Uplink Admission Control in WCDMA Systems

Transcription

1 Uplin Admiion Control in WCDMA Sytem Jeper Högberg T97 June 2001 Mater Thei Supervior: Ulf Nilon, Per Erntröm, Telia Reearch AB Examinator: Clae Trygger, Optimization and Sytem Theory, KTH

2 Abtract In future wirele communication ytem there will be a traffic mix of different ervice with different demand on data rate and delay. In addition to peech ervice, which are already preent in today ytem, there will alo be video telephony, , and WWW ervice The European 3 rd generation mobile ytem UMTS ha been deigned with all of thee ervice in mind. There will, however, be new demand on radio reource management in UMTS, which i much more flexible than, for example, GSM. In UMTS, uer hare the ame frequency. Thu, one uer will caue interference for other uer in the ytem. New uer will only be granted acce to the networ if they do not caue too much interference for already active uer. The idea i that it i much better to bloc the acce of a uer to the networ than having to drop already active uer. Such deciion are made by the admiion control routine. In thi thei, three different admiion control algorithm are tudied and compared. The aim i to compare the efficiency of the different algorithm, which i why only peech uer are conidered. If more ervice are included, the imulation become very complex and time conuming. Mot of the imulation are done with a traffic intenity, which give 10% dropping when there i no admiion control, i.e. all uer are admitted. Dropped mean that the effect from the mobile i lower than required and the ervice eion i interrupted. The imulation clearly how that the admiion control trategie improve the ytem performance by decreaing the dropping of call. A expected, ome of the new uer are bloced. The ta for thi thei ha been to tudy and implement new complex admiion control algorithm to compare them to more imple algorithm lie The Equivalent Bandwidth method. 2

3 Content ABSTRACT... 2 CONTENTS INTRODUCTION PROBLEM DEFINITION THESIS OUTLINE THE THREE GENERATIONS OF TELEPHONY A SHORT HISTORY OF UMTS THE UMTS VISION UMTS SERVICES AND APPLICATIONS UMTS QOS CLASSES CDMA TECHNIQUE FDMA TDMA CDMA WCDMA SPREADING AND DESPREADING RADIO NETWORK SYSTEM RADIO PROPERTIES PROPAGATION PATH LOSS SHADOWING AND SCATTERING MULTI-PATH PROPAGATION INTERFERENCE THE CELL MODEL MATHEMATICAL BACKGROUND IN UMTS THE UPLINK SINGLE CELL POLE CAPACITY NUMBER OF USERS ADMISSION CONTROL ADMISSION CONTROL PRINCIPLE SYSTEM MODEL PROPAGATION MODELS TRAFFIC MODEL FOR SPEECH HANDOVERS Macro diverity POWER AND RATE CONTROL (PARC) CALL DROPPING MECHANISM PERFORMANCE MEASURES THE PRINCIPLE OF THE SIMULATOR ADMISSION CONTROL ALGORITHMS UPLINK LOAD FACTOR WIDE BAND POWER-BASED ADMISSION CONTROL STRATEGY Approximating the rie in interference by uing differentiation Approximating the rie in interference by uing integration

4 Equivalent Bandwidth algorithm DERIVATIVE ALGORITHM WITH EXPONENTIAL EQUALIZATION START APPROXIMATIONS SIMULATION RESULTS SYSTEM LOAD WORKING POINT ADMISSION CONTROL PILOT STUDY SIMULATIONS DISCUSSION AND CONCLUSION FURTHER STUDIES REFERENCES APPENDIX A. ABBREVIATIONS APPENDIX B. STARTING VALUES APPENDIX C. PLOTS AND DATA

5 1. Introduction There are everal way to end digital information through the ether. The main technique ued in mobile communication are Frequency Diviion Multiple Acce (FDMA), Time Diviion Multiple Acce (TDMA), and Code Diviion Multiple Acce (CDMA). In FDMA the available frequency interval i divided into band, which are aigned to individual uer. Each uer can then tranmit in hi band without ignificant interference from other uer. In TDMA, time i intead divided into o-called lot. A uer who i aigned a lot can tranmit freely in it with all the ytem reource devoted to him. The GSM ytem ue a combination of FDMA and TDMA. In CDMA all uer hare the ame frequency band and time lot and are identified by their orthogonal ignal. Thi olution mae the ytem more flexible but may require more bandwidth for it tranmiion. The next generation ytem, the o-called Univeral Mobile Telecommunication Sytem (UMTS) will be baed on CDMA technique. In fact, to cope with the demand for high bit rate, the allocated bandwidth for UMTS i very large (about 5 MHz). Therefore one often refer to UMTS a a Wideband CDMA (WCDMA) ytem. A CDMA ytem ha in principle no abolute limit on the number of uer it can upport. Intead it i the interference from other uer in the ytem that limit the number of uer. Thi mean that in order to maintain good quality of already ongoing connection, the networ need to etimate the amount of interference new uer bring with them. If a new uer will caue the ytem to be overloaded, he mut not be allowed to connect to the networ. Thi i the purpoe of the admiion control in CDMA ytem. There are many different principle for admiion control in the literature, both imple and advanced. It i neceary, however, to implement ome ind of admiion control, epecially when the traffic will be a mix of ervice with different demand on data rate and delay. In thi thei we will tudy ome propoed admiion control routine in a imulated radio networ. The aim of the thei wor i decribed in more detail in the next Section. 1.1 Problem definition A tated previouly, the capacity of a WCDMA networ i limited by the amount of interference in the ether, yielding a o-called oft capacity limit. Thi i in contrat to a ocalled hard capacity limit, which i et by the available amount of hardware. If the load of the ether i allowed to increae exceively, the quality of ervice for exiting connection can no longer be guaranteed. In practice thi mean that the interference from other uer become too big. Thu, before admitting new uer into the ytem, an etimation of the increae in interference need to be done to enure that the new uer will not eriouly affect already exiting connection. Thi ta i performed by the networ admiion control routine. If the algorithm ued are not effective or not optimized, it will reflect badly upon the operator. To develop a deep competence about admiion control routine i thu very important to an operator. The objective of the thei are a follow. Achieve a qualitative undertanding of the function of a WCDMA networ lie UMTS. 5

