Bachelor Thesis HSDPA CQI Mapping Optimization Based on Real Network Layouts Supervisor: Prof. Dr.-Ing. Markus Rupp Assistant: Dipl.-Ing. Martin Wrulich by
Mar´ıa Elsa Feliz Fern´ andez Institute of Communications and Radio-Frequency Engineering
November 2007 - October 2008
Acknowledgements I would like to thank all those people who have somehow supported me during my work in this Bachelor Thesis and also during all my career in Madrid and Vienna. First I want to express my gratitude to Professor Markus Rupp for his supervision and warm wellcome, and also to all the staff of the Institute who did my time in the university more comfortable. I am specially grateful to my assistant Martin Wrulich for his constant support, guidance and patience in the development of this thesis, for the interest he expressed since the first day and the general support in my abroad experience. I am deeply grateful to my parents, who have always expressed their love, support and interest and have given me the opportunity to study abroad, and also for visiting Vienna with me. I want specially thank to my brother for helping me everytime he could, and all my friends in Madrid and Le´on. I want to thank also the friends that I met in Vienna who did my life there during six months really special and happy and showed their interest in my work, specially Corinna, Alberto, Cristina and the turkish girls. I thank their lovely frienship and support. I want to dedicate a special mention to Borja, for coming with me to the adventure of having an Erasmus experience in Vienna, for his constant support and love during this years, for his care and patience and for giving me courage in the difficults moments since we started our careers. Without his support and help this project would have never been possible.
i
Abstract In this bachelor thesis, the SISO HSDPA simulator developed for Mobilkom Austria AG shall be extended in order to handle real network layout data. The Mobilkom Austria AG will provide measured path-loss matrices of a HSDPA cluster. This data has to be converted in a suitable form to be analyzable in the simulator. Furthermore, a memory efficient loading of the data has to be implemented. Based on this real network data, an optimization of the CQI mapping of HSDPA mobiles shall be performed in order to find the mapping which maximizes the overall cell throughput. This mapping can be used to implement a suitable CQI ”re-mapping” at the Node-B, granting an optimum HSDPA network performance. The source-code has to be developed in MATLAB to ensure simple debugging and feature extendability.
ii
Contents 1 Introduction 1.1 Third Generation Services 1.2 Technology . . . . . . . . 1.3 Work environment . . . . 1.4 Objectives . . . . . . . . . 1.5 Structure of the thesis . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
1 1 1 2 3 3
2 HSDPA Principles 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . 2.2 HSDPA Standardization . . . . . . . . . . . . . . . . 2.3 HSDPA vs Release 4 DCH . . . . . . . . . . . . . . . 2.3.1 Radio resource management architecture . . . 2.4 HSDPA operation principle . . . . . . . . . . . . . . 2.5 HSDPA channels . . . . . . . . . . . . . . . . . . . . 2.5.1 HSDPA new channels . . . . . . . . . . . . . . 2.5.2 High-speed dedicated physical control channel
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
5 5 7 8 10 11 13 13 20
. . . . . . .
22 22 23 24 24 27 30 33
3 HSDPA Simulator 3.1 Introduction . . . . . . 3.2 System model . . . . . 3.3 Simulation process . . 3.3.1 Load settings . 3.3.2 Precalculations 3.3.3 Simulation loop 3.3.4 Results . . . . .
. . . . . . .
. . . . . . .
. . . . .
. . . . . . .
. . . . .
. . . . . . .
. . . . .
. . . . . . .
. . . . .
. . . . . . .
. . . . .
. . . . . . .
. . . . .
. . . . . . .
. . . . .
. . . . . . .
. . . . .
. . . . . . .
. . . . .
. . . . . . .
. . . . .
. . . . . . .
. . . . .
. . . . . . .
. . . . .
. . . . . . .
. . . . .
. . . . . . .
. . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
4 Real Network Layouts 36 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 4.2 Files structure and information . . . . . . . . . . . . . . . . . 38 4.3 Network setting based on real layout data . . . . . . . . . . . 39
iii
5 CQI Mapping Optimizations 5.1 CQI Basis in HSDPA . . . . . . . . 5.2 CQI mapping proposal . . . . . . . 5.2.1 CQI Table . . . . . . . . . . 5.3 CQI optimizations in the Simulator 5.3.1 Source Code Enhancements 5.3.2 Consequences . . . . . . . . 5.3.3 Results . . . . . . . . . . . . 6 Miscellaneous 6.1 New scenarios . . . . . . . . . . 6.2 More efficient implementation of pathloss generation . . . . . . . 6.3 Further enhancements . . . . . 6.3.1 Outage users . . . . . . 6.3.2 Throughput figures . . .
. . . . . . .
. . . the . . . . . . . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
. . . . . . .
45 45 49 50 50 52 53 56
59 . . . . . . . . . . . . . . 59 . . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
61 62 62 64
7 Conclusions 68 7.1 Simulator enhancement . . . . . . . . . . . . . . . . . . . . . . 69 7.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 7.3 Future enhancements . . . . . . . . . . . . . . . . . . . . . . . 72
iv
List of Figures 1.1
View coverage map about HSDPA deployment in the world, [1].
2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9
Estimated cell throughput per sector, [2]. . . . . . . . . Downlink data rates, [2]. . . . . . . . . . . . . . . . . . Fundamental properties of the DCH and HS-DSCH, [2]. RRM architecture, [2]. . . . . . . . . . . . . . . . . . . HSDPA Node B scheduling principle, [2]. . . . . . . . . HSDPA operation channels. . . . . . . . . . . . . . . . HS-DSCH channel coding chain, [3]. . . . . . . . . . . . 16QAM and 4QAM constellations, [2]. . . . . . . . . . Relative timing between HS-SCCH and HS-DSCH, [2].
. . . . . . . . .
6 7 10 11 12 13 15 15 19
3.1
Three main steps in the simulator: load settings, precalculations (i.e Node-B and users positions in order to prepare the network) and simulation loop to obtain the HSDPA data rate. Network layout with 7 and 19 base stations. . . . . . . . . . . Example of the grid positions generation in the serving cell. . Example of the users position in the serving cell. . . . . . . . . Overview of the basic steps in the simulator. . . . . . . . . . . Average data rates with RNC power control of the HS-DSCH.
25 28 28 29 30 34
3.2 3.3 3.4 3.5 3.6 4.1 4.2 4.3 4.4 4.5 4.6 5.1
. . . . . . . . .
. . . . . . . . .
. . . . . . . . .
Information extracted from the header of the files. . . . . . . Read pathloss from data files given by Mobilkom AG. . . . . Node-Bs positions and respective prediction files. . . . . . . Provisional grid, main BS and sector shape. . . . . . . . . . Last step: user positions located randomly and uniformly within the limits of the generated sector (green points). . . . Diagram of the overall handling real layouts data process. . .
. . . .
2
38 39 41 42
. 43 . 44
High-Speed Dedicated Physical Control Channel that carries the uplink. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
v
5.2
HS-DSCH link adaptation principle: (1) the UE reports lowquality channel information and the Node B allocates a low bit rate; (2) the UE reports high-quality channel information and the Node B allocates a high bit rate, [2]. . . . . . . . . . 5.3 Block diagram showing the received signal at the HSDPA user and report of the CQI to the serving HS-DSCH cell, [4]. . . . 5.4 CQI mapping. . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5 Coarse view of the simulation including slope and shift. . . . 5.6 Different values of the slope for the CQI mapping. . . . . . . 5.7 Different values of the shift for the CQI mapping. . . . . . . 5.8 Throughput as function of the slope value with shift = 0. . . 5.9 Throughput as function of the shift value with slope = 1. . . 5.10 Throughput as function of the slope and shift. . . . . . . . . 6.1 6.2 6.3 6.4 6.5 6.6 6.7 6.8
6.9
Four possible user positions configuration. . . . . . . . . . . Snapshot and exhaustive snapshot scenarios. . . . . . . . . . Fixed angle and fixed distance scenarios. . . . . . . . . . . . Network with 19 base stations and 3 sectors model. . . . . . Variation of the user pathloss with the distance. . . . . . . . Variation of the user pathloss with the angle. . . . . . . . . . HSDPA outage users as function of the power. . . . . . . . . Throughput as a function of the angle with two fixed BS - user positions distances (50 meters in the figure of the left and 250 meters in the right). . . . . . . . . . . . . . . . . . . . . . . Throughput as a function of the distances with two fixed angles between the users and the BS (80o in the figure of the left and 40o in the right). . . . . . . . . . . . . . . . . . . . . . .
vi
. 47 . . . . . . . .
