Institutionen för systemteknik

Transcription

1 Institutionen för systemteknik Department of Electrical Engineering Examensarbete Uplink TDMA Potential in WCDMA Systems Examensarbete utfört i Reglerteknik vid Tekniska högskolan i Linköping av Markus Persson LITH-ISY-EX--08/4085--SE Linköping 2008 Department of Electrical Engineering Linköpings universitet SE Linköping, Sweden Linköpings tekniska högskola Linköpings universitet Linköping

2

3 Uplink TDMA Potential in WCDMA Systems Examensarbete utfört i Reglerteknik vid Tekniska högskolan i Linköping av Markus Persson LITH-ISY-EX--08/4085--SE Handledare: Examinator: Rikard Falkeborn isy, Linköpings universitet Erik Geijer Lundin Ericsson AB Fredrik Gunnarsson isy, Linköpings universitet Linköping, 31 January, 2008

4

5 Avdelning, Institution Division, Department Division of Automatic Control Department of Electrical Engineering Linköpings universitet SE Linköping, Sweden Datum Date Språk Language Svenska/Swedish Engelska/English Rapporttyp Report category Licentiatavhandling Examensarbete C-uppsats D-uppsats Övrig rapport ISBN ISRN LITH-ISY-EX--08/4085--SE Serietitel och serienummer Title of series, numbering ISSN URL för elektronisk version Titel Title Potential med tidsdelad uplänk i WCDMA system Uplink TDMA Potential in WCDMA Systems Författare Author Markus Persson Sammanfattning Abstract The evolvement of the uplink in the third generation mobile telecommunication system is an ongoing process. The Enhanced Uplink (EUL) concept is being developed to meet the expected need from more advanced services, like video streaming and mobile broadband. One idea for further improvement in the EUL concept is to introduce Time Division Multiple Access (TDMA), which is studied in this master thesis. The master thesis assignment is to study the consequences of introducing TDMA in EUL. The goal has been to identify the gains and problems, and how they can be handled. A derived theoretical framework and system simulations, using a radio network simulator, are used. The overall conclusion is that there is a potentially large gain with an introduction of TDMA in EUL. Simulations in favorable conditions have shown that the system throughput can increase by 100% when there are only User Equipment (UE) that are using EUL in the system and by 50% when there is a mix of speech and EUL UE s. When using TDMA the uplink load also shows improvements, the mean is generally higher but the variance is generally smaller. Due to major differences in experienced interference between passive and active UE s, the signal quality will very a lot. The big variation in signal quality is identified as the main problem with introducing TDMA in EUL. It is shown that this problem can generate extreme high uplink load, which have a negative impact both on the resource efficiency and the coverage. Nyckelord Keywords 3G, UMTS, WCDMA, TDMA, EUL, Enhanced Uplink

6

7 Abstract The evolvement of the uplink in the third generation mobile telecommunication system is an ongoing process. The Enhanced Uplink (EUL) concept is being developed to meet the expected need from more advanced services, like video streaming and mobile broadband. One idea for further improvement in the EUL concept is to introduce Time Division Multiple Access (TDMA), which is studied in this master thesis. The master thesis assignment is to study the consequences of introducing TDMA in EUL. The goal has been to identify the gains and problems, and how they can be handled. A derived theoretical framework and system simulations, using a radio network simulator, are used. The overall conclusion is that there is a potentially large gain with an introduction of TDMA in EUL. Simulations in favorable conditions have shown that the system throughput can increase by 100% when there are only User Equipment (UE) that are using EUL in the system and by 50% when there is a mix of speech and EUL UE s. When using TDMA the uplink load also shows improvements, the mean is generally higher but the variance is generally smaller. Due to major differences in experienced interference between passive and active UE s, the signal quality will very a lot. The big variation in signal quality is identified as the main problem with introducing TDMA in EUL. It is shown that this problem can generate extreme high uplink load, which have a negative impact both on the resource efficiency and the coverage. v

8

9 Acknowledgments I would like to thank all at the Radio Resource Management department at Ericsson for a great master thesis experience. I especially want to thank my supervisor at Ericsson AB Erik Geijer Lundin for always take the time to answer all my questions, discuss my work or give me good advice. I would also like to thank my examiner Fredrik Gunnarsson and my university supervisor Rikard Falkeborn for showing interest in my work and making useful comments on my report. Finally a big thank to my family and friends for all the support given during the autumn. Stockholm, January 2008 Markus Persson vii

10

11 Abbreviations 3G 3GPP ACK ARQ BLER CDMA CIR CDF CN CQI db dbi DCH DPCCH DPDCH E-DCH E-DPCCH E-DPDCH EUL FDD FDMA HARQ HSDPA HSPA HSUPA IC ILPC kbps Mbps NACK NBI NMT OLPC OVSF 3rd Generation mobile communication system 3rd-Generation Partnership Project Acknowledgement Automatic Repeat Request Block Error Rate Code Division Multiple Access Carrier-to-Interference Ratio Cumulative Distribution Function Core Network Channel Quality Information Decibel Decibels over Isotropic Dedicated Channel Dedicated Physical Control Channel Dedicated Physical Data Channel Enhanced Dedicated Channel Enhanced Dedicated Physical Control Channel Enhanced Dedicated Physical Data Channel Enhanced Uplink Frequency Division Duplex Frequency Division Multiple Access Automatic Repeat Request High-Speed Downlink Packet Access High-Speed Packet Access High-Speed Uplink Packet Access Interference Cancellation Inner Loop Power Control Kilobit per second Megabit per second Negative Acknowledgement Narrow Band Interference Nordic Mobile Telephony Outer Loop Power Control Orthogonal Variable Spreading Factor ix

12 x PIC PSTN RBS RNC RoT RTT SF SIC SIR TDD TPC TTI TDMA UE UMTS UTRAN WCDMA WBI Parallel Interference Cancellation Public Switched Telephone Network Radio Base Station Radio Network Controller Rise over Thermal Round-Trip Time Spreading Factor Successive Interference Cancellation Signal to Interference Ratio Time Division Duplex Transmit Power Control Transmission Time Interval Time Division Multiple Access User Equipment Universal Mobile Telecommunication System UMTS Terrestrial Radio Access Network Wideband Code Division Multiple Access Wide Band Interference

13 Contents 1 Introduction Background Related Work Problem Statement Approach Thesis Outline Cellular Radio Systems History Propagation Distance Path Gain Shadowing Fading Fast Fading Antenna Gain Receiver Antenna Rake G-Rake Multiple Access FDMA TDMA CDMA Cellular System Uplink and Downlink Power Control UMTS WCDMA Network Architecture Uplink Power Control Interference Cancellation EUL EUL Channels Short TTI Scheduling Hybrid ARQ xi

14 xii Contents 3 Simulator Model Definitions Shannon s Theorem Noise Rise Interference Cancellation Propagation Receiver Antenna Network Layout Traffic Models Upload with EUL Upload without EUL Speech Mobility Synchronization UE Power Power Control Hybrid ARQ TDMA Interference Cancellation EUL Scheduling Delays System Logging Results CIR Overshoot Theoretical Simulations Performance Improvement Evaluation Theoretical Results Power Rushes Interference Compensation Simulations Minimum Timeslot Length Simulations Interference Cancellation Minimum Timeslot Length with SIC System Analysis TDMA Combined with CDMA Conclusions Further Work Bibliography 49 A CDF 51

