 1 Also, cvm =2rfd maximum Doppler frequency (in radians) where fd =v (fl c) v = mobile speed; / = transmit frequency; c = speed of light = 3 x 108 m/s and tv„ = co„, COS(2TOT / N) Doppler shifts for each oscillator n =1,2, . . . N o The real and imaginary parts, xc(t) and xs(t) respectively, of the signal are: — 0 xc(t) = V2 cos cc cos comt + 2 ^ cos /3n cos cont "0 xs(t) = V2 sin a cos comt + 2 ^ sin f3n cos coj where a = nl 4 p = nn I N0 The final composite output signal y(t) is y(t) = xc(i) cos coc t + xs(t) sin coc t 34 where coc is the carrier frequency in radians. In the m-file provided (in Appendix B), the carrier frequency was not used. Therefore, xc{t) and xs(t) were simply combined as real and imaginary components of a complex process without multiplication with cosine and sine. Performance of the Rayleigh process simulation software was tested before being put to use. Results of the tests (including a plot of the cumulative distribution) are in Appendix B; along with sample Rayleigh processes. 4.2 Conventional Receiver Demodulation, descrambling and despreading follow directly from the equivalent uplink blocks. In the conventional receiver, only the desired user's signal is received. Any other signal in the communication system is ignored. A rake receiver with as many rake fingers as the number of paths in the channel is used. In each rake finger, the effect of delay is reversed from the received signal. The resulting signal is then demodulated. Descrambling and despreading will decode the message signals from the various paths. Output of all the paths for each user are combined using maximal ratio combining. This is shown in Fig. 4.4 [35]. combine cdl scdl despread h — descramble \ * — demodulate despread descramble demodulate reverse delay despread descramble demodulate reverse delay Fig. 4.4: Rake receiver Output of the rake receiver is sent to a decision device to determine the data bits transmitted. B E R estimates are obtained by comparing the output of the decision device to the data bits actually transmitted. 35 4.3 Parallel Interference Cancellation As discussed in Chapter 2, to implement parallel interference cancellation, all interfering users' signals must be received, regenerated and subtracted from the received signal. The conventional receiver discussed in the preceding section is used to receive each interfering user's message signal. For the purpose of interference cancellation, message signals transmitted by all users in the communication system must be received. 4.3.1 Regeneration Signals are regenerated by spreading, scrambling and modulating the message signal received at the output of a rake receiver. Fig. 3.4 shows a regeneration block following a rake receiver. combine cdl despread scdl rake receiver descramble demodulate despread descramble demodulate reverse delay despread descramble demodulate reverse delay spread scramble modulate delay regenerate'' Fig. 4.5: Regeneration of received message signals The same channelization code (cdl in the figure) is used in all spreading and despreading blocks. Similarly, all scrambling and descrambling blocks use the same scrambling code (scdl in the figure). Perfect knowledge of channel coefficients (channel delay and phase for each path and for each user) is assumed. 36 4.3.2 Interference Cancellation A sum of the regenerated signals for all users in the system, with the exception of the desired user, constitutes the interference that needs to be removed. Interference so calculated is subtracted from the received signal. The residual signal is then passed through a rake receiver. A two-user example is shown in Fig. 4.6. Fig. 2.9 shows the flow of signals for any number of users. delay rake receiver cd2 regenerate scd2 cd2 scd2 rake receiver ) ' cdl I scdl 1 PIC receiver Fig. 4.6: Parallel Interference Cancellation 4.3.3 Multiple Cancellation Stages The PIC receiver shown in Fig. 4.6 above or in Fig. 2.9 can be cascaded for better results. To do so, signals of all users in the communication system have to be received and processed through the respective stages shown: rake receiver, regenerate and cancel out interference, and another rake receiver. Thus the PIC receiver will produce as output, signal(s) for each user in the system, desired or not. The number, type and purpose of the output of such a PIC receiver is similar to those obtained from the rake receiver. To obtain multiple cancellation stages, the output of each PIC receiver stage is provided as input to the next stage as shown in Fig. 4.7. The PIC receiver block in the figure is similar to the dotted-line box in Fig. 4.6, except that signals of all users in the communication system are processed by the PIC receiver blocks in Fig. 4.7. The channelization codes (cd) and the scrambling codes (sed) for all users in the system are required in the bank of rake receivers, and in each of the PIC receiver blocks. delay rake receiver cd delay PIC receiver sed cd sed decision delay PIC receiver cd sed decision PIC receiver cd sed decision Fig. 4.7: Multiple Parallel Interference Cancellation Stages 37 4.3.4 PIC Receiver Outputs The output of each rake receiver is sent through a decision device (cf. section 4.2 above). This makes a hard decision of ±1 on its input signal. In Fig. 4.5, the decision device follows the maximal ratio combining block. The signal inputs into the regeneration block are taken out of the rake receiver before maximal ratio combining is performed. Effectively, no hard decision is made on the signals before being regenerated. For the purpose of cascading, two types of output come out of each PIC receiver stage in Fig. 4.7. One is taken from the combiner output of the rake receiver in that