48 53 54 55 55 57 57 58
. . . . . . .
60 60 61 62 63 63 64
. 66
. 67
Abbreviations 16QAM - 16-Quadrature Amplitude Modulation 3G - Third Generation 3GPP - Third Generation Partnership Project AMC - Adaptive Modulation and Coding ARP - Allocation and Retention Priority ARQ - Automatic Repeat Request AWGN - Additive White Gaussian Noise BLER - Block Error Rate BS - Base Station BTS - Base Transceiver Station CDMA - Code Division Multiple Access CmCH-PI - Common Transport Channel Priority Indicator CPICH - Common Pilot Channel CQI - Channel Quality Indicator CSI - Channel State Information DEM - Digital Elevation Models FCS - Fast Cell Selection FCSS - Fast Cell Site Selection FP - Frame Protocol GGSN - Gateway GPRS Support Node GSM - Global System for Mobile Communications HARQ - Hybrid Automatic Repeat Request HSDPA - High-Speed Downlink Packet Access HS-DPCCH - Dedicated High-Speed Physical Control Channel HS-DSCH - High-Speed Dedicated Shared Channel HSPA - High-Speed Packet Access HS-PDSCH - High-Speed Physical Downlink Shared Channel HS-SSCH - High-Speed Shared Control Channels HSUPA - High-Speed Uplink Packet Access IR - Incremental Redundancy ITU - International Telecommunication Union vii
Max C/I - Maximum Carrier to Interference MCS - Modulation and Coding Scheme MIMO - Multiple Input Multiple Output MS - Mobile Station PDP - Power Delay Profile PDU - Protocol Data Unit PF - Proportional Fair QoS - Quality of Service RLC - Radio Link Control RNC - Radio Network Control RR - Round Robin RRM - Radio Resource Management SAW - Stop And Wait SF - Spreading Factor SGSN - Serving GPRS Support Node SINR - Signal to Noise and Interference Ratio SISO - Single Input Single Output SNR - Signal to Noise Ratio SPI - Scheduling Priority Indicator TCP - Transmission Control Protocol TBS - Transport Block Size TTI - Transmit Time Interval UE - User Equipment UMTS - Universal Mobile Telecommunications System UTRAN - UMTS Terrestrial Radios Access Network WCDMA - Wideband Code Division Multiple Access WSS - Widesense Stationary
viii
Chapter 1 Introduction 1.1
Third Generation Services
During the last decades, the mobile communication market evolution has led to demands for higher data rates and larger system capacity. To successfully satisfy these requirements, Third Generation systems must increase their spectral efficiency and support high user data rates, especially on the downlink direction of the communication path due to its heavier load. For this purpose, the 3GPP has standardized in Release 5 a new technology called High Speed Downlink Packet Access (HSDPA) that represents an evolution of the WCDMA radio interface. These technological enhancements can allow operators to enable new high data rate services, improve the QoS of already existing services, and achieve a lower cost per delivered data bit. Consumers are expected to acquire mobile data services if their contents add value to the consumer’s life by satisfying a concrete necessity or requirement. From the end user’s interest, the value provided by the service contents contribute to his cost-effectiveness, time-efficiency, or simple entertainment; for instance, rich content services like video telephony, audio/video clips, and map based information, or fast Internet access for business users.