15 Contents xiii B Decibel 52 B.1 db B.2 dbw and dbm B.3 dbi C E-DCH UE Categories 53

16

17 Chapter 1 Introduction During the past years, there has been a tremendous growth in both the number of users and amount of traffic in cellular systems. Today, the most widespread mobile communication is digital cellular, with more than 3 billion users, which is almost half of the world s population. The ordinary voice service is still the most important service that mobile users rely on and expect high quality from, but new services such as mobile broadband are growing stronger, with new demands on capacity. 1.1 Background Since the first release, Rel 99, of the third generations mobile telecommunication systems were introduced, the evolvement has continued. WCDMA Release 5 introduced the improved downlink, High-Speed Downlink Packet Access (HSDPA), with better resource utilization thus higher bit rates. The natural step with Release 6 was the introduction of the improved uplink, High-Speed Uplink Packet Access (HSUPA) or Enhanced Uplink (EUL). The algorithms used in EUL have since the introduction been improved. They have been improved to the extent that it is hard to gain much more with the use of the same technique. To further increase the capacity in the uplink a radical change has to be done. One promising idea to increase the capacity and improve EUL further is to change the access scheme, from Code Division Multiple Access (CDMA) to Time Division Multiple Access (TDMA) 1.2 Related Work Since TDMA in EUL is a fairly new concept there are not yet many studies published. In most studies where TDMA are compared to CDMA, there is only low 1

18 2 Introduction bit rate users, such as speech users, instead of high bit rate users which is more relevant in this study. Those studies often compare delays and the maximum expected number of users that can be admitted. In study [10], there is shown that a system with high packet load which uses TDMA have the same or higher throughput as if the system would use CDMA. But the main focus in this is study is more on the number of admitted users and delays, so it differs from this study and not directly comparable In study [8], the effects of discontinuous transmission in WCDMA systems is considered, which is similar to TDMA. It is shown that when users switch from transmitting to not transmitting and vice versa the channel conditions for the other users in the cell changes, which cause variation in the signal quality. Further more it is shown that this problem is getting worse with increasing bit rates. Another study in [7], a changed power control algorithm in presented. The algorithm is capable of compensate for large power variations faster than the present one. This might be an interesting approach to solve the problems that can occur when introducing TDMA in EUL. 1.3 Problem Statement UMTS networks use WCDMA as its air interface. If more than one User Equipment (UE) uses a high bit rate in the same cell, the interference increases and thus the coverage decreases. This forces the radio base station to reduce the bit rate for the UE s to maintain coverage. One idea to handle this problem and also improving the radio resource utilization is to change the access scheme to TDMA. The idea with TDMA is that the UE s should take turn in transmitting and hence experience less interference from other UE s. The master thesis assignment is to study the consequences of introducing TDMA in UMTS networks. What are the main problems and how can they be solved or the effects of them minimized. The goal is to show that there are potentially significant gains with an introduction of TDMA under favorable conditions. Moreover, to identify the main problems and limitations that the existing systems bring. All results shall both be explained theoretically and supported by simulations. 1.4 Approach The following steps are taken to answer the problems stated in Section 1.3. Create a reference case with CDMA Adjust the simulator for the new access scheme, TDMA

19 1.5 Thesis Outline 3 Develop and use a framework for a theoretical analysis Run simulations to identify the main problems Solve or minimize the effects of the discovered problem, verify with simulations. 1.5 Thesis Outline Chapter 2 gives an introduction to cellular system in general, followed by a part more focused on the third generation s mobile systems and particular the evolution on the Enhanced Uplink. In Chapter 3 there is a system model to give an understanding for the system. The first part contains an introduction of the main notations followed by how the dynamic system is modeled. The results are presented in the fourth chapter; the intention is to show what the main problems and possible solutions are. This is followed by the overall conclusions in Chapter 5.

20

21 Chapter 2 Cellular Radio Systems This chapter gives an introduction to cellular systems. Most focus is on Wideband Code Division Multiple Access (WCDMA), which is the air interface, and the concept of the Enhanced Uplink (EUL). A more detailed description of WCDMA can be found in [5, 6] 2.1 History The first generations (1G) mobile systems were introduced in the early 1980 s; they were analog networks and design for speech services. Several standards were used around the world; one of them was the Nordic Mobile Telephony (NMT) which was introduced in the Nordic countries in During the 1980 s the second generation (2G) of mobile networks were developed, which used digital communication. With digital technology the capacity and quality were increased and it was possible to develop more interesting services. Europe launched the GSM project which presented a standard in the mid 1980 s. In the mid 1990 s GSM was developed to support packet data, which often are referred to as 2.5G. Demands for more advanced services e.g., video telephony, streaming video and web browsing pushed the development further. The result was the third generation mobile systems (3G) with the first standard settled in In Europe the standard 3G system was named Universal Mobile Telecommunications Services (UMTS) with the WCDMA as its air interface. These systems use the radio resources more efficient and can therefore offer higher data rates and increased capacity compared to 2G. 5

22 6 Cellular Radio Systems 2.2 Propagation When radio signals are transmitted and propagate in air or close to obstacles some basic propagation mechanisms will occur, such as reflection, diffraction and distance attenuation. Those mechanisms result in a varying and unpredictable signal which will cause attenuation of the received power. The total attenuation or power gain, g, is the ratio between the received power, P rx, and the transmitted power, P tx, and can be expressed as, P rx P tx = g = g p g s g f g a < 1 (2.1) where, g p is the distance path gain, g s is the shadowing fading, g f is the fast fading and g a is the antenna gain. All factors of the total power gain are described in more detail below Distance Path Gain The attenuation due to the distance between the transmitter and the receiver is described by the path gain. The path gain is often modeled as, g p = C R n < 1 (2.2) where C, n are constants and R is the distance in meters. In the simplest of all propagation models, free space propagation, n = 2. Here all obstacles that may affect the field is disregarded. In more realistic models for telecommunication, often derived from the Okumura-Hata propagation model, n often varies between 3 and 5. [2] Shadowing Fading Shadowing fading occurs due to large objects or terrain obstacles e.g., hills and large buildings between the transmitter and receiver. As a transmitter, or receiver, moves through an environment it will be shadowed from the receiver, or transmitter, by different obstacles. This cause the received signal power to fluctuate. Since the objects are on a large scale it will take some time for the transmitter to move out of a shadow region, this makes the shadowing fading quite slow. A common model for these variations is a log-normal distribution, an example on how it can look is in Figure 2.1. [2] Fast Fading Fast fading, or multipath fading, is an effect of propagation that results in radio signals reaching the receiving antenna by two or more paths. Those paths have different length and therefore the received signal components will vary in phase. This may generate constructive or destructive interference, resulting in rapid variations in received signal power. The standard statistical model for fast fading is a distribution called Rayleigh distribution, an example is shown in Figure 2.1. [2]