stage. This goes to a decision device. The second output is a set of all the rake finger output signals taken just before the combiner. These are the inputs to the combiner and the regeneration blocks in Fig. 4.5. They are supplied to the next PIC receiver stage (if one exists). B E R achieved at each stage is obtained by comparing the output of the decision device (at the stage) with the message data sent by the user. 4.4 Group-wise Serial Interference Cancellation Group-wise Serial Interference Cancellation (GSIC) uses the conventional receiver as the first stage and the PIC receiver as the internal multiuser detection receiver. Signals of all users in the communication system are received using the conventional receiver. This enables estimation of the received signal strengths for each user. Users are ranked and grouped based on the signal strengths such that the first group consists of the strongest signals and the last group is made up of the weakest. Following the grouping of signals, the first group is regenerated (from the rake receiver outputs) and subtracted as interference from the received signal. The residual is passed through a PIC receiver where signals in the second group are received. The output of the PIC receiver for signals in the second group is regenerated and subtracted from what remains of the received signal (which has the first group already cancelled out). The residual is again sent through a PIC receiver to receive message signals for the third group. This process continues until the desired user's group has been received. If the desired user is in the last group, all the signals transmitted will be received before the process stops. On the other hand, if the 38 desired user is in the first group, processing stops before the first PIC receiver (that receives the second group) is entered. The number of groups is predetermined based on the number of users in the communication system. For instance, two groups are used for a four-user system; three groups for a nine-user system and five groups when the number of users in the system is 25. 4.5 Successive Interference Cancellation Simulation of the Successive Interference Cancellation (SIC) is similar to the GSIC. The main difference is the number of groups to which the users are separated. In GSIC, the number of groups used is always less than the number of users. Therefore, some or all of the groups will contain more than one user. In SIC, the number of groups is equal to the number of users. So, each group contains exactly one user. 4.6 Calculation of Signal to Interference Ratio (SIR) For a communication system with N active users, SIR is calculated as «=i where Sd is the average power of the desired user and S„ is the average power of each interfering user. The power, S,i and S„ are calculated as: S d = i s d J x S F i=\ L 1=1 where Sdj and S„j represents the power in a single path, / for each desired and interfering user respectively. SF is the spreading factor of the desired user. Thus, the total power for each user is the sum of the power in the different paths for the user. In the simulations, Stl and S„ are calculated in the channel, after all channel processing has been done but before the signals are added. Only the power in the D P D C H (containing payload) is calculated. 39 4.6.1 Obtaining multiple SIR points. We required simulations at more than a single SIR point to plot a curve. Varying SIR was achieved by fixing the transmit power of the interfering users' signals while varying the transmit power of the desired user. Throughout the simulations, the transmit power level of interfering users was set to 0 dB. Depending on the number of users and the channel situations, the transmit power level of the desired user varied widely from -20 dB to 10 dB. Since some channels consisted of multiple paths and Rayleigh fading, the received signal level could vary significantly, compared with transmit power level. So, SIR was calculated in the channel, just before all the signals for the different paths and users were added. 40 Chapter 5 Results Results of the simulations discussed in Chapter 4 are presented here. General discussions of the results are included. Conclusions are drawn based on the results. Simulation results are plotted as bit error rates (BER) versus the signal to interference ratio (SIR) in dB. No (Gaussian) noise was added to transmitted signals in the channel (see Fig. 4.1). Reliability of the simulation software was checked by plotting results obtained from it alongside standard theoretical results. Examples of such plots are provided in Appendix B. Results obtained using Parallel Interference Cancellation method (PIC) are presented first. This is followed by discussions about the results from the Successive Interference Cancellation (SIC) method and then the Group-wise Serial Interference Cancellation (GSIC) method. Towards the end of the discussion on the PIC receiver, a comparison of all three methods is presented. 5.1 Parallel Interference Cancellation Results of simulations employing the PIC receiver are presented in this section, in comparison to the conventional receiver. The PIC receiver was tested for the following limitations: • different channel situations • number of users in the communication system • scrambling code generation method • different spreading factors • multiple interference cancellation stages It was also compared with the other two methods investigated: SIC and GSIC. Except where the two methods of scrambling code generation (described in section 4.1.3) are being compared, all simulations employed the long scrambling code. 5.1.1 Different Channel Situations Simulations were run to create a system with two users both transmitting at a spreading factor (SF) of 64 in the following channel situations for comparison: • no fading 41 • 2-path Rayleigh fading, user equipment speed at 3 km/h • 3-path Rayleigh fading, user equipment speed at 3 km/h The result is presented in Fig. 5.1 below. The performance of the conventional receiver is shown as continuous lines and labeled as 'BER woe' (bit error rates without cancellation) in the legend. There are three continuous lines, one each for the three different channel situations. The performance of the PIC receiver is shown in dashed-lines and labeled as 'BER wc' (bit error rates with cancellation) in the legend. 