1.2
Technology
High-Speed Downlink Packet Access, or also known as HSDPA, is a mobile telephone protocol in the High-Speed Packet Access (HSPA) family of third generation (3G) technologies designed to reduce the latency of the link and increase data transfer rates and the capacity of such networks through the transfer of data using a cellular phone. HSDPA is associated with the vari1
ous Universal Mobile Telecommunications System (UMTS) networks. Current HSDPA deployments support down-link speeds of 1.8, 3.6, 7.2 and 14.4 Mbps. The first phase of HSDPA has been specified in the 3rd Generation Partnership Project (3GPP) Release 5. The second phase of HSDPA is specified in the 3GPP Release 7 and has been named HSPA Evolved or also HSPA+; it can achieve data rates of up to 25 Mbps, [1]. As a difference with other WCDMA channels, the High-Speed Downlink Shared Channel (HS-DSCH) lacks two basic features - fast power control and variable spreading factor. Instead, it presents an improved downlink performance by using adaptive modulation and coding (AMC), fast packet scheduling and fast retransmissions at the base station, known as hybrid automatic repeat-request (HARQ), together with a shorter 2-ms Transmission Time Interval (TTI). Figure 1.1 shows the coverage map of deployed HSDPA technology around the world.
Figure 1.1: View coverage map about HSDPA deployment in the world, [1].
1.3
Work environment
Due to the importance of the HSDPA technology, a SISO-HSDPA System level simulator was developed in a collaboration of the Institute of Communications and Radio Frequency Engineering and Mobilkom Austria AG. 2
The source-code is based on MATLAB and it simulates a mixed network in which both UMTS and HSDPA traffics are present. MATLAB is a high-level technical computing language and interactive environment for algorithm development, data visualization, data analysis, and numeric computation. It is an extended engineering tool and has enhanced the tradicional languages. It allows solving technical computing problems and a wide range of applications like signal and image processing, communications, easy matrix manipulation, plotting of functions and data, implementation of algorithms, creation of user interfaces, and interfacing with programs in other languages. The extension of the initial simulator is also developed in MATLAB, to ensure extendability and simple debugging.
1.4
Objectives
Let me briefly sketch the motivation of this bachelor thesis. The first goal of the simulator was to evaluate the HSDPA throughput performance in the mixed traffic network; and based on this groundwork this thesis should develop two enhancements, namely: • Extend the SISO HSDPA Simulator in order to handle real network layout data: this is the main functionality on which the work has been focused; the Mobilkom Austria AG provided measured path-loss matrices of a HSDPA cluster, including parameters like antenna gain patterns, the height of the antennas and the Nodes-B positions. This data has to be converted in a suitable form to be analyzable in the simulator. • CQI mapping optimizations: an optimization of the CQI mapping of HSDPA mobiles shall be performed in order to find the mapping which maximizes the overall cell throughput. This optimization can result in a new mapping at the Node-B that grant an optimum HSDPA network performance.
1.5
Structure of the thesis
This thesis report is organized as follows: • Chapter 1: a short introduction about the technology and work environment, objectives and outline of the thesis is described. • Chapter 2: provides a general overview of the HSDPA technology, like basis and key features. 3
• Chapter 3: this chapter describes the initial SISO-HSDPA Simulator, including the system model and simulation process. • Chapter 4: the conversions of the real data matrices into useful information for the simulator, the overall process and necessary new functionalities and modifications in the initial source-code for handling the new real data layout are explained in this section. • Chapter 5: this chapter outlines the CQI basis in HSDPA, the current CQI mapping and the mapping optimizations investigated in the thesis, as well as the new functions including in the original simulator for this purpose. • Chapter 6: includes a description of some investigations and new functionalities for the enhancement of the simulator that do not fit into the previous chapters. • Chapter 7: summarises the main conclusions of the bachelor thesis.
4
Chapter 2 HSDPA Principles This chapter covers high-speed downlink packet access (HSDPA) principles for wide-band code division multiple access (WCDMA) - the new key feature included in Release 5 specifications and enhanced further in Release 6 specifications. HSDPA has been designed to increase downlink packet data throughput, compared to the rates provided by the Release 4 (also called Release 99) WCDMA specifications by means of fast physical layer (L1) retransmission and transmission combining as well as fast link adaptation controlled by the Node B.