23 2.3 Receiver Antenna 7 Figure 2.1. A typical attenuation of the received signal in a mobile radio link where the fast and shadowing fading is illustrated. [2] Antenna Gain Antenna gain is the ratio of the power density of an antennas radiation pattern in the direction of strongest radiation to that of a reference antenna. The reference antenna is often an isotropic antenna and the gain is expressed in dbi (decibels over isotropic). The gain of an antenna is a passive phenomenon; no power is added by the antenna but redistributed to provide more power in a certain direction and less power in others. [2] 2.3 Receiver Antenna To efficient collect signal energy that has been dispersed in time by multipath fading in a cellular radio system, an efficient receiver antenna has to be used. Today s radio systems employ Rake receivers, which is described in the following section Rake Due to multipath fading the received signal may consists of more than one copy of the transmitted signal. The difference between the signal components are delay and attenuation. A Rake receiver consist of multiple correlation receivers called fingers, those finger can be adapted to the different delays of the signal components. Each finger creates an image of a delayed component of the transmitted signal. The Rake receiver combines all images to minimize the affects of multipath propagation which reduce the own interference. The problem with a Rake receiver is that it assumes that the interference is white noise, but with multipath propagation the interference is colored. A Rake receiver is therefore not the best solution in a CDMA system. [4, 9] G-Rake The Generalized Rake (G-Rake) is an improved Rake which can handle colored interference better. One improvement in the G-Rake receiver is that there are

24 8 Cellular Radio Systems extra interference fingers. Those fingers are used to capture information of the interference and make a better approximation. [4, 9] 2.4 Multiple Access Multiple access allows several users to transmit over the same radio resource or carrier frequency. There are a lot of different schemes that are well suited for wireless communication. This section will briefly describe the most common in telecommunication with focus on Time Division Multiple Access and Code Division Multiple Access FDMA Frequency Division Multiple Access (FDMA) is a scheme which allows several users to share the radio spectrum. The available bandwidth is divided into a number of band-pass channels. Between the channels there is often a certain gap or guard interval to ensure that channels do not interfere with each other, those guard interval leave some of the bandwidth unused. Users are assigned one or more channels and transmit on those frequencies throughout time, illustrated in Figure 2.2(a) TDMA Time Division Multiple Access (TDMA) is another scheme which allows users to share the same frequency channel. By dividing the frequency channel into a number of timeslots, the users can be separated in time. Users are assigned one or many time slots in which they are allowed to transmit using the full system bandwidth, illustrated in Figure 2.2(b). To achieve non-overlapping signals guard interval, like in FDMA, and synchronization in time is needed. [5] (a) FDMA (b) TDMA Figure 2.2. FDMA and TDMA.

25 2.4 Multiple Access CDMA Code Division Multiple Access (CDMA) is a multiple access scheme where users are separated by codes. Each users is assigned a unique code sequence (spreading code) and can therefore use the entire frequency domain the whole time, thus there is no physical separation in frequency nor in time for different users. This results in mutual interference between the users. On the transmitting side the user data is multiplied by the assigned spreading code. The spreading code consists of chips and there are chip rate, R c, chips per second. The spreading results in a spread signal in the frequency domain, therefore the name spreading code. The ratio between the chip rate and the data rate, n, is referred to as the spreading factor (SF) or processing gain. The spreading codes are so-called Orthogonal Variable Spreading Factor (OVSF). Those codes are arranged so that they are perfectly orthogonal to each other, even if the data rates are different. This holds as long as the assigned codes for the users are selected from different branches in the code tree. The code tree is defined by the tree structure in Figure 2.3. On the receiver side the spread signal is multiplied by the same spreading code Figure 2.3. Code tree that is generating OVSF codes of varying length. and the original data sequence can be retrieved. If another orthogonal spreading code is used to decode the same spread signal, the result will be interpreted as noise or interference. The spreading and despreading procedure is illustrated in Figure 2.4. Since all other users signals will be interpreted as interference, care must be taken so the interference caused by other users does not exceed the own signal power. If that happens the own signal will drown in interference and may not be retrievable from the despreading. As the wanted signal is despreaded the despreading gives a gain compared to other spread signals. This gain is called the processing gain, G p, and can be calculated from Equation 2.3. Here, R c is the

26 10 Cellular Radio Systems Figure 2.4. Spreading and despreading procedure. The data is retrieved when the original spreading code is used in the despreading procedure, using another code results in noise. chip rate and the R b is the bit rate. G p = 10 log ( R c R b ), [db] (2.3) The spreading and despreading procedure in the frequency domain is illustrated in Figure 2.5. Figure 2.5. Spreading and despreading in the frequency domain. 1. The original Signal; 2. Spreading of the Original signal; 3. Transmitting and experience interference; 4. Adding Wide Band Interference (WBI) and Narrow Band Interference (NBI); 5. Despreading; 6. Filtering

27 2.5 Cellular System Cellular System A cellular system is a radio network made up of a number of cells each served by a base station. These cells are used to cover different geographical areas to provide network access for many mobile users Uplink and Downlink In a cellular system there is transmission in two directions, from the user to the base station and vice versa. The transmission from the user to the base station is called the uplink and in the other way is called the downlink. There is several ways to implement communication in both directions, Frequency Division Duplex (FDD) is the one commonly used in CDMA systems. FDD means that the uplink and downlink are separated in frequency, one carrier frequency for the uplink and one for the downlink. Time Division Duplex (TDD) is another method to separate the uplink from the downlink. Here one carrier frequency is used and the uplink and downlink take turn in transmitting Power Control Power control is one of the most central and important functions in a cellular system. The main purpose is to control the transmitted power of simultaneously transmitting users so the interference in the system is minimized and in the same time provide sufficient signal quality at the receiver side. The signal quality is often described by the Signal to Interference Ration (SIR). SIR is the ratio between the wanted signal power and the total interference power. The interference power can be signals from other users or from background noise. In CDMA systems there are many users transmitting at the same time on the same carrier frequency. These users can be far away, close to the edge of the cell, to very near the base station when they transmitting. Users transmitting with very high power at a close distance would make signals from users far away with too little power undetectable. This is referred to the near-far problem. 2.6 UMTS The Universal Mobile Telecommunication System (UMTS) is often called third generation mobile radio system (3G), and is the leading 3G technology today. The specification for UMTS has been specified by third-generation partnership project (3GPP) which is a joint standardization project. 3GPP is continuously developing UMTS to get higher bit rates and more efficient usage of the radio resources. Members of the 3GPP are companies such as mobile telephony system manufactures and operators.

28 12 Cellular Radio Systems WCDMA To transfer information over the air, UMTS uses the Wideband Code Division Multiple Access (WCDMA) as the air interface, which is based on CDMA described in Section A chip rate of 3.84 Mcps is used on the original signal and then spread out and transmitted on the 5 MHz carrier. The 5 MHz bandwidth is considered wide, compared to many other schemes which operate at 1.25 MHz Network Architecture UMTS networks are at a high-level system point of view divided in three subsystems, namely the core network (CN), UMTS Terrestrial Radio Access Network (UTRAN) and User Equipment (UE). An overview of the architecture can be seen in Figure 2.6. The core network routes traffic to external networks like the Internet and the public switched telephone network (PSTN). In the UTRAN each Radio Network Controller (RNC) controls several Radio Base Stations (RBS), also known as Node-B, and communicates with the CN. The RBS handles a numbers of cells (sectors) which each covers a geographical area. The main task for the RBS is to handle the radio signaling and traffic to the UE s. Resources allocation like UE-service requests is shared by the RNC and RBS. Figure 2.6. UMTS network architecture Uplink Power Control The shared resource in the uplink is the interference headroom. In WCDMA systems power control becomes crucial for efficiency and stability when the interference headroom must be managed. Without it, a single overpowered UE could block a whole cell. The power control is divided into two parts, the inner loop and outer loop. The control mechanism is illustrated in Figure 2.7. The goal is to keep the SIR just high enough to provide the required signal quality for the service in use. By keeping the SIR as low a possible the UE is not us-