6 Implement a number of relevant admiion control algorithm in a imulation environment and tudy their performance in relatively realitic cenario. Thi hould initially be done for imple traffic model. 1.2 Thei outline Chapter 1 contain a hort hitory of telephony and a hort decription of the main goal of UMTS. Ch. 2 i a hort preentation of the expected ervice in UMTS. Ch. 3 preent the principle of CDMA, while Ch. 4 decribe WCDMA in more detail. The phyical networ ytem i hown in Ch. 5 and radio propertie are conidered in Ch. 6. Ch. 7 decribe the mathematical bacground, while ch. 8 i a hort decription of admiion control. Ch. 9 decribe the ytem model, while Ch. 10 how the principle of the imulator. In Ch. 11 the mathematic algorithm ued for the admiion control are preented. In Ch. 12, the imulation reult are collected. The thei end with a dicuion and concluion in Ch The three generation of telephony The hitory of mobile telephony i uually divided into three generation, ee Figure 1. The firt generation, which tarted in the middle of the 20 th century and i called the telephony era, upported only peech ervice. It exited during the time from the innovation to about An example of a firt generation ytem i NMT, which actually i till in operation. In the beginning of the 1990, the econd generation wa introduced. Firt generation Second generation Third generation Figure 1. The evolution of mobile ervice. The econd generation i called the telephony-and-data era and upported ervice lie peech, voice mail, Short Meage Service (SMS), Local Area Networ (LAN) acce, information ervice, internet/intranet acce and image download. GSM (Global Sytem for Mobile communication) i an example of a econd generation ytem. The poibility to end pacet data in a GSM networ wa introduced through GPRS (General Pacet Radio Service), which i tarting to become available to the public during Sometime GSM together with GPRS i called a 2.5 generation networ. GSM/GPRS ha evolved into GSM EDGE (EnhanceD rate for GSM [or ometime Global] Evolution), which can upport higher bit rate than GSM/GPRS, but i till quite limited. Thu, in order to upport really high bit rate it wa decided that a new mobile ytem hould be developed, at leat within Europe. Thi i the third generation ytem. The third generation ervice are almot the ame a thoe in the econd generation. Some new ervice are high peed internet/intranet acce, video retrieval and video conferencing/ 6

7 video telephony. For an overview of new multimedia ervice in UMTS, ee [9]. In Europe the third generation ytem i nown a UMTS. It alo include GSM EDGE, which i nown GERAN (GSM EDGE Radio Acce Networ) within UMTS. UMTS will coexit with GSM/GPRS for quite a while, which mean that today GSM phone will not become uele a oon a UMTS i introduced. 1.4 A hort hitory of UMTS The tory of UMTS tarted in 1985 in the ITU (International Telecommunication Union) with a tudy on FPLMTS (Future Public Land Mobile Telecommunication Sytem). The main goal of FPLMTS were to achieve a truly mobile ytem with integrated peech and data ervice that would alo be uitable a a modern telecommunication infratructure for developing countrie. In Europe FPLMTS ha changed name to UMTS. Some major miletone in the development of UMTS were [2]: A global identification of an unued frequency pectrum for FPLMTS wa made by WARC (World Adminitrative Radio Conference) in The migration plan for evolving UMTS from GSM were etablihed in The radio interface between the uer and the bae tation wa choen in GPP (3 rd Generation Partnerhip Project) wa founded in Tet ytem running in Sweden and around the world in The firt et of UMTS pecification, the o-called Releae 99 of UMTS, wa approved in December The UMTS viion UMTS will provide the mobile phone uer with all ind of ervice from today peech to future ervice lie , Internet ervice, mobile video telephony and mobile multimedia conferencing. UMTS will be of great benefit to the whole of ociety including indutry and buinee, public ervice, educational intitution a well a private uer. The UMTS ytem i pecified and deigned to provide a flexible and future proof architecture baed on tandardized interface which will allow new ervice to be developed, both by operator and by third partie and offered by many different ervice provider. UMTS will mae it poible to provide the cutomer with all ort of combination of communication and information feature baed on different type of media component uch a audio, text, video and graphic [2]. 7

8 2. UMTS ervice and application The bet nown new feature of UMTS i the high uer bit rate. On circuit-witched connection the expected maximum bit rate i 384 bp, while on pacet-witched connection it i 2 Mbp. Higher bit rate mae it poible to introduce new ervice uch a video telephony and quic downloading of data from the Internet. At the tart of UMTS mot of the traffic will be peech, but later the hare of web browing and will increae. It i difficult to predict the pace at which the hare of web browing and will occur; traffic will move from circuit-witched connection to pacet-witched connection. At the tart of UMTS ervice not all of the Quality of Service (QoS) function (peech, video telephony, WEB and ), will be implemented, and therefore delay-critical application uch a peech and video telephony will be carried on circuit witched bearer. Later, it will be poible to upport delay-critical ervice uch a pacet data with QoS function. 2.1 UMTS QoS clae Application and ervice can be divided into different group, depending on how they are conidered. In UMTS four traffic clae have been identified a (ee, for example, [1]): 1. Converational cla: In thi cla the time relation between information entitie in the tream need to be preerved. There are alo trong demand on the delay. Application within thi cla are voice ervice, video telephony and video game. Data integrity i not alway a neceity. 2. Streaming cla: Here the time relation between information entitie alo need to be preerved, but the requirement on the delay are omewhat le tringent than for the converational cla. Application within thi cla are treaming multimedia lie radio and video. Data integrity may not alway be completely neceary. 3. Interactive cla: In thi cla the tranmiion of information follow a requetrepone pattern and the data i expected within a reaonable amount of time. The data integrity i important. The application within thi cla are mainly web related, lie browing and networ gaming. 4. Bacground cla: For application within thi cla the data integrity i important but there i no real requirement on the delay. The application are, for example, . The main ditinguihing factor between the different clae are data integrity and how delay-enitive the traffic i. In the initial phae of UMTS the converation and treaming clae will be tranmitted a real-time circuit witched connection over the WCDMA air interface, while the interactive and bacground clae are tranmitted a non-real time pacet data. 8