2.1
Introduction
HSDPA, also called 3.5G, is the evolution of the third generation (3G) and is considered the previous step before the fourth generation (4G), the future High-Speed Mobile Network. HSDPA and High-Speed Uplink Packet Access (HSUPA) are the components of the High-Speed Packet Access (HSPA) family. HSPA is an upgrade of the network infrastructure and it is part of the WCDMA 3G network. As an enhancement of UMTS, HSDPA was designed to improve the quality of service, increase the peak data rates (currently speeds supported by HSDPA are 1.8, 3.6, 7.2 and 14.4 Mbps). Also compared to UMTS, the spectral efficiency is significantly increased, and this allows more users being able to use high data rates on a single carrier. The fundamental techniques used in HSDPA to achieve this improvements are Adaptive Modulation and Coding (AMC), extensive multi-code operation and a fast and spectrally efficient retransmission strategy. The assignment of the HS-DSCH (High-Speed Downlink Shared Channel) among the users on 5
a TTI basis (1 TTI = 2 ms) is coordinated by a fast scheduler. Higher cell capacity and higher spectral efficiency are required to provide these higher data rates and new services with the current base station sites. Figure 2.1 illustrates the estimated cell capacity per sector per 5MHz with WCDMA, with basic HSPA and with enhanced HSPA in the macro-cell environment.
Figure 2.1: Estimated cell throughput per sector, [2].
HSDPA is able to satisfy the most demanding multimedia applications such as email attachments, Power Point presentations or web pages. An HSDPA 3.6 Mbps network can provide a 3MB music file in 8.3 seconds and a 5 MB video clip in 13.9 seconds. Speeds achieved by HSDPA reach 14.4 Mbps but currently most network operators provide speeds up to 3.6 Mbps, with the rollout of 7.2 Mbps quickly growing. It is important to note that the total available downlink speed within one sector is split among all the active users. Also, HSDPA can coexist with Release 4 in the same frequency band of 5 MHz. In Austria four HSDPA operators are giving service, Mobilkom Austria, Hutchison 3 Austria and ONE Austria serving HSDPA data rates of 7.2 Mbps, and Mobilkom Austria serving HSUPA (with a data rate of 1.4 Mbps). Currently only Telstra (in Australia) is serving HSDPA data rates of 14.4 Mbps. There are 185 commercial HSDPA networks in 92 different countries; the current deployment of HSPA networks in the world and the 6
HSDPA data rates supported are shown in Table 2.1.
Data rate 0-3.6 Mbps 3.6-7.2 Mpbs 7.2-14.4 Mpbs
Networks 56 91 38
Table 2.1: Current HSDPA commercial networks and data rates, [1].
2.2
HSDPA Standardization
High-speed downlink packet access (HSDPA) was standardized as part of 3GPP Release 5 with the first specification version in March 2002. Highspeed uplink packet access (HSUPA) was part of 3GPP Release 6 with the first specification version in December 2004. HSDPA and HSUPA together are called ’high-speed packet access’(HSPA). The first commercial HSDPA networks were available at the end of 2005, as we can see in Figure 2.2, and many improvements have been introduced in the Release 6, 7, and 8.
Figure 2.2: Downlink data rates, [2]. 7
HSPA is deployed on top of the WCDMA network. Both of them can share all the network elements in the core network and in the radio network including base stations, Radio Network Controller (RNC), Serving GPRS Support Node (SGSN) and Gateway GPRS Support Node (GGSN). WCDMA and HSPA are also sharing the base station sites, antennas and antenna lines. 3GPP creates the technical content of the specifications, based around work items, though small changes can be introduced directly as ”change requests” against specification, but it is the organizational partners that actually publish the work. In addition to the organizational partners, there are also market representation partners, such as the UMTS Forum, part of 3GPP. With bigger items a feasibility study is done usually before rushing in to making actual changes to the specifications, [2]. A feasibility study for HSDPA was started in March 2000 in line with 3GPP principles, having at least four supporting companies. Motorola and Nokia supporting the start of the work from the vendor side and BT/Cellnet, TMobile and NTT DoCoMo from the operator side. The study was finalized for the TSG RAN plenary for March 2001 and there were issues studied to improve the downlink packet data transmission over Release 4 specifications. Physical layer retransmissions and BTS-based scheduling were studied as well as adaptive coding and modulation. The study also included some investigations for multi-antenna transmission and reception technology, titled MIMO (Multiple Input Multiple Output), and also Fast Cell Selection (FCS), [2].