29 2.6 UMTS 13 ing too much power and therefore does not create more interference then necessary. The SIR target is steered by the Outer Loop Power Control (OLPC), which is located in the RNC, and is regulated to achieve a specific block error rate (BLER). BLER is the ratio between erroneous received data blocks to the total number of received data blocks. If the quality is better than wanted the outer loop decreases the SIR-target and if it is too low it will increase the target. The outer loop is updated every Transmission Time Interval (TTI) which is the time frame in which a UE may transmit over the air interface, e.g., every 2 ms. The Inner Loop Power Control (ILPC) is located in the RBS and is updated with a frequency of 1.5 khz; this is to be able to react to fast variation in the channel condition, like fast fading. It makes a rapid estimation of the SIR and regulates it towards the SIR-target. Depending on if the measured SIR is higher or lower than the SIR-target, the RBS sends a Transmit Power Control (TPC) command to the UE to either raise or lower its transmission power by a fix step size. Figure 2.7. Uplink power control Interference Cancellation In CDMA systems UE s create interference to each other that affects the performance; Interference Cancellation (IC) is a method to improve the performance. The principle with IC is first to detect the information of an interfering UE, then reconstruct the interfering signal and cancel it from the total received signal. Detection of the desired UE s is finally performed. IC can be performed in two ways, Parallel Interference Cancellation (PIC) and Successive Interference Cancellation (SIC). PIC is when the contribution of all interfering UE s is cancelled simultaneously in a parallel manner while SIC cancel

30 14 Cellular Radio Systems the contribution of the strongest remaining interferer iteratively. In general SIC takes more time while saving hardware compared to PIC. On the other hand PIC is preferable when the amount of interference from each UE is similar. Conversely, SIC is preferable in the case when UE s interferer with different signal strengths. [3, 6] 2.7 EUL The evolution of the WCDMA-standard is an ongoing process, as a summary, the evolution of WCDMA is illustrated in Figure 2.8. The first step was the introduction of High-Speed downlink Packet Access (HSDPA) in Release 5 or just Rel 5. The next step, in Rel 6, was to complement the improved downlink with the Enhanced Uplink (EUL). HSPDA and EUL are often jointly referred to as High-Speed Packet Access (HSPA). The main goals with HSPA are: Reduced delays Increased data rates Increased capacity Increased high data rate availability To achieve this, a lot of functionality is moved from the RNC to the RBS, e.g., decision making functions is moved to the RBS, resulting in less signaling and reduced delays but also reduced knowledge of the surrounding when the information in the RNC no longer are available. With Rel 6, WCDMA-systems support data rates up to 5.7 Mbps in the uplink, compared to 2 Mbps in the first Release, R99. The general principles behind the most important features introduced with EUL are described below. Figure 2.8. The evolution of WCDMA EUL Channels EUL introduces a new set of channels, the Enhanced Dedicated Channel (E-DCH), which is only used by EUL UE s. This channel can co-exist with the existing Dedicated Channel (DCH), which consist of the Dedicated Physical Control Channel (DPCCH) and Dedicated Physical Data Channel (DPDCH), from earlier releases. The E-DCH consist of two channels, the Enhanced Dedicated Physical Control

31 2.7 EUL 15 Channel (E-DPCCH) which carries uplink control signaling while the Enhanced Dedicated Physical Data Channel (E-DPDCH) carries the data. The E-DPCCH is only used when data is transmitted with EUL from the UE, while the DPCCH is always transmitted. The DPCCH contain e.g., control information such as Channel Quality Information (CQI) for the downlink and the RBS uses this channel for SIR measurement. The power control, described in 2.6.3, regulates the UE power based on the measured SIR. When the E-DCH channel is used, the power level is regulated together with the DPCCH. When data is transmitted using EUL the power level for the E-DPCCH and E-DPCCH are set with two offsets relative to the DPCCH power. The power for E-DPDCH is, E DP DCH p DP CCH, illustrated in Figure 2.9. Figure 2.9. Offsets for E-DPCCH and E-DPDCH The offsets are of great importance, especially the one for the E-DPDCH which is based on bit rate and the efficiency of the channel coding that the UE uses. In general higher bit rates demand higher power which leads to a higher offset for higher rates Short TTI The available radio resources are allocated between the UE s accessing the system at certain intervals, the length of those intervals is referred to Transmission Time Interval (TTI). A short TTI, 2 ms instead of 10 ms, results in shorter delays due to a reduced Round-Trip Time (RTT). A 2 ms long TTI consist of 3 slots, e.g., TPC are transmitted every slot.

32 16 Cellular Radio Systems Scheduling Scheduling is another important feature, in the EUL concept it is moved from the RNC to the RBS for shorter delays. The scheduler controls when a UE are allowed to transmit and with which maximum rate. The maximum rate for each UE is chosen so that the total interference at the RBS does not exceed a certain threshold, a fast scheduling can respond to changes in the system faster and therefore minimize the risk for the system to overstep the threshold Hybrid ARQ If it is possible to decode the received data an acknowledgement (ACK) is replied to the UE, if the decoding fails there is erroneous packet and a negative acknowledgement (NACK) is replied instead. All erroneous packets have to be retransmitted by the UE. One approach to handle those retransmissions is to use Automatic Repeat Request (ARQ). Instead of ignoring the erroneous data completely, the Hybrid ARQ (HARQ) scheme combines the information from the earlier transmission with the new information from the retransmission. This procedure is called soft combining and increases the chance of a successful decoding, simply because an erroneous packet still provides useful information together with the retransmission. As long as the UE receive a NACK, the packet will be retransmitted. In order to be able to transmit the whole time, no need to wait for an ACK or NACK due to the processing delay, multiple HARQ processes can be used. The retransmission will always take place in the same process in the next cycle. A HARQ scheme is illustrated in Figure Figure Operation of 4 parallel HARQ-processes.

33 Chapter 3 Simulator Model To be able to evaluate the advantages and the disadvantage with the introduction of a TDMA access scheme in EUL, a series of system simulations were done. All the simulations were done in a WCDMA radio network simulator with support for HSPA; it is developed by Ericsson AB and written entirely in Matlab. The simulator models everything from fast and slow fading, traffic models to different high level network functionality such as load control and scheduling. First in this chapter a number of notations and definitions will be introduced, followed by a more detailed description of the system model. 3.1 Definitions The introduced equations in this section will not include the effects of delays in the system; they will rather describe an average stationary point or the dynamics where the mean values of the different distributions are used. Consider a network with one base station serving M UE s. As discussed in Section 2.2, all signals transmitted over an air interface experience a basic path gain. The basic path gain between each UE to the base station is defined in Definition 3.1. Definition 3.1 Path Gain G = ( g 1,..., g M ) < 1 where g i is the total uplink path gain between UE i and the base station. The total received power at the base station from UE is expressed as C tot i = p i g i (3.1) where p i is the power used by UE i. The power received from the DPCCH channel DP CCH is denoted C, the remaining channels are given by the DPCCH power times i 17