10 10 10"' & 10° TO tr o LU m 10" 10 10 10 - * - BER woe, no fading x- BER wc, no fading -e- BER woe, 2-paths O BER wc, 2-paths BER woe, 3-paths BER wc, 3-paths Parallel Interference Cancellation No. of users = 2 S F = 64 -30 -25 -20 -15 -10 -5 0 5 Signal to Interference Ratio (dB) 10 15 20 Fig. 5.1: Performance in different channel situations (with 2 users) The result shows that, compared to the conventional receiver, the PIC receiver provides performance improvement in every channel situation. This improvement is about 17 dB in the 3-path channel at a BER of 10"4. The performance improvement when there is no fading in the channel is much higher (about 35 dB). The result obtained for similar simulation conditions as above, but with four users in the system is shown in Fig. 5.2. 42 10" 10 10 E 10 10 1 0 ' 10 Para l le l Interference Cance l la t ion N o . of users = 4 S F = 64 — * - B E R woe, no fading - x - B E R wc , no fad ing - e - B E R woe, 2-paths -O- B E R wc , 2-paths - * - B E R woe , 3-paths * B E R w c , 3-paths -30 -20 -10 0 10 S igna l to Interference Rat io (dB) 20 30 Fig. 5.2: Performance in different channel situations (with 4 users) 5.1.2 Number of Users in the Communicat ion System A communication system with a 3-path Rayleigh fading channel was simulated while changing the number of users. In the result, shown in Fig. 5.3, performance of the PIC receiver is better than that of the conventional receiver for the different number of users considered. This performance improvement is however reduced as the number of users increase. The improvement is about 4 dB at B E R of 10"4 for a system with 25 users. This is much lower than the 17 dB improvement observed for the same system when there are only two users. Please see Fig. 5.4. The diminishing performance as the number of users increase is due to the effect of detection errors on interference cancellation. Increasing the number of users in the communication system raises the noise floor for each user. That increases the error rate for each interfering user, which then affects performance improvement obtained from interference cancellation. 43 10 10 10 B 10" ; a. m 10 1 0 ' 10 10 - * - B E R woe, 2 users - x - B E R wc, 2 users - & - B E R woe, 4 users - O - B E R wc, 4 users B E R woe, 9 users - * - B E R wc, 9 users -a- B E R woe, 25 users i j - B E R wc, 25 users Parallel Interference Cancellat ion No. of paths = 3; speed = 3 km/h S F = 64 -15 -10 -5 0 5 10 Signal to Interference Ratio (dB) Fig. 5.3: Performance as number of users change 15 20 18 16 14 12 \ i ® -"Paral le l Interference Cancellation -No. of paths = 3; speed = 3 km/h S F = 64 B E R = 1.0E-4 i i i E 10 <a 8 10 15 Number of users 20 25 Fig. 5.4: Performance improvement at BER of 10" for different number of users 44 In order to be able to observe performance improvement for interference cancellation, results have to be compared with those obtained when no interference cancellation was employed. That is the case in the results presented above. If, on the other hand, the SIR were calculated using one of the interfering users as a base (instead of the desired user), the effect of detection errors on interference cancellation will be clearer. 5.1.3 Scrambling Code Generation Method The two methods of generating scrambling codes available, resulting in long or short codes, were used in the simulation of a two-user communication system with a 3-path Rayleigh fading channel. The simulation results show that there is no significant performance difference due to the choice of scrambling code generation method used. Both methods produced similar performance improvements in results for the PIC receiver as compared to the conventional receiver. The results are in Fig. 5.5 and Fig. 5.6. 10" IO"1 10" S 10° (TJ I— p 10 1 0 ° t -1 0 " % 10 Parallel Interference Cancellation No. of paths = 3; speed = 3 km/h No. of users = 2 SF = 64 - * - BER woe, 'short' scrambling codes - x - BER wc, 'short' scrambling codes -©- BER woe, 'long' scrambling codes -C— BER wc, 'long' scrambling codes 5 10 Signal to Interference Ratio (dB) 15 20 Fig. 5.5: Performance difference due to scrambling code generation method (with SF = 64) 45 10u 10" 10" 2 10 rr g LU m 10"* 10 10c 10 BER woe, 'short' scrambling codes BER wc, 'short' scrambling codes BER woe, 'long' scrambling codes BER wc, 'long' scrambling codes Parallel Interference Cancellation No. of paths = 3; speed = 3 km/h No. of users = 2 SF = 16 -15 -10 0 5 10 15 Signal to Interference Ratio (dB) 20 25 30 Fig. 5.6: Performance difference due to scrambling code generation method (with SF = 16) 5.1.4 Different Spreading Factors Capacity for different data rates is important in 3GPP systems. Simulation of a system at different spreading factors (SF) was therefore done. Data rates obtained at the different spreading factors (using chip rate of 3.84 * 106) is listed in Table 5.1. SF Raw Data Rate (kbps) 64 60 16 240 8 480 The results (see Fig. 5.7 and Fig. 5.8) show that as the data rate increases (spreading factor reduces), performance improvement obtained from use of the PIC receiver reduces. This is because the interfering user has a higher error rate at lower spreading factors. The detection errors suffered by the interfering user affect performance of the interference cancellation for the desired user. 46 Signal to Interference Ratio (dB) Fig. 5.7: Performance at different spreading factors 18 50 100 150 200 250 300 350 400 450 500 Data rate (kbps) Fig. 5.8: Performance improvement at BER of IO - 4 for different data rates This simulation was done for a system with two users. As seen earlier, increasing the number of users will reduce the performance improvement obtained. Thus, at high data rates of 480 kbps (and higher), improvement can only be expected when the number of users is small (less than 10 for instance). 