2.3
HSDPA vs Release 4 DCH
In Release 4 specifications basically exist three different methods for downlink packet data operation: dedicated channel (DCH), forward access channel (FACH) and downlink shared channel (DSCH). The most interesting comparison is between Release 4 and HSDPA dedicated channel; the FACH is used either for small data volumes or when setting up the connection and during state transfers. In connection with HSDPA, the FACH is used to carry the signalling when the terminal has moved. However the DSCH has been replaced with the high-speed DSCH of HSDPA. The Release 4 based DCH is the key part of the system, and Release 5 8
HSDPA is always operated with the DCH running in parallel. If the service is only for packet data, then at least the signalling radio bearer (SRB) is carried on the DCH. In case the service is circuit-switched, then the service always runs on the DCH. In Release 5, uplink user data always go on the DCH (when HSDPA is active), whereas in Release 6 an alternative is provided by the Enhanced DCH (E-DCH) with the introduction of high-speed uplink packet access (HSUPA). In the case of multiple services, the reserved capacity is the sum of the peak data rate of the services. The main functionality for the DCH is the fast power control in addition to encoding the data packet provided by the RNC. Furthermore, soft handover is supported for the DCH. As a difference with Release 4, HSDPA introduces some methods for improving downlink packet data in terms of capacity and bit rates. The key differences between the HS-DSCH (HSDPA dedicated channel) and the Release 4 DCH-based packet data operation are as follows: • Lack of fast power control. Instead, link adaptation selects the suitable combination of codes, coding rates and modulation to be used. • Support of higher order modulation than the DCH. With 16-Quadrature Amplitude Modulation (16QAM) the number of bits carried per symbol is doubled in favourable conditions compared to the quadrature phase shift keying (4QAM) in Release 4. • User allocation with base station based scheduling every 2ms, including fast physical layer signalling. With DCH the higher layer signalling from the RNC allocates semi-permanent code (and a spreading factor) to be used. The transmission time interval (TTI) is also longer with the DCH, allowing values such as 10, 20, 40 or 80 ms. (The longest is limited in the specific case of small data rates that have a spreading factor of 512). • Use of physical layer retransmissions and retransmission combining, while with the DCH - if retransmissions are used - they are based on RLC level retransmissions. • Lack of soft handover. Data are sent from one serving HS-DSCH cell only. • Lack of physical layer control information on the HS-PDSCH. This is carried instead on the HS-SCCH for HSDPA use and on the associated DCH (uplink power control, etc). 9
• Multicode operation with a fixed spreading factor. Only spreading factor 16 is used, while with the DCH the spreading factor could be a static parameter between 4 and 512. • The DCH may use both turbo-coding or convolutional coding, while in HSDPA only turbo-coding is used. This was motivated by the fact that turbo codes make it possible to increase data rate without increasing the power of a transmission, or they can be used to decrease the amount of power used to transmit at a certain data rate. • No discontinuous transmission (DTX) on the slot level. The HS-PDSCH is either fully transmitted or not transmitted at all during the 2-ms TTI. The main differences are summarised in Figure 2.3:
Figure 2.3: Fundamental properties of the DCH and HS-DSCH, [2].