34 18 Simulator Model a power offset. To simplify, the E-DPCCH and E-DPDCH channels are grouped into one. Ci tot can then be divided into two parts, C tot i = C DP CCH i (1 + EUL i ) = C DP CCH i where EUL i is the power offset for the E-DCH channels for UE i. tot i (3.2) The sum of thermal noise and other non power controlled interference will be referred to as background noise, denoted N. The total interference power at the base station, I tot, is the background noise plus the received signal power from all UE s. The total interference power will be referred to as interference or total interference. Definition 3.2 Total Interference is defined as I tot = M i=1 C tot i + N = M p i g i + N The signal quality for each UE in a radio system is often measured in Carrier-to- Interference Ratio (CIR). CIR is the ratio between the received carrier power and the experienced interference before the despreading in the receiver. In a CDMA radio system SIR is measured after despreading. SIR and CIR are related as SIR = SF CIR, where SF is the spreading factor. In UMTS networks the signal quality is measured on the DPCCH channel. The power control makes the decisions based on that measurement, so only the DP CCH Ci part of Ci tot is included in the CIR measurement. i=1 Definition 3.3 Carrier-to-Interference Ratio is defined as where γ i is the CIR for UE i. DP CCH C γ i = i I tot Ci tot Due to multipath propagation the signal components reach the receiver at different times, described closer in Section This may cause the signal to interfere with itself. The orthogonality factor, α, describes how well the receiver can handle those delays. Here α is constant but in reality it varies between UE s and time. The diversity gain, d, describes the signal gain due to the use of several receiving antennas. Also d varies with time and mobility, but it is modeled as a constant. Both α and d affect the signal quality and therefore also CIR. Definition 3.3 is an ideal situation but can be rewritten with regard to α and d as

35 3.1 Definitions 19 Definition 3.4 CIR is defined as DP CCH Ci d γ i = I tot (1 α)ci tot Definition 3.4 is the definition for CIR that is going to be used throughout this study Shannon s Theorem In the late 40s Shannon published his Communication Theory, which included the formulation of Shannon s Theorem. Shannon s Theorem define a theoretical upper limit for the bit rate based on bandwidth and CIR. [11] Theorem 3.1 Shannon s Theorem R = W log 2 (1 + γ) [Mbits], where R is the maximum attainable bit rate in Mbps, W is the bandwidth in MHz and γ is experienced CIR This maximum bit rate is a theoretical restriction and can never be achieved in a real system. The theorem will however be used as a model to map CIR-values to a corresponding rate that the UE s are allowed to use. Figure 3.1. Maximum theoretical bit rate as a function of CIR. Notice that in this study W is constant so higher rates demand higher CIR and the increased demand on CIR is not linear to the bit rate, illustrated in Figure 3.1. Higher bit rates inevitable result in a nonlinear increasing interference in the system.

36 20 Simulator Model Noise Rise A common way to express the interference at the base station is to describe it as rise over thermal noise (RoT), or just noise rise, Λ. Noise rise is the ratio between the total interference, I tot, and the background noise, N. Definition 3.5 Noise rise, Λ is defined as Λ = Itot N = M i=1 Ctot i N + N = 1 + M i=1 CDP CCH i N tot i Noise rise is the limited resource in the base station. If the noise rise is high the UE has to use more power to maintain its CIR. UE s have an upper limit on the power that they are allowed to use, so if the maximum power is used and the signal quality is too low the UE s are power limited. If the noise rise is too high in the base station, UE s far away from the base station can not make themselves heard. To maintain coverage and stability in the base station a upper limit is set on the noise rise, Λ Λ max (3.3) Definition 3.4 combined with Equation 3.2 can be rewritten as DP CCH Ci = γ i d + (1 α)γ i tot i 1 I tot (3.4) By combining Equation 3.4 with Definition 3.5, Λ can be expressed as Λ = Itot N = 1 1 M i=1 γ i tot i d+(1 α)γ i tot i Equation 3.3 gives a upper limit on the sum in Equation 3.5, 0 M i=1 1 (3.5) γ i tot i d + (1 α)γ i tot < 1 (3.6) i Equation 3.6 leads to a limit on how high rates that the individually UE s are allowed to use Interference Cancellation The following model is an SIC, which is described closer in Section This model will only have one iteration, so only the most interfering UE, i k, will be canceled. UE i k is experience the total interference, I tot and the other UE s experience a canceled interference, I ic, modeled as I ic = N + Ci tot k (1 η) + Ci tot I tot (3.7) i i k where η is the efficiency for the SIC. Since I ic is lower than I tot the UE s that not are the i k can use lower power to maintain their signal quality. A system using

37 3.2 Propagation 21 SIC will therefore have lower total interference compared to a system without SIC. To have the same amount of interference in a system with a implemented SIC that are using one iteration compared to a system without an SIC, UE i k shall fulfill the following condition C tot i k = C totic i k (1 η) (3.8) 1 With an SIC UE i k can create 1 η times more interference without any change in performance. This means that UE i k can use a higher bit rate. 3.2 Propagation Section 2.2 described in more detail what fading is. In the simulator the distance attenuation, or path gain, is derived from the Okumura-Hata propagation model. The slow fading is derived in the simulator with a log-normal distribution. The simulation is in an urban environment and the fast fading is modeled through the standardized model named 3GPP Typical Urban. The model is a map consisting of precomputed fast fading values. The antenna gain is derived from implemented Omni-antenna diagrams in the simulator. 3.3 Receiver Antenna The G-Rake receiver makes a better approximation of the interference than a Rake receiver. The orthogonality factor, α, is thus lower when a G-Rake is implemented. Therefore a G-Rake antenna is used in this study. 3.4 Network Layout The network used in all simulations is a single site with a single cell. This means one RBS that handle all UE s in the simulation. 3.5 Traffic Models It is possible to use many types of traffic models in the simulator, i.e. speech, interactive, download etc. This study will cover two kinds of models for uploading a large file at different rates; one using EUL and one not using EUL, the third one is pure speech. All three are further described in this section Upload with EUL The standard UE is using a model for uploading a large file over the E-DPDCH channel, which means EUL is used. The file is so large so the UE s will be active

38 22 Simulator Model during the whole simulation, regardless of bit rate. The number of UE s will be set for each simulation individually and all of them are supporting category 6 1, which means they are supporting bit rates up to 5.76 Mbit. This one is used if nothing else is stated Upload without EUL The second model is also a large file that shall be uploaded, but this model is not supporting EUL and therefore use the DPDCH channel. All these UE s will get a bit rate of 64 kbps. As in the previous model the number of UE s will be set for each simulation and they stay active during the whole simulation Speech The data in the speech model is transmitted over the DPDCH channel and demands a bit rate of 12.2 kbps. In this model the UE s are not active the whole time. Each call lasts for a certain time which is exponentially distributed with a mean time of 90 seconds. As in the other models, the number of UE s will be set for each simulation. 3.6 Mobility All UE s have the same mobility; their speed is 3 km/h, which is a rather low mobility. When new UE s enter the system they are initially placed randomly in the cell according to an uniform distribution. 3.7 Synchronization As mentioned in Section 2.4.2, TDMA needs a guard interval too minimize the risk that two or more UE s are transmitting at the same time and causes extra interference for each other. In this study the system and all UE s are perfectly synchronized, so no guard interval is needed and therefore not further considered. 3.8 UE Power Within the standard UE s are not allowed to use more than 21 2 dbm when transmitting. If a UE is far away from the serving RBS, it can not use the highest bit rates. This is because the received power in the RBS is too low to maintain the signal quality that is required for the higher bit rates. This study focuses on UE s with good channel condition and that are capable of using high bit rates, so instead of place the UE s closer to the RBS, the maximum available power in the UE s are increased to 50 dbm. 1 See appendix C for more information about UE categories 2 See Appendix B