9 3. CDMA technique A dicued previouly, UMTS will ue WCDMA technique in which all uer hare the ame frequency band. Thi i only true for uplin and downlin eparately, though. Uplin refer to tranmiion from the uer to the bae tation, while downlin refer to tranmiion from the bae tation to the uer, ee Figure 2. In thi Section we decribe how thi technique wor in omewhat more detail. We begin, however, with a hort introduction to FDMA and TDMA. uplin downlin Figure 2. The definition of uplin and downlin tranmiion. 3.1 FDMA In FDMA, the entire available frequency range i divided into band, ee Figure 3. The technique i characteried by the continuou acce of the uer in a given frequency band without any ignificant interference from other uer. The main drawbac of FDMA i that when a uer i idle, hi hare of the bandwidth cannot be ued by anyone ele. Thi lead to a wate of capacity. FDMA i thu not very flexible; on the other hand it relie on the ue of proven technique. time f1 f2 f3 f4 frequency Figure 3. In FDMA the available frequency range i divided into eparate band, which are aigned to different uer. There i no iginificant interference between uer. The figure i taen from [5]. 3.2 TDMA In TDMA, time i divided into lot, ee Figure 4. Each time lot i pre-aigned to a uer. During uch a lot, a uer i allowed to tranmit freely, and the ytem entire reource are devoted to thi uer. The lot aignment are periodic, and each period i uually referred to a a frame. A uer could be aigned one or more time lot during a cycle thereby increaing hi capacity for tranmiion. The drawbac of TDMA i that each uer ha a fixed allocation 9

10 of a time lot whether or not it ha data to tranmit. Again thi i a wate of capacity, epecially ince many application tranmiion are quite burty. time frequency Figure 4. In TDMA time i divided into lot, which are allocated to different uer. Within uch a lot all ytem reource are devoted to one uer. The figure i taen from [5]. 3.3 CDMA In CDMA, all uer hare the ame frequency band all the time. Thi mae CDMA a much more flexible ytem. A uer i only allocated the reource he need, e.g. when he need to end omething he i aigned the neceary bandwidth, which i then releaed when the tranmiion i finihed. Thu, when he i idle the reource are available for other uer. To identify individual uer, orthogonal ignal in conjunction with matching filter at the receiving tation are ued, ee Figure 5. Each uer i aigned a particular code equence, which modulate the carrier frequency with the digital data modulated on top of it. Even when everal tation employ the ame code, the effect of interference i minimied by the ability of the receiver to loc onto one pacet while all other overlapping pacet appearing a noie. CDMA i imple to operate ince it doe not require any tranmiion ynchroniation between the mobile tation, but the throughput can be low. By throughput we mean the total number of bit tranmitted per econd. code time frequency Figure 5. In CDMA, uer hare the ame frequency range all the time. Intead they are identified by orthogonal code. The figure i taen from [5]. 10

11 3.4 Power control Power control i an eential feature of all CDMA ytem. A fat and accurate power control i abolutely neceary in the uplin to detect uer and avoid o-called near-far problem. In the downlin, power control i ued to decreae the interference but it i le critical for ytem operation. The near-far effect i preent when an interfering tranmitter i much cloer to the receiver than the intended tranmitter, ee Figure 6. Although the cro-correlation between code A and B i low, the reulting ignal from the correlation between the received ignal from the interfering tranmitter and code A can be tronger than the correlated ignal received from the intended tranmitter. The reult i that proper data detection i not poible. Receiver, code A Intended tranmitter, code A Interfering tranmitter, code B Figure 6. The near-far effect in a CDMA ytem. The tranmitter with code B mae it impoible for the receiver to hear the tranmitter uing code A if power control i not ued. The figure i taen from [2]. Even if the power control wor perfectly, ome power related problem might occur. One of them i caued by the coctail party effect, decribed by Claude Shannon in 1949 a: More and more people can come, and they would all pay equally, o to pea. If more people were there, gradually the noie level would increae on each channel. But everyone would till tal, even though it might be a pretty noiy coctail party by that time A long a bae tation and terminal can increae their output power thing wor fine. However, when thi i no longer poible the radio lin will fail. Therefore, it i very important to have good power planning in order to enure that the capacity and the coverage are table. 11