2.3.1
Radio resource management architecture
The radio resource management (RRM) functionality with HSDPA and HSUPA is significantly changed compared to Release 4. In Release 4 the scheduling control was purely based in the radio network controller (RNC) while in the base station (BTS or Node B in 3GPP terminology) mainly a power control related functionality (fast closed loop power control) was located. In Release 4 if there were two RNCs involved in the connection, the scheduling was distributed. The serving RNC (SRNC) - the one being connected to the core network for that connection - would handle the scheduling for 10
dedicated channels (DCHs) and the one actually being connected to the base transceiver station (BTS) would handle the common channels. Due to the BTS based scheduling, the overall RRM architecture changed. The SRNC will still retain control of handovers and is the one which will decide the suitable mapping for quality of service (QoS) parameters. With HSDPA the situation is simplified because, as there are no soft handovers for HSDPA data, the utilization of the Iur interface can be avoided by performing SRNC relocation, when the serving high-speed downlink shared channel (HS-DSCH) cell is under a different controlling RNC (CRNC). Thus, just a single RNC could be enough for the typical HSDPA scenario, [2]. Figure 2.4 shows the new RRM architecture.
Figure 2.4: RRM architecture, [2].
2.4
HSDPA operation principle
HSDPA is based on a fast Node B scheduling where the Node B estimates the channel quality of each active HSDPA user on the basis of the physical 11
layer feedback received in the uplink. Scheduling and link adaptation are then conducted on a fast pace depending on the scheduling algorithm and the user prioritization scheme. The general HSDPA operation principle is shown in Figure 2.5.
Figure 2.5: HSDPA Node B scheduling principle, [2].
The other new key technology is physical layer retransmission. In Release 4 when the data has not been received correctly, is necessary to retransmit it again from the RNC. In Release 4 there is no difference in physical layer operation, regardless if the packet is a retransmission or a new packet. With HSDPA the packet is first received in the buffer in the BTS. The BTS keeps the packet in the buffer even if has sent it to the user and, in case of packet decoding failure, retransmission automatically takes places from the base station without RNC involvement. So, the terminal can combine the transmissions, capturing the energy of both. Using a radio link control (RLC)-acknowledged mode of operation, RLC layer acknowledgement is provided in the RLC layer as would be done for Release 4 based operation.
12
2.5 2.5.1
HSDPA channels HSDPA new channels
Several new channels have been introduced for HSDPA operation. For user data there is the high-speed downlink shared channel (HS-DSCH) and the corresponding physical channel. For the associated signalling needs there are two channels: high-speed dedicated physical control channel (HS-DPCCH) in the uplink direction and high-speed shared control channel (HS-SCCH) in the downlink. In addition to the basic HSDPA channel covered in Release 5 specifications, there is now a new channel in Release 6 specifications the fractional dedicated physical channel (F-DPCH) - to cover for operation when all downlink traffic is carried on the HS-DSCH. The channels needed for HSDPA operation are shown in Figure 2.6.
Figure 2.6: HSDPA operation channels.
High-speed downlink shared channel The HS-DSCH is the transport channel that carries the actual user data. In the physical layer the HS-DSCH is mapped onto the high-speed physical 13
downlink shared channel (HS-PDSCH). An important property of the HS-DSCH is that it can dynamically allocate the resource. When the Node-B decides which user is going to be served, the data is sent continuously during the 2-ms TTI, so there is no discontinuous transmission (DTX) on the slot level like with the DCH. With DTX the downlink interference generated is reduced, but it keeps the code resource occupied according to the highest data rate possible on the DCH, because the code resource reservation is not changed when moving to a lower data rate; (the only way to reduce resource consumption is to reconfigure the radio link, but this takes time in reconfiguring the data rate to a new smaller value, and then a new reconfiguration to upgrade the data rate again). As a difference to DTX, with HS-DSCH, once there are no more data to be transmitted for that user, there is no transmission on the HS-DSCH again for the same user, but the resources in the according 2-ms are allocated to another user. Let’s see the important technical apects: • Adaptative Modulation and Coding: To cope with the dynamic range of the signal-to-noise ratio (Es/No) at the UE, HSDPA adapts the modulation, the coding rate and number of channelization codes to the instantaneous radio conditions. The combination of the first two mechanisms is denominated Adaptive Modulation and Coding (AMC). The channel coding is simpler than the corresponding DCH one, because in the HS-DSCH there is no need to deal with DTX or compressed mode, and there is only one transport channel active at a time because fewer steps in multiplexing/de-multiplexing are needed. The HS-DSCH channel coding chain is illustrated in Figure 2.7. – 16QAM : While the DCH only uses 4QAM modulation, the HS-DSCH may additionally use the higher order modulation 16QAM. HS-DSCH incorporates this modulation to increase the peak data rates for users served under favourable radio conditions. Support of 4QAM is mandatory for the mobile, despite the support of 16QAM is optional for the network and the UE. The inclusion of this high order modulation introduces some complexity challenges for the receiver terminal, which needs to estimate the relative amplitude of the received symbols, whereas it only requires the detection of 14
Figure 2.7: HS-DSCH channel coding chain, [3]. the signal phase in the 4QAM case. The turbo encoder is in charge of the data protection. The 16QAM constellation rearrangement depends on the transmission number, because the symbols in the constellation do not have the same error probability. The 16QAM and 4QAM constellations are shown in Figure 2.8.