39 3.9 Power Control Power Control The power control, described in Section 2.6.3, is implemented in the simulator. The step size in power that the UE take when order from the RBS is to go up or down is 1 db Hybrid ARQ HARQ with soft combining are used as described in Section 2.7.4, 8 parallel HARQ queues, or processes, are used and each is one TTI long. The minimum delay is defined by the TTI length; this is the time during which data is transmitted. If a retransmission occurs an additional delay is introduced which is specified by the HARQ Round-Trip Time, which depends on the number of processes, RT T = N HARQ T T I. In the case with 8 processes which are 2 ms each, the extra delay for each retransmission is 2 8 = 16 ms TDMA Within the standard for UMTS it is possible to allocate the uplink resources for each TTI, or for each HARQ-process. The idea is to give a group of UE s one or more HARQ-processes each for a certain time. During a HARQ process only one UE is allowed to transmit. This is a way to implement a TDMA access scheme and it is used in this study. The transmission time, T, is the number of HARQ processes in row multiplied by the TTI length, illustrated in the two following examples. Example 3.1 In both figures in Figure 3.2 two UE s have four HARQ-processes each. The four assigned processes do not have to be in a row, they can be scattered in any way. The difference between the two figures is that in Figure 3.2(a), T is N HARQ T T I length = 4 2 = 8 ms, and in Figure 3.2(b) T is N HARQ T T I length = 1 2 = 2 ms. As described in Section the retransmission is done in the same process as the first transmission, in this example it is no problem because the UE s have the same processes during many cycles. There are just a few combinations of the number of UE s and transmission time that fulfill these conditions.

40 24 Simulator Model (a) Four HARQ processes are in a row, T is 8 ms. (b) One HARQ process in a row, T is 2 ms. Figure 3.2. Two UE s with four HARQ processes each. Example 3.2 In Figure 3.2 there is three UE s with T = 6 ms. This combination does not fulfill the condition that they can have the same HARQ process during more than 1 cycle. If there is a retransmission, the scheduler must give that process back to the UE that have to do the retransmission, illustrated in Figure 3.3(b). (a) T is 6 ms, notice that each UE have different HARQ processes each cycle. (b) An error in HARQ process 4. In next cycle UE 2 still have HARQ process 4. Figure 3.3. Three UE s with three HARQ processes each.

41 3.12 Interference Cancellation Interference Cancellation The SIC described in Section is implemented in the simulator. In this study the effects of SIC is only studied with a TDMA access scheme, the UE s take turn in transmitting so there is only one active UE s at the same time. That extra interference that UE s can create are translated to a higher bit rate. The SIC is implemented in two ways to make a comparison; one way is to do the IC before the channel decoding, pre-decoding SIC, and the other way is to do it after the channel decoding, post-decoding SIC. The benefit with pre-decoding SIC is that the delay to get the data, the most important data in this study is the TPC, is increased by only 1 slot, the disadvantage is that the efficiency depends on the coding rate. With post-decoding SIC the delay increases with 3 slots instead, but the efficiency is high and is independent of the coding rate EUL EUL is simulated with all the functionality described in Section 2.7. The short TTI, 2 ms, is used throughout this study Scheduling The scheduler controls which UE s and when they are allowed to transmit, and with which rate. The assigned rate is based on which rate is needed for the present service used by the UE and the amount of interference they are allowed to create. Speech UE s have the highest priority and are granted a transmission rate first. All speech UE s are granted resources corresponding to a bit rate of 12.2 kbps. Upload UE s without EUL have the next highest priority and are next to get resources. If there is enough resources, they get grants to use a bit rate of 64 kbps, otherwise they share the available resources equally by a scheduling algorithm. Speech UE s and upload UE s without EUL are put on a CDMA access scheme, which is described in Section The remaining resources are distributed to the EUL UE s. Those UE s are set to use a TDMA access scheme, described in Section 2.4.2, since they take turn in transmitting one UE is granted all the remaining resources. In most simulations that are carried out in this study there is only EUL UE s, in those cases the active UE gets all available resources, except that part that is needed for the DPCCH channels Delays It takes some time for the commands from the RBS to reach and be decoded by the UE s. Those delays are implemented in the simulator and give dynamics to

42 26 Simulator Model the simulated network. The two delays that affect this study is the transmit power command, TPC, and the rate grant command. The time it takes both for the algorithm in the RBS to calculate the values and the time over the air interface are listed below, TPC - 2 slots - 4 ms Rate Grant - 15 slots - 30 ms 3.15 System Logging All simulations are run with a simulation time of more than 100 seconds, which is long enough to get reliable statistics. The initial 10 seconds of the simulations are disregarded in order to let the system stabilize. UE performance measures such as bit rate; power-usage etc is logged for each TTI or slot depending on the update frequency. System data e.g., noise rise and system throughput are logged for each TTI, some UE data e.g., TPC are logged for each slot.

43 Chapter 4 Results The results will be presented and analyzed in this chapter. All simulations are carried out with the simulator described in Section 3. The analysis will be from a system perspective, and thus little or no concern is taken to the fairness between the UE s. 4.1 CIR Overshoot The main problem with introducing a TDMA access scheme in EUL is the large variation in experienced interference for the UE s. The large variation in interference is due to large difference in transmission power when the UE s take turn in transmitting. A transmitting, or active, UE experiences only interference from the DPCCH channels from the not transmitting, or passive, UE s, which is a rather small interference. A passive UE experiences interference both from the DPCCH from the other UE s and the E-DPDCH channel from the active UE, which is significant. This major change in channel conditions affects the signal quality Theoretical Definition 3.4 gives an expression for the signal quality, γ. When TDMA is used C tot varies a lot for each UE depending on if the UE is passive or active. The total received carrier power, C tot described in Equation 3.2, for an active UE is C tot = C DP CCH (1 + EUL ) (4.1) and CIR for the same UE is expressed in Equation 4.2. γ = C DP CCH d I tot (1 α)c DP CCH (1 + EUL ) (4.2) The total received carrier power for a passive UE is C tot = C DP CCH (4.3) 27

44 28 Results and the CIR for the same UE is expressed in Equation 4.4 γ = C DP CCH d I tot (1 α)c DP CCH (4.4) Since EUL >> 1, C DP CCH for the active UE can be reduced and still maintain the same CIR, provided that I tot is constant. The power control strives towards a CIR-target which is independent of if the UE is passive or active. When switching from passive to active the CIR gets too high, this creates a CIR overshoot. Conversely when switching from active to passive the CIR gets too low and generate a CIR undershoot. It takes a while for the UE to adjust the power so the CIR reaches the CIR-target, the time is depending on how large EUL is and the step size in the power control. The principle appearance of CIR when two UE s take turn in transmitting is illustrated in Figure 4.1. Figure 4.1. Two UE s take turn in transmitting. When EUL is high the UE is active. The resulting over- and undershoot for each UE is shown.