12 4. WCDMA WCDMA i a wideband Direct-Sequence Code Diviion Multiple Acce (DS-CDMA) ytem. The uer information bit are pread over a wide frequency band by multiplying the uer data with quai-random bit (called chip) derived from CDMA preading code. To upport very high bit rate, up to 2 Mbp, the ue of a variable preading factor and multicode connection i upported. The chiprate in UMTS i 3.84 Mcp, which lead to a carrier bandwidth of about 5 Mhz. In WCDMA every bit i coded into chip and the number of chip from one bit i called the preading factor. The inherently large carrier bandwidth of WCDMA upport high uer data rate, but alo highly variable data rate. Each uer i allocated frame of 10 m duration, during which the uer data rate i ept contant. Every ervice type ha it own bit rate and it i regulated every 10 m. 4.1 Spreading and depreading Two different type of code equence are ued in order to pread the ignal. They are referred to a channelization code and crambling code, repectively. The channelization code are o-called Orthogonal Variable Spreading Factor (OVSF) code, characterized by different preading factor (SF). The SF alo define the length of the code. For example, a OVSF code with preading factor 4 i four chip long, while a OVSF code with preading factor 256 i 256 chip long. Two OVSF code of the ame length, i.e. with the ame SF, are alway orthogonal. For example, two 4-bit code: ω { 0,1} X ( ω) { 1,1 } X ( 0) 1 X ( 1) 1 Y ( ω) { 1,1 } Y ( 0) 1 Y ( 1) 1 S X ( 0,0,1,1) ( 1,1, 1, 1) T Y ( 0,1,1,0) ( 1, 1, 1,1 ) S T [( 1,1, 1, 1) ( 1, 1,1,1 )] Every uer in the ytem i given a unique preading code, which eparate the ignal from other uer ignal. To receive the coded ignal, a depreading code i ued. The code for preading and depreading are the ame. All information ent in the ytem i coded a binary number. An example of a preading and depreading procedure when an OVSF code with SF 8 i ued, i hown in Figure 7. 12

13 Bit Chip 1-1 Data preading 1-1 Spreading code 1-1 Spread ignal Data * code depreading 1-1 Spreading code 1-1 Data Spread ignal * code Figure 7. Spreading and depreading, uing a preading factor of 8. The OVSF channelization code have preading factor, ee Figure 8, that vary from 4 to 256 in both uplin and downlin. An SF 512 may optionally be ued in the downlin. The orthogonality property give le interference between uer. The number of different code, and conequently the number of imultaneou uer of a certain ervice, i directly related to the preading factor: there are 256 different code with SF 256 but only four with SF 4. A peech connection baed on the 12.2 bp AMR (the peech bitrate of UMTS) mode require a SF of 128 wherea a 128 bp data connection need SF 16 to reach the QoS goal. 13

14 C ch,4,0 (1,1,1,1) C ch,2,0 (1,1) C ch,4,1 (1,1,-1,-1) C ch,1,0 (1) C ch,4,2 (1,-1,1,-1) C ch,2,1 (1,-1) C ch,4,3 (1,-1,-1,1) SF 1 SF 2 SF 4 Figure 8. The code-tree for generating OVSF code. In figure 8, the channelization code are uniquely decribed a C ch,sf,, where SF i the preading factor of the code and i the code number, 0 SF-1. Each level in the code tree define channelization code of length SF, correponding to a preading factor of SF in figure 8. The generation method for the channelization code i defined a (ee [14]): C ch,1,0 1, C C ch,2,0 ch,2,1 C C and in general we have: ch,1,0 ch,1,0 C C ch,1,0 ch,1, , C C C C C C ch, 2 ( n+ 1),0 ch, 2 ( n+ 1),1 ch, 2 ( n+ 1), 2 ch, 2 ( n+ 1),3 : ch, 2 ( n+ 1), 2 ( n+ 1) 2 ch, 2 ( n+ 1), 2 ( n+ 1) 1 C ch, 2 n C ch, 2 n C ch, 2 n C ch, 2 n : C ch, 2 n, 2 C ch, 2 n, 2,0,0,1,1 n 1 n 1 C C C C C C ch, 2 n,0 ch, 2 n,0 ch, 2 n,1 ch, 2 n,1 : ch, 2 n, 2 n 1 ch,2 n, 2 n 1. The leftmot value in each channelization code word correpond to the chip tranmitted firt in time. In the downlin, crambling code are ued to eparate different cell and ector from each other. All channel tranmitted from one bae tation are ynchronou at chip level and ymbol can be hifted by multiple of 256 chip to each other. Thi preerve the good correlation characteritic of the OVSF code (orthogonal in an environment of infinite bandwidth and no reflection). Since the ame crambling code i ued in the entire cell (ector) in downlin, eparation between uer i in the hand of the channeliation code. 14

15 In the uplin, all mobile ue the ame code et of OVSF code and they are therefore ditinguihed only by the crambling code, which i pecific to each uer. The purpoe of the channelization code in uplin i to eparate different channel from the ame UE (Uer Equipment) if neceary. There are about 2 24 crambling code. The propertie and generation of thee code are outide of the cope of thi thei. For further tudie ee [1], [2], [14]. 5. Radio networ ytem Here i a hort and imple decription of the general tructure of a third generation mobile networ ytem. It i very imilar to the econd generation networ. The bae tation and the radio networ controller are new feature. CN Core Networ GERAN GSM Edge Radio Acce Networ UTRAN UMTS Terretrial Radio Acce Networ MSC Mobile Switching Center SGSN Serving GPRS Support Node RNC Radio Networ Controller UE Uer Equipment GSM Global Sytem for Mobile communication GPRS General Pacet Radio Service RNS Radio Networ Subytem Node B Bae tation The MSC and the SGSN function lie witchboard for circuit witched and pacet witched data repectively. The RNC control baically all radio reource and, for example, adjut the ignal power ent from UE and Node B. 15

16 CN MSC SGSN GERAN GSM / GPRS RNC RNC Node B Node B RNS UE RNS UE UTRAN Figure 9. Phyical view of the networ ytem. The Core Networ i the already exiting phyical networ. UTRAN i the future phyical networ for UMTS and RNS i one bae tation erving it mobile. 6. Radio propertie A problem with radio i that the available pectrum i limited. The le pectrum needed per ubcriber the more ubcriber can be accommodated in the networ. There i a need for good modulation technique and efficient acce method to ue the ether properly. A radio ignal i affected by different phyical phenomenon on it way from tranmitter to receiver. Here we are going to preent the mot important phenomena. They are path lo 16