5.1.5 Multiple Interference Cancellation Stages A communication system with four users was simulated in a 3-path Rayleigh fading channel. The PIC receiver employed had multiple stages. Performance of each stage is compared with one another and with the performance of the conventional receiver. Results are shown in Fig. 5.9. The second stage of interference cancellation shows a significant improvement (of about 2.5 dB) over the first stage. That is, a two-staged PIC receiver is much better than a single staged. Subsequent stages after the second show only marginal performance improvement and are not recommended. Therefore, use of multiple interference cancellation stages improves performance of the PIC receiver to a limit. Parallel Interference Cancellation No. of paths = 3; speed = 3 km/h -15 -10 -5 0 5 10 15 20 Signal to Interference Ratio (dB) Fig. 5.9: Effect of multiple interference cancellation stages 48 5.1.6 Comparison with SIC and GSIC The other two interference cancellation schemes, SIC and GSIC are discussed in the next sections. Some of the results presented under those discussions are combined here to compare the receivers with the PIC receiver. Fig. 5.10 shows the results for a system with nine users at SF of 64 in a 3-path Rayleigh fading channel employing PIC, SIC and GSIC receivers. The figure shows that there is no significant difference between the three different schemes at this data rate and number of users. It is noted however that when SIR is greater than 10 dB, performance improvement obtained from SIC and GSIC receivers tend towards 0 dB. Also, the SIC receiver is better than the GSIC receiver, which is better than the PIC receiver, at low values of SIR (although the difference is very small). o 10 - H - BER WOC - » BER WC, SIC •O BERWC, OSIC •B BERWC, PIC - 6 No. o l paths = 3; speed = 3 km^i No. of users = 9 SF = 64 10 - 1 5 - 1 0 - 5 0 5 Signal to Interference Ratio (d B) 10 15 Fig. 5.10: Performance of the different interference cancellation schemes 49 5.2 Successive Interference Cancellation Results obtained from simulations employing SIC receiver are presented in this section. Most of the observations are similar to those of the PIC receiver. There is one major difference: the performance improvement tends to 0 dB at high SIR. 5.2.1 Different Channel Situations Results of simulations in different channel conditions are shown in Fig. 5.11. Similar to the PIC receiver, SIC receiver delivers performance improvement in comparison to the conventional receiver in all channel situations. At BER of IO - 4, the improvement is about 12 dB in a 3-path Rayleigh fading channel. More improvement is observed when the channel has no fading. j " s i i i i i i i i i I -25 -20 -15 -10 -5 0 5 10 15 20 Signa l to Interference Rat io (dB) Fig. 5.11: Performance in different channel situations 50 5.2.2 Number of Users in the Communicat ion System Fig. 5.12 is the result of a simulation to show effect of increasing number of users. The performance improvement for the 4-user case is better than the improvement for the 9-user case. In addition, for the 9-user case, the BER curve due to cancellation tapers towards the curve for the conventional receiver. This occurs from about 6 dB of SIR. The reason for the tapering is that the cancellation scheme stops benefiting the desired user when the desired user's signal is stronger than the interfering users'. The SIC algorithm ensures signals arriving with the strongest power are received and cancelled out before the rest. So, when the desired user's signal is received with higher power compared with the interfering users' signals, the desired user does not benefit from the interference cancellation done. 10 l ,0 B E R woe , 4 use rs B E R wc , 4 use rs B E R woe , 9 use rs -O™ B E R wc , 9 use rs S u c c e s s i v e Interference Cance l la t ion N o . of paths = 3; s p e e d = 3 km/h S F = 64 -15 -10 -5 0 5 10 15 20 Signa l to Interference Rat io (dB) Fig. 5.12: Performance as number of users change 51 5.3 Group-wise Serial Interference Cancellation For the most part, the discussions about the PIC and SIC receivers apply to GSIC as well. The results presented here show the main similarities and differences of GSIC with the other two methods considered. 5.3.1 Different Channel Situations As in the preceding section, this is a presentation of the results for three different channel conditions simulated. The result is in Fig. 5.13. Observations from the results are similar to those discussed in the preceding section. That is, improvement in performance is observed in all channel situations for the GSIC receiver relative to the conventional receiver. 