45 4.1 CIR Overshoot 29 If the transmission time is short, the height of the CIR overshoot can be so high that the power control has not enough time to adjust the power so that the CIR reaches the CIR-target. The principle appearance of CIR when two UE s take turn in transmitting with a too short transmission time is illustrated in Figure 4.2. Figure 4.2. Two UE s take turn in transmitting with a short transmission time. CIR never reaches the CIR-target. Compared to a CDMA system where the UE s are active the whole time, there is no CIR under- or overshoot. Hence, CIR will follow the CIR-target the whole time Simulations The affects of introducing TDMA in EUL can also be seen in the simulator. In Figure 4.3, measured CIR and the CIR-target is plotted for different bit rates. As shown in Section 4.1.1, it takes a while for the UE to adjust the power so the CIR reaches the CIR-target after switching from transmitting to not transmitting and vice versa. It takes longer time to adjust the power when higher bit rates are used. For the highest bit rate, there is not enough time to reach the CIR-target, all this is illustrated in Figure 4.3. As illustrated in Figure 4.3, for the lowest bit rate there is basically no over- or undershoot, CIR is on target the whole time. With a bit rate of 3 Mbps, the overand undershoot is clearly visible but there is enough time for the UE to reach the CIR target. With the highest bit rate, the over- and undershoot is even larger. There is not always enough time to reach the CIR target. The principles from the theoretical part are observed in the simulator as well.

46 30 Results Figure 4.3. Simulations of CIR and CIR-target for varying bit rates when two UE s take turn in transmitting. The over- and undershoots are larger for higher bit rates. 4.2 Performance Improvement Evaluation The idea with introducing TDMA in EUL is to be able to offer higher bit rates to many UE s simultaneously. This leads to a higher cell throughput which is the sum of all UE s bit rate in the cell. In the following simulations the UE s are not power limited, they all use the traffic model upload with EUL, which is described closer in Section Theoretical With TDMA the UE s take turn in transmitting, therefore they do not creating any interference for each other. As there is just one active UE at a time, the RBS shall give the active UE all the available uplink resources. When the RBS have to serve a lot of UE s the control channels, mainly the DPCCH, take more and more of the uplink resources. This results in that the maximum available bit rate decreases with increasing number of UE s in the system. In a CDMA system when the UE s transmit simultaneously, the bit rate will depend more on the number of UE s. More UE s increase the amount of interference in the system, and lower bit rates have to be used to not create too much noise rise.

47 4.2 Performance Improvement Evaluation Results The cell throughput is plotted against the number of UE s in the cell in Figure 4.4. There is both throughput for a system using CDMA, which is the reference case, and a system using TDMA with different transmission times, T. Figure 4.4. System throughput for a CDMA system and a TDMA system with different transmission times. For short transmission times, the system throughput is not as high expected. As seen in Figure 4.4, TDMA with a transmission time longer than 3 HARQ processes, or 3 TTI which is 6 ms, behaves as expected. The throughput is almost constant for a low number of UE s then start to decrease when the DPCCH take more and more resources. The gain with introducing TDMA compared to CDMA is clearly shown here. The system throughput is more than 100% higher for the TDMA system compared to the same system using CDMA for two or more UE s. Note that if there is only one UE, the active UE in both systems are alone and gets all the resources, therefore there is no difference between the two systems. For transmission times shorter than 3 TTI, the system does not behave as expected. The rate is so high that the transmission time is too short for the CIR to reaches the CIR-target. When this occurs the system can be unstable. In Figure 4.5, the CDF 1 of the noise rise are plotted. The curves are for three UE s in the system, one curve is the CDMA system and the two others are the TDMA system with a transmission time of 1 and 4 TTI. The threshold for the noise rise is set to 12 db. This means that the system tries to keep the noise rise below this value to maintain coverage. 1 Se appendix A

48 32 Results Figure 4.5. CDF of the noise rise for three UE s. A CDMA system and a TDMA system with a transmission time of 1 and 4 TTI. The TDMA system with a transmission time of 1 TTI have very high and many noise rise peaks. As seen in Figure 4.5, the TDMA system with a long transmission time, 4 TTI, has very few values above 12 db, which is what the system aiming at. This means that the UE s are using the available resources efficiently. The reference case, CDMA, has some high peaks but nothing extreme. The TDMA case with a short transmission time, 1 TTI, which corresponds to the lowest system throughput in Figure 4.4, has an extremely bad curve. More than 30% of the values are above the threshold. This system uses most resources and still has the lowest throughput. The problems seen in this section are Power rushes when a short transmission time is used Dramatically decreased system throughput when a short transmission time is used. 4.3 Power Rushes The TDMA systems with a short transmission time in Section had very high noise rise peaks and used a lot of resources. What happens in the system is that the UE s trigger each other to use more and more power. This increase in power usage is to maintain the signal quality due to more interference in the system. In Figure 4.6 the cycle is illustrated. During the first interval, all UE s use a rather low bit rate. Therefore there is a lot of available resources, the RBS then allow the UE s to use higher bit rates.

49 4.3 Power Rushes 33 Figure 4.6. Noise rise and used bit rate in the system. The granted bit rate is set so when the rate is used the noise rise in the system shall be below 12 db. It is a delay of 30 ms for UE to receive the grant and be able to use the higher bit rate. During the second interval, all UE s start to use the higher allowed bit rate. Since more power is used the interference, or noise rise, in the system increases. During the third interval the UE s use the granted bit rate. When the UE s using those bit rates, the noise rise is supposed to be below 12 db. Seen in Figure 4.6, the noise rise passes the 12 db threshold. The RBS reacts to the increased noise rise, but it takes 30 ms before the UE s start to use the lower bit rate assign by the RBS for lowering the noise rise. During those 30 ms the noise rise building up and have peaks around 30 db. The next question is why the granted bit rates, which are based on that the noise rise shall be below 12 db, generate so high noise rise peaks. TDMA is used, which generates CIR overshoot as described in Section 4.1. It takes a certain time to compensate for the CIR overshoot. If the transmission time is not long enough, there is no enough time to decrease the power to reach the target. What happens then is that the passive UE s have to increase their power to maintain their CIR. In the following discussion, Equation 4.4 is used for expressing CIR for the passive UE s and Equation 4.2 is used for the active UE. In Figure 4.7 three UE s take turn in transmitting, UE 1 starts to transmit and creates extra interference for UE 2 and UE 3. UE 1 gets an overshoot in CIR and start to decrease its power while UE 2 and UE 3 gets an undershoot and start to increase their power. The

50 34 Results transmission time is 1 TTI, which is 3 slots, so the UE s going up or down 3 db. In the second TTI, UE 2 starts to transmit and gets a CIR overshoot and starts to decrease the power. UE 1 experiences the extra interference that UE 2 creates which results in a too low signal quality and start to increase the power. Since UE 2 increased the power during the first TTI even more interference was created for UE 3 that gets a new undershoot and have to increase the power even more. During the third TTI UE 3, which has increased the power with 6 db, starts to transmit. UE 1 and UE 2 gets even more interference, and have to increase their power. During this cycle, all UE s goes up 6 db and goes down 3 db which results in a 3 db increase during 3 TTI. More and more power is used, which leads to power rushes and increasing noise rise in system. To avoid this, the transmission time have to be so long that CIR reaches the CIR target. Figure 4.7. A system with three UE s and a transmission time of 1 TTI. The UE s trigger each other to use more and more power, which leads to noise rise peaks. In the rest of this chapter a stable system means that transmission time is long enough so the power control has time to compensate for the CIR overshoot. 4.4 Interference Compensation One way to compensate for the large change in interference is to scale the power, by a factor k, that the active UE is transmitting with. If k is correctly calculated then CIR is the same when a UE goes from being passive to active, and therefore does not create any CIR overshoot which results in a more constant power usage. The CIR shall be same regardless if the UE is active or passive, hence Equation 4.2 and Equation 4.4 shall be equal. Furthermore Equation 3.1, 4.1 and 4.3 is also used. Thus k is a function of bit rate,, and expressed in Equation 4.5 d + (1 α)γ p active = p passive k = p passive d + (1 α)γ(1 + ) (4.5)