17 including hadowing, cattering and multipath propagation. We alo preent the meaning of interference. The cell model i alo decribed. The model decribing the propagation of radio wave in citie lie New Yor or Hong Kong differ greatly from thoe that decribe the propagation in the north of Sweden. Below we decribe ome of the effect that are relevant for path lo calculation. 6.1 Propagation path lo Path lo or attenuation of the ignal caue the received ignal to get weaer further away from the tranmitter. Path lo mae it difficult to get ufficient ignal trength level, but it alo reult in a lower interference from unwanted tranmitter far away from the receiver. Within the telecom buine we tal about path gain and path lo. Thee concept are defined a follow. P Path gain (6.1) received P tranmitted 1 Path lo (6.2) Path gain Path lo 2 d (6.3) Here i a contant and d the ditance between the tranmitter and receiver, ee Figure 10. The path lo itelf i dimenionle. It follow that Path lo a d and then P received 0 far away from the tranmitter. The definition of path lo may eem omewhat trange, but i common in telecommunication. d Figure 10. The path lo between tranmitter and receiver depend on the ditance d between them. A d, the received power will go to zero. Relation (6.3) aume a line of ight condition between tranmitter and receiver and that there are no reflection interacting with the direct radio wave. It i thu a very imple model. There are, however, more complicated model that tae more phenomena into account. In general, the exponent in (6.3) are not 2, but general contant determined by, for example, meaurement and approximation. The theoretical model are then compared to actual meaurement and contant, lie in (6.3) above, can be tuned. Correction are due to hadowing, cattering, and multipath propagation (ee Figure 11), which are decribed below. 17

18 6.2 Shadowing and cattering If the radio path doe not have free line of ight between tranmitter and receiver, the obtacle will caue hadowing. The mobile phone i normally located in a low poition and the tranmiion will mot liely be affected by hadowing object, for example building and hill. When the mobile phone move around, variation in ignal trength due to the character of the object can be meaured and contribute quite a lot to the variation in the path lo. The actual received ignal i often tronger than predicted. Thi i due to the fact that when the radio wave trie a rough urface, the reflected energy i pread out in all direction due to cattering, thereby providing additional energy at the receiver. Flat urface that have much larger dimenion than the wavelength can be modelled a reflective urface. In other cae the roughne of the urface induce propagation effect that mut be taen into account. For more detail on hadowing and cattering, ee [10]. The fading due to hadowing and cattering can be modelled a log-normally ditributed with a zero-mean and a tandard deviation of σ (ee [12]). 6.4 Multi-path propagation In an urban environment with a lot of reflecting object near the tranmitter and receiver another effect occur, called multi-path propagation. Since the tranmitter i not normally tranmitting directly toward the receiver but rather in a wide ector toward him/her, there will be a lot of ray reflected by obtacle and received by the receiver, ee Figure 11. Different reflection will have different time delay and effect on the phae of the radio wave. Normally a uer receive everal wave and the reulting wave can be very trong or very wea. If two wave with the ame frequency have a phae difference of 180 degree, they will completely cancel each other out. The phenomenon i called multi-path fading and can be quite ubtantial. Thi effect i not conidered in the imulator. 18

19 Figure 11. Multipath propagation. The effect can be quite ubtantial. 6.5 Interference Interference i the term for the unwanted ignal that the receiver experience. Mot interference i due to uer in adjacent cell. Reuing a frequency in different cell i limited by co-channel interference. Co-channel interference decribe the relation between the deired ignal effect and the undeired ignal effect, both uing the ame carrier frequency. Figure 12 how the principle of interference between two bae tation. The mobile want to liten to the ignal C from the carrier bae tation but alo receive interference I from the interfering bae tation. The larger the interference I i, the harder it will be for the uer to ditinguih the deired ignal C. Eventually it will become impoible. Carrier Interferer C I Figure 12. Interference. The interfering bae tation mae it harder for the mobile to liten to the ignal from the carrier bae tation. Although the channelization code are orthogonal in theory, a conequence of the effect decribed i that the waveform will not be perfectly orthogonal when received. They can thu impact each other. Since all uer occupy the ame part of the pectrum, the effective noie will be the um of all other uer ignal. The receiver correlate it input with the deired 19

20 noie, enhancing the ignal to noie ratio at the detector. Becaue the interference i ummed, the ytem i no longer enitive to wort cae interference, but rather to average interference. 6.6 The cell model The bae tation and their coverage area can be modelled a hexagon tructure with the tation in the middle. There are two type of cell: omnidirectional antenna cell and multiple ector ite cell, ee Figure 13. Omnidirectional antenna cell have the bae tation in the center of the hexagon. Multiple ector cell have the bae tation located at the border of the cell. Here we preent the multiple ector ite a a three ector ite. bae tation Omni directional cell Three ector cell Figure 13. The omnidirectional antenna cell model and the three ector antenna cell model. Only omnidirectional cell are ued in the imulation. In thi thei only omni-directional cell are modelled. A an approximation of the bae tation in a region the hexagon are placed cloe to each other in a regular pattern, ee Figure 14. In reality thi i not the alway the cae due to hadowing and multi-path propagation. 20

21 Figure 14. A regular cellular ytem in theory. Typical cell diameter for peech in UMTS with three ector ite lie between 1800 m and 1100 m for the up lin and 3500 m and 1500 m for the down lin, depending on the load of the cell. Thi how that it i often the uplin that i the limiting lin, ee, for example, the imulation reult in [12]. It i well nown that the traffic load will affect the coverage of an UMTS cell. Thi i ometime referred to a cell breathing, becaue the cell hrin with increaing load and grow with decreaing load. Thi effect i inherent in all CDMA ytem and it ha to be reconed with when planning the networ. Thi add complexity to the problem of optimiing the coverage and the capacity of the ytem. In FDMA/TDMA ytem uch a GSM, the coverage i relatively independent of the capacity, which mae planning the coverage of the networ much eaier than in CDMA ytem, where we have to tae into account the interference from other uer already at low traffic load. 7. Mathematical bacground in UMTS A tated previouly, we focu on the uplin ince it i the limiting lin in mot cae. Thi i due to the relatively low output power of the mobile. In thi Section we preent ome mathematical formula and expreion that are important when tudying the capacity and coverage of a WCDMA networ. One can do baically the ame analyi for the downlin cenario a well. The mathematic will be very imilar. 21