10 10 1<f F P 10 ' 10 10 10" - * - BER woe, no fading - x - BER wc, no fading -e- BER woe, 2-paths - O - BER wc, 2-paths - * - B E R woe, 3-paths * B E R wc, 3-paths Groupwise Serial Interference Cancellation No. of users = 4 S F = 64 -25 -20 -15 -10 -5 0 5 Signal to Interference Ratio (dB) 10 15 20 Fig. 5.13: Performance in different channel situations 52 5.3.2 Number of Users in the Communicat ion System The effect of increasing number of users is demonstrated in Fig. 5.14 below. The result shows that the performance improvement obtained goes towards 0 dB for the 9-user case at high SIR values. In addition, the 4-user case offers more performance improvement. 10 10 10" <5 or | 10"3 LU 10 10' 10" G r o u p w i s e Ser ia l Interference Cance l la t ion N o . of paths = 3; s p e e d = 3 km/h S F = 64 - * - B E R woe , 4 use rs x - B E R wc , 4 users - © - B E R woe , 9 use rs -o- B E R wc , 9 use rs -15 -10 -5 0 5 10 Signa l to Interference Rat io (dB) Fig. 5.14: Performance as number of users change 15 20 53 Chapters Conclusions In this thesis work, we have been able to demonstrate usability of some multiuser detection methods using the Third Generation Partnership Project (3GPP) Standard. This chapter is a summary of our findings. Discussions on each of the three methods investigated follows. These are the Parallel Interference Cancellation (PIC), Successive Interference Cancellation (SIC) and the Group-wise Serial Interference Cancellation (GSIC) receivers. Areas requiring further studies are also provided. 6.1 Similarities between the different receivers Any of the three multiuser detection methods investigated is equally suitable for a 3 GPP system at low SIR, low data rates and a small number of users. Al l three methods give appreciable performance improvement over the conventional receiver. The improvement obtained varies based on the data rate, number of users and the method used. In a communication system with four users for instance, the PIC receiver saves about 10 dB in a 3-path Rayleigh fading channel at a BER of 1 x IO - 4. More improvement can be obtained with a two-stage cascaded PIC receiver. Similarly, both the SIC and GSIC receivers offer improvements of about 12 dB and 10 dB respectively for the desired user in these same conditions. General limitations on the receivers are discussed below. Reducing the capacity of the communication system as mentioned in the descriptions can be used to offset part of the limitations and obtain more performance improvement from the multiuser detection receiver. 6.1.1 Data rates Spreading factors of 64 or 16 are observed to provide significant performance improvements. These give raw data rates of 60 and 240 kbps at the Physical layer. Lower spreading factors (at 8 or lower) produce performance improvement only for very small number of users (2 or 4) or in channels without significant degradations (such as fading). 54 6.1.2 Number of users Very good performance improvement is obtained with any of the multiuser detection receivers when the number of users is small (less than 25). Increasing the number of users in the system increases the noise floor, and thus the error rates of interfering users. High data rates also lead to increased error rates for interfering users due to the lower spreading gain. Thus, in both cases, performance improvement obtained from interference cancellation receivers is reduced. For a large number of users, high spreading factors or channels with low degradations are required to obtain good performance improvements. 6.2 Differences between the different receivers For SIC and GSIC receivers, an additional limitation is that the SIR as observed by each desired user should be below 0 dB. At higher SIR, performance improvement from use of either multiuser detection receivers approaches 0 dB. No improvement is obtained at very high SIR. A 3GPP system employing power control will therefore not benefit from use of these two methods when the power control is effective in bringing the SIR of each desired user above 0 dB. The PIC receiver is better at about 0 dB. Both the SIC and GSIC receivers are better than the PIC receiver is when desired user's SIR is much lower than 0 dB. Performance of the GSIC receiver is always lower than that of the SIC receiver. This is because the GSIC receiver employs a PIC receiver internally to receive grouped signals. 6.3 Practical Applications From the foregoing analysis, we conclude that a 3GPP communication system where the number of users is not very large and the data rate requirements of each user is low will benefit significantly from the use of a PIC receiver. Examples of such systems will be in small-scale communication systems such as Wireless Local Area Networks (LANs) based on the 3GPP Standard. Care must be taken for situations when SIR is much greater than 0 dB. Performance of the PIC receiver could be worse then. Commercial cell phone use with a large number of users per base station will not find good use in any of these cancellation schemes. A workaround for such situations will be to use sectors in the base station to divide the number of users into groups such that the number of users in each group is low (for 55 example 25 or less). Such sectorization of cells is often employed in practice. Then, any of the three multiuser detection schemes discussed here can be applied. 6.4 Areas of further studies In the 3GPP Standard, Wideband C D M A can be used with FDD or TDD. We have investigated performance in the FDD environment only. Use of multiuser detection in T D D could also be investigated. If time allowed, we planned to implement the PIC receiver using V H D L . This remains an important study area. It will confirm that the receiver can be implemented in practice. It will also provide an insight into the cost, relative to the conventional receiver. 