51 4.5 Minimum Timeslot Length Simulations In Figure 4.8 the system throughput and the CDF of the noise rise is plotted for the TDMA system with the scaling factor implemented. The transmission time is 1 TTI, which is the worst case from Section 4.2. Seen in Figure 4.8(a), if the power for the active UE s is scaled by k, the system throughput increases. It does not reach the system throughput as systems with longer transmission times in Section 4.2. That is because α and d in Equation 4.5 is modeled as constant but in reality and in the simulator they are not constant. The system throughput can be improved by optimizing k. This optimization is not done in this study, but to show that the throughput can be increased k are scaled by 1.2. This leads to higher throughput, seen in Figure 4.8. Shown in Figure 4.8(b), the scaling factor, if properly calculated, can improve the resource efficiency. It can even make an unstable system stable. Since the UE s do not have the information needed to calculate k, this is not possible to do within the standard today. The idea with this is to show that it is possible to compensate for the CIR overshoot and minimize the risk for getting an unstable system. 4.5 Minimum Timeslot Length Seen in Section 4.2, a TDMA system need a certain transmission time to behave properly, and Section 4.1 showed that the CIR overshoot is larger for higher bit rates. Hence it will take longer time for a system where the UE s are using high bit rates to stabilize compared to a system where the UE s are using lower bit rates. The time that is needed is depending on how long it takes for the power control to increase or decrease the power so CIR is the same after the switch from not transmitting to transmitting, or vice versa. This gives the condition γ passive = γ active = γ (4.6) The path gain in Equation 3.1 is assumed to be equal when transmitting and not. g passive = g active (4.7) Since the power control work in db and the difference in used power by the UE s is expressed in that scale by using the following logarithmic rule ( ) ppassive (p passive ) db (p active ) db = (4.8) p active To get an expression for the height of the CIR overshoot for different bit rates, Equation 4.2, 4.4, 4.6, 4.7 and 4.8 are combined. The result is expressed in Equation 4.9. db

52 36 Results (a) With the scaling factor implemented the system throughput is improved. (b) CDF of the noise rise, with three UE s in each system. With a correctly calculated k, the noise rise distribution is improved. Figure 4.8. The TDMA system with the scaling factor implemented.

53 4.5 Minimum Timeslot Length 37 (p it ) db (p i ) db = d + (1 α)(1 + EUL i t ) d + (1 α)γ (4.9) To get γ( ) system simulations with a constant bit rate was carried out for a set of data rates. A link simulation is a simulation with only one UE that using one fixed bit rate. Mean CIR, γ, and offset,, are measured during the simulations. The results from the system simulations are used in Equation 4.9 to get the height of the CIR overshoot which is then translated to the time it takes to recover, or converge, the CIR before the transition to or from being active for different data rates. The result is plotted in Figure 4.9. The difference is expressed in db. The power control step size is 1 db, so a CIR overshoot with a height of 1 db takes 1 slot to compensate for assuming no power control errors. Figure 4.9. The number of slots it takes to converge for different data rates. The +2 db line in Figure 4.9 is due to the delay for the TPC. It takes 2 slots for the UE to get the information, if the power shall increase or decrease, from the RBS. This results in that the UE can take 2 steps in wrong direction. This makes the used power by the UE can differ up to 2 db from ideal power. To be sure that the CIR overshoot is compensated for, 2 extra slots are needed. Figure 4.9 strengthen the discussion in Section 4.2. When bit rates around 3-4 Mbps are used it takes around 7 slots to be sure that the power control has compensated for the CIR overshoot. Every TTI consist of 3 slots, so the time it takes corresponds to 3 TTI.

54 38 Results Simulations A set of simulations with two UE s take turn in transmitting was carried out to see if the theoretical part holds. The UE s used a transmission time of 5 TTI, which is long, and that is to be sure that the system has time to converge. Like the system simulations, a fixed bit rate was used for each simulation. The time it took for the UE s to reach the CIR-target, or compensate for the CIR overshoot, was measured and the result is plotted in Figure Figure The simulated result of the number of slots it takes to converge for different data rates. The simulations are in line with the theoretical assessments. In the simulations, effects from all kinds of fading are included. Therefore the simulated values represent the time it takes for the power control to compensate for 95% of all CIR overshoots. The results from the simulations are in line with theoretical part, the values are on the upper limit, the +2 db line. Also these results strengthen the discussion in Section Interference Cancellation With interference cancellation, the interference created by the transmitting UE for the passive UE s depends on the efficiency, η, in the SIC. As described in Section 3.12, the extra interference that the active UE can create is first translated to a higher bit rate. If the efficiency is high enough, the active UE s will be granted the highest bit rate and still the experienced interference for the passive UE s will be lower. The disadvantage with introducing an SIC is the extra delay for the TPC. Two kinds of SIC is implemented for comparison, the pre-decoding SIC has an extra delay of 1 slot and the post-decoding SIC has an extra delay of 3 slots. Those extra delays make the used power, due to extra delay for TPC, to oscillate

55 4.6 Interference Cancellation 39 with higher amplitude. The decrease in experienced interference will affect height the time it takes to converge, or the height of the CIR overshoot will be lower, hence the passive UE s can use a lower power and still maintain their CIR. This affects how short transmission times that can be used and still maintain a stable system Minimum Timeslot Length with SIC The time it takes for the system with an SIC to converge can be calculated in the same way as in Section 4.5. The only difference is Equation 4.4, the passive UE s experience a CIR depending on the efficiency, η, in the SIC as C γ = I tot η(1 α)(1 + EUL )C DP CCH passive d DP CCH active (1 α)cdp CCH passive (4.10) By combining Equation 4.2, 4.6, 4.7, 4.8 and 4.10, an expression of the time it takes for converge with an SIC is received and expressed in Equation 4.11 (p passive ) db (p active ) db = d + γ(1 η)(1 α)(1 + EUL ) d + (1 α)γ (4.11) To get the time it takes to converge for different bit rates, the same system simulations as in Section 4.5 are used in Equation As described in Section 3.12, a post-decoding SIC has a high efficiency, here modeled as 95%, and a pre-decoding SIC a efficiency depending on the rate, here modeled by both an efficiency of 50% and 95%. The extra delay is illustrated with higher safety margin, +3 and +5 db respectively. The result is illustrated in Figure Even if the efficiency in the pre-decoding SIC is only 50%, the time it takes to guarantee that the CIR has converged is comparable to the post-decoding SIC which has an efficiency of 95%. This is because the post-decoding SIC has a longer delay for retrieving the data. If the efficiency in the pre-decoding SIC is increased, the time to converge will be shorter then the post-decoding SIC. Similar simulations are done as in Section 4.5.1, the only difference is that the pre-decoding SIC and post-decoding SIC are implemented. The result is illustrated in Figure The values from the simulations follow the theoretical assessment; they are below the upper limit. The result for the post-decoding SIC are well below the upper limit. The best results are obtained with an efficient pre-decoding SIC, but the efficiency in the pre-decoding SIC depends on the bit rate. If the efficiency is not high enough the post-decoding SIC can perform better.

56 40 Results (a) Pre-decoding SIC with an efficiency of both 50% and 95%. (b) Post-decoding SIC Figure The number of slots it takes to converge for different data rates when an SIC is used.