22 7.1 The uplin In the uplin, we receive the ignal from all the UE at the ame location, viz. at the bae tation antenna. It i alo here that interference from other cell will be received. The limiting factor in mot cae will be the maximum output power of the UE, becaue it ha to overcome noie, other cell interference (intercell) and interference from other uer within the cell (intracell) at the bae tation antenna. The quality of a radio lin can be characterized by it Signal-to-Interference Ratio (SIR) value. Thi i baically the quotient between the effect of the deired ignal and the interference, a received at the bae tation. One hould note, though, that meauring thee quantitie in real life can be quite complicated. A ervice, lie peech ha ome neceary SIR value for it to wor properly. Thee are found through imulation and meaurement in tet ytem. The ervice alo ha a neceary bit rate, which i determined by how much information it need to convey. Thi i a nown value, which we denote by R. Speech in UMTS, for example, need a bit rate of 12.2 bp. G i b() i i G b () i G b ( ) Figure 15. Path gain between mobile and bae tation. 22

23 The received ignal power i Gib() i Pi, while the total amount of interference i N 0 + G b P, i.e., the bacground noie N 0 together with the um of all other uer i () i ignal trength at the bae tation. P i i the tranmitted power from uer i. G b(i) i the path gain between uer and the bae tation uer i i connected to, ee Figure 15. Here b(i) i hort-hand notation for the bae tation uer i i connected to. We can now write down the equation that decribe the connection between the SIR value and power, the o-called power control equation. For uer i, it i (ee [12]) SIR i W R i N o + G ib ( i ) ( G b ( i ) P ) i P i. (7.1) There i one uch equation for each uer in the ytem, which mut be olved for the power P i. W i the chiprate 3.84 MHz. To implify the calculation we introduce the equivalent bandwidth η : η R SIR. (7.2) i i i The equivalent bandwidth can be viewed a ome ort of meaure of the amount of radio reource the ervice will need ince it i directly proportional to both the bit rate and the neceary SIR. Equation (7.1) and (7.2) now yield ηi N W o + ( Gb( i) P ) Gib( i) Pi i (7.3) ( Gb i) P ) ( Gb( i) P ) Gib() i Pi i Equation (7.3,7.4) give: (. (7.4) ηi N o + () i ) W ηi W N N G η ( Gb P ) Gib() i Pi Gib( i Pi ( G P ) G () P + i () + i o b i ib i i 1 W W N Pi G ib o ib o ηi W η G 1 + W () i i ib() i ηi + 1 W + η G () i i ib() i ηi W + ηi η ηi 1 W η + i 1 W ( Gb() i P ) ( Gb() i P ). (7.5) 23

24 Thi can be rewritten in matrix form according to: Ρ Α Ρ + Β, (7.6) Ρ min i { 1... N} N 1 ( ( Ι Α) Β), P ) i max i,, number of uer. (7.7) Here we have introduced P max a the maximum output power of a UE. In UMTS, P max i W. Equation (7.7) thu decribe the fact that a UE can not tranmit with higher power than P max. The matrice A and B are defined according to Α i η G i W + ηi G b ib () i () i (7.8) Β i ηi N W + ηi Gib o () i (7.9) { 1... N} N i,, number of uer One big problem i that it i not effective to compute the invere of a very large matrix. In thi tudy it i ueful to mae iteration to olve thi ytem numerically. The iteration proce i continued a long a the relative difference between two conecutive value i larger than ξ P ( n) ( n 1) i P P n i i ξ (7.10) 7.2 Single cell pole capacity The pole capacity i the capacity when the bacground noie rie, the total interference to thermal noie ratio ( ee page 33), diverge toward infinity. We will now focu on a ingle cell urrounded by ome tatic interfering cell in order to undertand the pole capacity better. In thi ituation the power control equation can be olved analytically. The olution i (ee [12] for detail), P i λ W η i ( N + oci) ( W + η ) i 1 W + η o η. (7.11) where λ i the number of uer per unit time and oci i the other to own (tatic) cell interference ratio. We can now define the cell load X according to 24

25 X ηi. (7.12) W + η i i Equation (7.11) can now be written a P i λ W ηi ( N o + oci) ( W + η )( 1 X ) i. (7.13) If too many uer are allowed into the ytem, o that X>1, the ytem will collape. The output power need to be infinite in order for all uer to fulfil the SIR requirement. In thi analyi, the variou ervice that UMTS provide are differentiated through their different η value. The increment of the load when adding one uer of ervice number will be X η (7.14) W + η If we have only one ervice preent in the cell characterized by η, the maximum number of uer in the cell i max W K + 1, (7.15) η which follow from equation (7.12) and the fact that X1 define the maximum cell load poible. Since the maximum number of uer i roughly W/η, η ha a role imilar to the channel bandwidth in a FDMA ytem, which explain the name equivalent bandwidth. Intead of taing the um over all uer in equation (7.12), we can group the uer according to the ervice they are uing. Then the load become X η K W + η, (7.16) where K i the number of uer in the cell uing ervice number. 7.4 Number of uer It i time to introduce three level of activity for the uer: inactive, in eion and active. The inactive uer are of no interet to u other than a a pool of uer that may tart a eion. The uer who are in eion are currently uing the ervice, which doe not necearily mean that data i being tranmitted over the radio interface. Thi occur with a certain probability ν depending on the type of ervice. The number of active uer i thu alway maller than or equal to the number of uer in eion. Thi model i baed on the behaviour of peech uer, where the probability of activity i called voice activity factor and ha a value around 50 %. Alo, the ditribution of the number of uer in eion, M, i needed in order for the model to be complete. It i well nown that the inter arrival time and the call holding time for peech are exponentially ditributed, leading to a Poion ditribution for M [8]. When 25