6.4.1 Effect of Power Control Fast closed loop power control at 1500 Hz is always used when a link is established between a user equipment and a base station. It is quite effective in a Rayleigh fading channel, helping to bring the SIR towards a desired level. Open loop power control is used when no link exists between the user equipment and the base station. This happens: • during an initial call set-up. • when packet transmission is done from user equipment to the base station in the absence of an established link. If transmission of packets continues for a designated period, a dedicated channel will be established. Then, closed loop power control will be used. During a session where open loop power control is used, the user equipment estimates the power required for transmission from a reference signal broadcasted by the base station. Considering the difference in the uplink and downlink transmit frequencies, the power used as a result of this measurement will only be a rough estimate, at best, of the appropriate power level. From the foregoing, it can be observed that open loop power control will not be effective in combating effects of Rayleigh fading in the channel. It is only used in situations where closed loop power control cannot be used. In the absence of open loop power control however, user equipment will have no other means of determining the appropriate transmit power level. 56 If closed loop power control is employed, and the target SIR is 0 dB or more, the system will not benefit from use of either the SIC or the GSIC receiver. On the other hand, if the target SIR is the same for all the users in the system, both SIC and GSIC receivers can be used to obtain significant performance improvement. These receivers could also be used if the system is using open loop power control instead of closed loop control. 57 References 1. C. Huang, "An Analysis of C D M A 3G Wireless Communications Standards", IEEE 49th Vehicular Technology Conference, 1999, Vol. 1, pp 342 - 345. 2. R. Prasad and T. Ojanpera, "A Survey on C D M A : Evolution Towards Wideband C D M A " , IEEE 5th International Symposium on Spread Spectrum Techniques and Applications, 1998. Proceedings. Vol. I ,pp323 - 331. 3. E . Dahlman, P. Beming, J. Knutsson, F. Ovesjo, M . Persson and C. Roobol, "WCDMA-The Radio Interface for Future Mobile Multimedia Communications", IEEE Transactions on Vehicular Technology, vol. 47, no 4, Nov. 1998, pp 1105 - 1118. 4. H. Holma and A. Toskala, " W C D M A for UMTS: Radio Access for Third Generation Mobile Communications", John Wiley & Sons, 2000. 5. "General U M T S Architecture", 3rd Generation Partnership Project; Technical Specification Group Services and System Aspects, Technical Specification (3GPP TS 23.101). ftp://ftp.3gpp.org/specs/2002-09/Rel-4/23_series/23101-400.zip. 6. "Radio Interface Protocol Architecture", 3rd Generation Partnership Project; Technical Specification Group Radio Access Network, Technical Specification (3GPP TS 25.301). ftp://ftp.3gpp.org/specs/2002-09/Rel-4/25_series/25301-440.zip. 7. "Services provided by the Physical Layer", 3rd Generation Partnership Project; Technical Specification Group Radio Access Network; Technical Specification (3GPP TS 25.302). ftp://ftp.3gpp.org/specs/2002-09/Rel-4/25_series/25302-460.zip 8. "Physical layer - General description", 3rd Generation Partnership Project; Technical Specification Group Radio Access Network; Technical Specification (3GPP TS 25.201). ftp://ftp.3gpp.org/specs/2002-09/Rel-4/25_series/25201-430.zip 9. "Physical channels and mapping of transport channels onto physical channels (FDD)", 3rd Generation Partnership Project; Technical Specification Group Radio Access Network; Technical Specification (3GPP TS 25.211). ftp://ftp.3gpp.org/specs/2002-09/Rel-4/25_series/2521 l-460.zip 58 10. "Spreading and modulation (FDD)", 3rd Generation Partnership Project; Technical Specification Group Radio Access Network; Technical Specification (3GPP TS 25.213). ftp://ftp.3gpp.org/specs/2002-09/Rel-4/25_series/25213-430.zip 11. L. Wang, A . H . Aghvami, W. G. Chambers, "On the Capacity of a Wideband Multi-Rate C D M A System With Multi-Service in the Presence of Fading and Power Control Error", IEEE International Conference on Telecommunications, ICT 2001. 12. S. Verdu, "Multiuser Detection", Cambridge University Press, 1998. 13. J. G. Proakis, "Digital Communications", 3rd ed., New York; McGraw-Hill, 1995. 14. S. Moshavi, "Multiuser Detection for D S - C D M A Communications", IEEE Communications Magazine, October 1996, pp 124 - 136. 15. A. Duel-Hallen, J. Holtzman and Z. Zvonar, "Multiuser Detection for C D M A Systems", IEEE Personal Communications, April 1995, pp 46-58. 16. S. Verdu, "Minimum Probability of Error for Asynchronous Gaussian Multiple-Access Channels", IEEE Transaction on Information Theory, January 1986, pp 85 - 96. 17. A. Klein, G. K. Kaleh, P. W. Baier, "Equalizers for Multiuser Detection in Code Division Multiple Access Mobile Radio Systems", IEEE 44th Vehicular Technology Conference, 1994, Vol 2 of 3, pp 762-766. 18. K. I. Pedersen, T. E. Kolding, I. Seskar, J. M . Holtzman, "Practical Implementation of Successive Interference Cancellation in D S / C D M A Systems", IEEE 5th International Conference on Universal Personal Communications, 1996, ICUPC'96. Proceedings of, Part 1 (of 2) pp 321-325. 19. M . K. Varanasi and B. Aazhang, "Multistage detection in asynchronous code-division multiple-access communications" IEEE Transactions on Communications, Vol. 38, No. 4 , April 1990, pp 509 - 519 20. A. Grant and C. Schlegel, "Convergence of Linear Interference Cancellation Multiuser Receivers", IEEE Transactions on Communications, Vol. 49, No. 10, October 2001, pp 1824 - 1834. 21. D. Divsalar, M . K. Simon and D. Raphaeli, "Improved Parallel Interference Cancellation for C D M A " , IEEE Transactions on Communications, Vol. 46, No. 2, February 1998, pp 258 - 268. 22. A . - L . Johansson, "Group-wise successive interference cancellation in multirate C D M A systems" Vehicular Technology Conference, 1999 IEEE 49th , Vol. 2, 1999, pp 1435 - 1439. 59 23. A . - L . Johansson, L. K. Rasmussen, "Linear group-wise successive interference cancellation in C D M A " , Spread Spectrum Techniques and Applications, 1998. Proceedings., 1998 IEEE 5th International Symposium on, Vol. 1, pp 121 - 126. 24. F. van der Wijk, G. M . J. Janssen, R. Prasad, "Groupwise successive interference cancellation in a D S / C D M A system", Personal, Indoor and Mobile Radio Communications, 1995. PIMRC'95, Sixth IEEE International Symposium on, Vol. 2, 1995, pp 742 - 746. 25. A. Hottinen, H. Holma, A. Toskala, "Performance of Multistage Multiuser Detection in a Fading Multipath Channel", Personal, Indoor and Mobile Radio Communications, 1995. PIMRC '95., Sixth IEEE International Symposium on , Volume: 3 , pp 960. 26. A . Hottinen, H. Holma, A. Toskala, "Multiuser Detection for Multirate C D M A Communications", IEEE International Conference on Communications, 1996. ICC '96, Vol. 3 pp 1819 - 1823 27. S. Hamalainen, H . Holma, A. Toskala, "Capacity evaluation of a cellular C D M A uplink with multiuser detection", Spread Spectrum Techniques and Applications Proceedings, 1996., IEEE 4th International Symposium on , Volume: 1 , 1996, pp 339 - 343. 28. M . Mozaffaripour, R. Tafazolli, "Multisensor Partial Parallel Interference Cancellation for U M T S Uplink", 3G Mobile Communication Technologies, 2001. Second International Conference on; pp 392 -395. 29. E . Del Re, R. Fantacci, D. Marabissi, S. Morosi, "Low Complexity Selective Interference Cancellator for a W C D M A Communication System with Antenna Array", IEEE Global Telecommunications Conference, 2001. G L O B E C O M '01. Vol. 1 pp 480 - 484. 30. M . Ammar, T. Chonavel, S. Saoudi, "Multi-stage SIC Structure for Uplink U M T S Multiuser Receiver over Multipath Rayleigh Channels", Vehicular Technology Conference, 2001. V T C 2001 Spring. IEEE V T S 53rd, Vol. 4 pp 2514 - 2518. 31. P. Pirinen, S. Glisic, "Sensitivity of Advanced Wideband C D M A Network Capacity to Various Channel and System Parameter Imperfections" Communications, 2001. ICC 2001. IEEE International Conference on, Vol. 10, pp 3000 - 3004. 60 32. " U E Radio Transmission and Reception (FDD)", 3rd Generation Partnership Project; Technical Specification Group Radio Access Network; Technical Specification (3GPP TS 25.101). ftp://ftp.3gpp.org/specs/2002-09/Rel-4/25_series/25101-450.zip. 33. "BS Radio Transmission and Reception (FDD)", 3rd Generation Partnership Project; Technical Specification Group Radio Access Network; Technical Specification (3GPP TS 25.104). ftp://ftp.3gpp.org/specs/2002-09/Rel-4/25_series/25104-450.zip. 34. W. C. Jakes, "Microwave Mobile Communications", John Wiley & Sons Inc., 1974. 35. T. S. Rappaport, "Wireless Communications, Principles and Practices", Prentice Hall Inc., 1996 61 Appendix A Abbreviations | A list of abbreviations used in this document, and their meanings, is provided in this appendix. 3G Third Generation 3GPP Third Generation Partnership Project A W G N Additive White Gaussian Noise BER Bit Error Rate BPSK Binary Phase Shift Keying BS Base Station C D M A Code Division Multiple Access C N Core Network D C H Dedicated Channel D L Downlink D P D C H Dedicated Physical Data Channel D P C C H Dedicated Physical Control Channel D S - C D M A Direct Sequence Code Division Multiple Access FDD Frequency Division Duplex F D M A Frequency Division Multiple Access G M S K Gaussian Minimum Shift Keying GSIC Group-wise Serial Interference Cancellation ISI Intersymbol Interference M A I Multiple Access Interference M A C Medium Access Control M L S E Maximum Likelihood Sequence Estimation M M S E Minimum Mean Square Error OSI Open System Interconnect 62 OVSF Orthogonal Variable Spreading Factor PCS Personal Communication System PIC Parallel Interference Cancellation PN Pseudonoise PSTN Public Switched Telephone Network QPSK Quadrature Phase Shift Keying RRC Radio Resource Control SF Spreading Factor SIC Successive Interference Cancellation SIR Signal to Interference Ratio T D D Time Division Duplex T D M A Time Division Multiple Access T r C H Transport Channel U E User Equipment U L Uplink U M T S Universal Mobile Telecommunication System U T R A N U M T S Terrestrial Radio Access Network W - C D M A Wideband Code Division Multiple Access 63 Appendix B Simulation Software Test Tools Some of the tools used to check quality of the output from the simulation software are provided in this appendix. The complete communication system was built in modules. Each block was tested as it was written. Al l the tests for individual blocks are omitted, except for the Rayleigh fading simulator. As the blocks were being connected to each other, the entire communication system was also tested. Results of some of these tests are provided here. The Raleigh fading simulator is discussed first. B.7 Rayleigh Fading Simulation Jake's model of a Rayleigh process was used to write the Rayleigh generation software. This was discussed in section 4.1.5.1. A listing of the Matlab m-file is provided below. This is followed by results of two tests performed on the generator: the level crossing rate and the cumulative density function (cdf). Examples of what the generated process looks like are shown afterwards. B.1.1 Software The Matlab m-file containing the source code for the Rayleigh fading simulator is listed below. IRAYLGEN Rayleigh process generator using Jakes model % r = raylgen(Ns, f s , f d , ntau) simulates Rayleigh f a d i n g f o r