26 26 tudying tatitic from meaured traffic load in the GSM networ, the traffic in GSM i found to be Poion ditributed; it i therefore reaonable to chooe the Poion ditribution in thi model. The Poion ditribution ha one parameter, referred to a the eion traffic, denoted by ρ, which i equal to the mean of the ditribution. Thi parameter i referred a the eion traffic, denoted ρ. Thi give ( ) e m m M P m ρ ρ!. (7.17) The ubcript i for ervice type. Given a certain number of uer in eion, the number of active uer, K, will follow a binomial ditribution: ( ) ( ) m m m M K P ν ν 1, (7.18) which give u ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) [ ] ( ) ( ) ( ) e e e e for formula Taylor m e m e m e m e m m m e m M K m P M P K P x m m m m m m m m m m m m m m m ν ρ ν ρ ρ ρ ρ ρ ρ ρ ρ ρ ν ρ ν ν ρ ρ ν ν ρ ρ ν ν ρ ν ν ρ ν ν ν ρ +!! '! 1!! 1!! 1!! 1! 1!!!! ) ( (7.19) Here we have tudied [6] to mae thi derivation. The reult i that the number of active uer alo follow a Poion ditribution with mean ρ ν, i.e., ( ) Po K ν ρ. (7.20)

27 8. Admiion control CDMA ytem have no abolute limit on the number of uer that can be upported in each ector. Thi number i determined by the multiple acce interference that i generated at the bae tation by all the uplin ignal and by the propagation condition (path lo and hadowing). In order to maintain the quality of the radio connection an admiion control i needed to prevent the ytem from getting overloaded. If the interference level get too high no new uer are admitted. The increae in the interference level caued by an incoming uer need to be etimated by the admiion control. Multiple acce interference in WCDMA can be approximated a Gauian noie but it inherently conit of received ignal of WCDMA uer. Therefore, multiple acce interference i highly tructured, and can be taen into conideration in the receiver to improve the performance. The admiion control allow a new uer into the radio acce networ if the admiion doe not caue an exceive interference in the ytem. Thu, the admiion control ha a big reponibility for the tability and high capacity of the WCDMA networ. The admiion control i performed for the uplin and downlin tranmiion eparately becaue the traffic load can be aymmetric. A uer i admitted into the ytem if both uplin and downlin admiion control requirement are fulfilled. In thi Thei only the uplin i conidered (ee, for example, [3] for downlin admiion control imulation). The baic admiion control procedure i decribed in the next Section. 8.1 Admiion control principle. Newly arrived real time uer will be controlled by an admiion control function to determine whether there i enough capacity for them to be accepted. The procedure for the new uer in every cell i a follow: 1. New uer in each cell will be regitered in the current admiion control function. 2. The already active real time uer and the new uer will be taen into account in the calculation of the cell load. 3. If the new uer i accepted, he/he will be included among the uer. If bloced, the new uer will of coure be rejected to the cellular ytem. 4. If there are other new uer, tep 2 and tep 3 will be repeated. 5. The procedure will finih when there are no more propective uer in the cellular ytem. 27

28 9. Sytem model The computer imulation are done in an environment of 16 hexagonal cell with a cell radiu of 500m. A wraparound technique i ued to avoid border effect. We have a bae tation placed in the middle of the cell and omni-directional antenna are ued. 9.1 Propagation model The received power of the ignal Prj j can be expreed a from tranmitter (mobile uer) i at receiver (bae tation) P rj G P, (9.1) ij ti where r tand for received and t for tranmitted. The path gain G i defined a G 1 (9.2) L S where L i the propagation path lo and S i the hadow fading. The path lo i baed on the Oumura-Hata formula L db δ + ε log( d ), (9.3) δ +ε log 10 ( d ) L 10 (9.4) whereδ i an attenuation contant baed on the carrier frequency (2 GHz), the height of the tranmitting and receiving antenna repectively.ε i the ditance attenuation coefficient with a value of Note that (9.3) of the ame form a in equation (6.3), but with other exponent and in db. Here d i the ditance between a bae tation and a mobile uer. The hadow fading S i log-normally ditributed with a zero-mean and a tandard deviation of σ. When computing different quantitie in radio ytem, the magnitude are often o mall that we ue db intead of it real phyical value to implify when doing the calculation. The relation between real value and value in db i b a (9.5) where a i the real value and b i the value in db. 9.2 Traffic model for peech In GSM, peech i the dominant ervice and will be o alo in UMTS. The ervice can be of different qualitie. The characteritic of peech are low delay and low variation in the delay. The SIR-requirement for peech i 7.9 db (ee [7] and [12]). A contant bitrate of 12.2 bp i needed and the ervice i ymmetric in uplin and downlin. The voice activity i 0.5, which 28