maximum % doppler frequency, f d at the sampling r a t e , f s . The number of % samples returned i s Ns. ntau i s an o p t i o n a l time delay which % d e f a u l t s to 1. I t i s an int e g e r which i s s c a l e d a p p r o p r i a t e l y by % d i v i d i n g by f s . Q. O % Reference: Microwave Mobile Communications % by W. C. Jakes % John Wiley & Sons Inc., 1974 % Author: Kassim Olawale. Date: 19 May, 2001 i f n a r g i n == 3, ntau = 1; end % d e f a u l t delay i f ntau i s not provided t = (ntau : ntau+Ns-1)/fs; % define time s c a l e r e l a t i v e to f s Nl = 8; % number of o s c i l l a t o r s , l e s s 1 N = (2*N1 + 1) * 2; wn = 2 * p i * f d * [ c o s ( 2 * p i * [ l :N1]/N) 1]; r i = 2 * [cos(pi*[1:N1]/Nl) c o s ( p i / 4 ) / s q r t ( 2 ) ] ; r q = 2 * [sin(pi*[1:N1]/Nl) s i n ( p i / 4 ) / s q r t ( 2 ) ] ; wnt= wn' * t ; r = r i * cos(wnt) + j * r q * cos(wnt); 64 B.1.2 Level Cross ing Rate and Cummulative Distribution Function Measured results are plotted against theoretical values below. Ray le igh S imu la to r Tes t Resu l t s : C D F theoret ica l O fd = 5 .2778 H z fd = 211.1111 H z 10 -20 -10 Ray le igh Enve lope Leve l (dB) Ray le igh S imu la to r Tes t Resu l t s : L C R 10 20 r 102 L-~ 10 10 10 • © • • • o- - - -1 • O fd = 5 .2778 H z - x fd = 211.1111 H z -45 -40 -35 -30 -25 -20 -15 -10 Leve l above m e a n (dB) Fig. B. 1: Level crossing rate and cumulative distribution results B.1.3 Sample Rayleigh Process Two channel conditions are simulated. The first channel had 3 paths Rayleigh fading and the user equipment was at 3 km/h. The second was a 4-path Rayleigh fading channel, with user equipment traveling at 120 km/h. The simulation lasted for 100 ms (or 0.1 s). This will cover 10 frames (each having 10 ms in duration). Results are presented in the figures below. The results show that all the Rayleigh processes generated for the different paths are independent. They also show, as was expected, that the user equipment speed affects the rate of change of the Rayleigh envelopes. 65 Fig. B.2: Sample processes for a channel with 3 paths -10 h Simu la ted Ray le igh fading 0.06 T i m e (s) 0.12 0.06 T i m e (s) 0.12 0.06 T i m e (s) 0.12 - N o . o f |paths = 4 Mob i le s p e e d = 120 km/h 0.02 0.04 0.06 T i m e (s) 0.08 0.1 0.12 Fig. B.3: Sample processes for a channel with 4 paths B.2 The Complete Communication System Some of the results of the tests conducted in the course of the research are presented in this section. Simulation results are compared with relevant (known) theoretical plots. B.2.1 Four users without noise or fading A communication system with four users was simulated. Fading was not introduced in the channel and no noise was added. The intention was to test the system at its simplest form. This test assumes that interfering users can be approximated as Gaussian noise. It compares simulation results to the theoretical probability of error, Pe given by: Pe=Q(fir) where Q(x) = — f e(_*2 l2)dx In *< and yis the signal to noise ratio (here replaced by signal to interference ratio: SIR). Results are provided for two data rates in Fig. B.4 and Fig. B.5 below. Plots obtained from simulation lie on the theoretical plot. 1 0 ° 10-2 S 10"3 ce g LU CQ 10"* 10"5 icr6 10"7 - 4 - 2 0 2 4 6 8 10 12 Signal to Interference Ratio (dB) Fig. B.4: Comparison of a four-user syst m with theoretical results (SF = 64) 68 BER simulated BER calculated Used long scrambling codes No. of users = 4 SF = 64 No. of bits = 4.8E+06 10 10 10 2 10" <o nc o LU CQ 10-" 10' 10 10 Used long scrambling codes No. of users = 4 SF — 8 No. of bits = 3.84E+07 2 4 6 Signal to Interference Ratio (dB) BER simulated BER calculated 10 12 Fig. B .5 : Comparison of a four-user system with theoretical results (SF = 8) B.2.2 Two users in a 3-path Rayleigh fading (no noise) A communication system with two users was simulated. Transmission was through a 3-path Rayleigh fading channel with equal average power. Gaussian noise was not added. This test compares the simulation result to three plots of theoretical probability of error: • Q(-) function (assumes interfering user approximates Gaussian noise) • Rayleigh fading function (assumes receiver monitors only one path) • Maximal ratio combining (assumes diversity intended with number of fingers in rake receiver) The probability of error in Gaussian noise, Pe was provided in the preceding section. The probability of error in a Rayleigh fading channel, Pejude is given by: :,fade 2 1- r Ay for y » 1 6 9 where yis the average value of the signal to noise ratio (here replaced by signal to interference ratio: SIR). The probability of error, PeMnc when maximal ratio combining is employed is: e,MRC 1 g f z - l + yO k=0 1 where /J 7 and Yc is the average signal to noise ratio per path (again, this was replaced by signal to interference ratio: SIR). Also, f 1 W 2 Z - l ^ e,MRC l L for yc » 1 J The maximal ratio combining curves are the most appropriate in this case. This is because the outputs of the rake receiver in the simulation were combined using maximal ratio combining. The plots (in Fig. B.6) show the simulation result falls right on the theoretical maximal ratio combining curves. It lies between plots of the Q(-) function and Rayleigh fading curve. 10 10 10 10 10 Number of users = 2 No. of paths = 3; speed = 3 km/h S F = 64 No. of bits = 1.2E+06 5 10 Signal to Interference Ratio (dB) B E R simulated 0.0 B E R calc (fade) B E R calc (MRC) 15 20 Fig. B.6: Comparison of a two-user system in a three-path Rayleigh fading channel with theoretical results 70 

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:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0065594/manifest

Comment

Related Items