29 mean that we pea approximately half of the time during a call. Arrival to the ytem are modelled a a Poion proce. The mean time of call i 90, ee [8]. 9.3 Handover One of the fundamental RRM (Radio Reource Management) function in the radio networ i different type of handover. For an overview of radio reource management function in UMTS, ee [13]. A handover i performed when the mobile i connected to a bae tation and move and change to another bae tation. If the ditance between a mobile uer and a bae tation i very large, a dropping can occur. By uing handover, the radio lin can be aved and the QoS requirement upheld. There are mainly three type of handover: 1. Hard handover; the lin to the old bae tation i diconnected before a lin to the new bae tation i et up. The mobile uer can thu have only one radio lin connection with one bae tation at a time. 2. Soft handover; the UE can be connected to two (or more) bae tation at the ame time. After moving further away from the old bae tation, the radio lin i diconnected and the UE continue with the new bae tation. 3. Softer handover; baically the ame a oft handover. The only difference i that the handover are between two ector on the ame ite Macro diverity A diverity method called macro diverity i ued in ofter handover. The diverity technique tranport the ame information along a number of communication path between the mobile uer and the bae tation. The mobile uer will dynamically chooe the tranmitter with the bet communication path with a combining method. It i a ueful method to combat the hadow fading and other terrain effect. By uing a diverity technique the performance of data tranmiion over fading channel can be improved and the hadowing of tranmitted ignal by large obtacle uch a mountain, hill or building can be avoided. The ue of oft/ofter handover decreae the capacity of a networ in term of the number of uer it can upport. Therefore it i not adviable to have too many uer in oft/ofter handover. In the imulation, a hard handover trategy i ued. A uer can only be connected to one bae tation at a time. 9.4 Power and rate control (PARC) By uing power control, SIR can be efficiently controlled. If variable tranmiion rate are available we can efficiently utilize the radio pectrum. Due to the different QoS requirement in UMTS, the need for power control and rate control i obviou. It i natural to ue a combination of thee RRM (Radio Reource Management) function, power control and rate control. For real time ervice a tolerable minimum average rate mut be guaranteed, while delay inenitive bet effort ervice may temporarily reduce their tranmiion rate and attempt to utilize any exce networ capacity. If the cellular ytem become heavily congeted, the tranmiion rate of bet effort ervice may be lowered even to zero. The effective 29

30 tranmiion rate are cloely related to the SIR. If the data rate of one uer i increaed, the other uer will experience an increaed interference power for a hort time. The other uer will experience le interference with a decreaed data rate, but for a longer time. 9.5 Call dropping mechanim A dropping condition, Prequiremen t > Pmax i ued but the call do not drop immediately. By uing a leay bucet technique, the call will be dropped after a number of tranmiion failure. After a failed tranmiion the bucet will decreae by one unit and when there i no unit left in the bucet, dropping will occur. After a ucceful tranmiion the bucet will increae by two unit, but only up to a given threhold. A threhold of 1 unit i ued for real time ervice lie peech correponding to : : : : N : : : : After a ucceful tranmiion the bucet will increae by two unit After a failed tranmiion the bucet will decreae by one unit Figure 17. The tructure of the leay bucet. 9.6 Performance meaure The ervice quality i normally determined by the probability of blocing and dropping. Blocing occur when a newly arrived uer i denied acce to the cellular radio ytem. Dropping occur when an active uer i diconnected from the cellular radio ytem. There are everal reaon why dropping occur. An active uer may move into another cell, where there i no capacity available or SIR may drop below the minimum SIR requirement. Dropping of an ongoing ervice i in general more annoying than blocing a new uer. To maintain a high QoS, the bet trade-off between thee two ervice quality meaurement mut be found. A lower blocing probability will reult a higher dropping probability and vice vera. The dropping probability P d i defined a the ratio between the number of dropped eion and the number of arrival to the ytem during the imulation time. The blocing probability P b i defined a the ratio between the number of bloced eion and the number of arrival to the ytem during the imulation time. 30

31 10. The principle of the imulator The imulator i a dynamic imulator implemented in MATLAB. Each imulation i done a follow. 1. A log-normal map i created to model the effect of hadowing and cattering. 2. Initial ervice eion and uer are created. 3. Simulation loop: Create new active uer (mobile) and tart new ervice eion, according to a Poion proce. Create new in-eion load. Different load for every ervice type. Allocate radio reource to active ervice eion. Admiion control. Handover. Power control. Calculate tranmiion quality (received power and SIR level). t t + dt Compare tranmiion quality with quality demand. End dropped or completed ervice eion. Collect tatitic. Remove completed and dropped ervice eion. Move active uer. Update poition of uer. Add new uer if there are any. Figure 18. Principle of the imulator. The imulation end when t t(end of ervice eion). A typical value of t(end of ervice eion) i 2000 and dt 0.1. In the imulation program, the mobile are modelled to move every time tep (dt) and their new poition (xy), xy i a complex vector, depend on their velocity (vel) and acceleration (acc) at the time tep n-1, where n correpond to the current time tep. The movement of the uer in each time tep i calculated a follow. 31

32 The acceleration i given by acc n φ r γ vel (10.1) n n where φ and γ are contant and r i a normal ditributed random number. The uer velocity then follow a veln +1 veln + accndt. (10.2) Thi lead to the new uer poition xyn +1 xyn + velndt. (10.3) Statitical propertie of acc and vel are aumed (e.g the average of all mobile uer velocity i aumed to be Rayleigh ditributed) and then φ and γ can be expreed a: σ φ acc (10.4) dt 2 σ acc 1 1 dt σ vel γ (10.5) dt where acc mean and vel mean are the input variable to thi mobility model. For further detail ee [15] and [16]. 11. Admiion control algorithm Here we preent three algorithm, viz. The Derivative, The Integration and The Equivalent Bandwidth algorithm. The Derivative and The Integration method are decribed in [1] and [4]. The Equivalent Bandwidth method i decribed in [11]. We mut derive formula for the increae in load and interference. We mut alo find out value of the threhold. In Section 11.1 the load factor for The Derivative and The Integration algorithm i preented a an equivalent bandwidth. In Section and Section the increae in interference level i derived for The Derivative and The Integration method repectively. In Section the load for The Equivalent Bandwidth algorithm i preented. Start approximation for the threhold are computed Uplin load factor The firt tep i to preent the ACSIR (Admiion Control SIR) for The Derivative and The Integration algorithm. The only difference from equation 7.1 i that the voice activity level i included and that the power and interference doe not include the path lo. The reaon that 32