Figure 2.8: 16QAM and 4QAM constellations, [2].
15
QAM is a digital modulation that transports data by changing the amplitude of two carrier signals. These two waves, generally sinusoidal, are in the same frecuency but with a phase difference of ninety degrees; both signal paths - I and Q - carry information. It is used for the data transmission with a high speed by channels with restricted bandwidth. By having more constellation points 16 instead of 4 - now 4 bits can be carried per symbol instead of 2 bits per symbol with 4QAM. In reception the use of higher order modulation like 16QAM introduces additional decision boundaries, as shown in Figure 2.8. Signal quality needs to be better when using 16QAM instead of 4QAM. Because of this, with 16QAM it is not sufficient to figure out the phase correctly but also the amplitude needs to be estimated for more accurate phase estimate. The channel coefficients can be estimated from the common pilot channel (CPICH), which directly gives phase information. The offset of the HS-DSCH data channel to the CPICH however has to be signalled in order to estimate the amplitude information. This suggests that at the base station - during the 2-ms transmission power changes should be avoided. In the system there can be other traffic that is consuming code space as well - such as circuit switched speech or video calls - which cannot be mapped on HSDPA. Thus, radio resource management will then determine the available code space for the scheduler at the BTS, [2]. – Bit scrambling: The bit scrambling functionality was introduced to avoid long sequences repeating the same symbol, as long sequences of ’0s’ or ’1s’. These could occur with some type of content, and especially when not using ciphering at higher layers. In such a case the terminal would have difficulties with HS-DSCH power level estimation and, thus, physical layer scrambling operation was introduced. Operation is the same for all users and is purely for ensuring good signal properties for demodulation, [2].
16
– HS-DSCH Link Adaptation HSDPA utilizes link adaptation techniques to substitute power control and variable spreading factor. The HS-DSCH link-adaptation algorithm at the Node-B is very dynamic, and adjusts the transmit bit rate on the HS-DSCH every 2-ms TTI. It is based on the physical layer CQI being provided by the terminal. Various sources contribute to the time-variant SINR at the user even though the HS-DSCH transmit power is assumed to be constant. The total transmit power from the serving HS-DSCH cell is time variant due to the transmission of the power controlled DCHs; the downlink radio channel is time variant if the user equipment is somehow moving; and finally, the experienced inter-cell interference at the user position is also time variant. For the purpose of HS-DSCH link adaptation, the user therefore periodically sends a CQI to the serving HS-DSCH cell on the uplink high-speed dedicated physical control channel (HS-DPCCH), [4]. Using link adaptation, the network will also gain from the limitation of power control dynamics in the downlink. As signals in the downlink cannot use a too large dynamic range to avoid the near-far problem between signals from the same source, the downlink power control dynamics is also limited. While in the uplink a 71-dB or more dynamic range is used, in the downlink only around 10 to 15 dBs can be utilized. The exact number depends on the implementation, channel environment and spreading factors applied. This means that for users close to the base station the power level transmitted is higher than necessary for reliable signal detection. Using link adaptation, there is a difference of a few decibels in the signal strength, just by changing from 4QAM to 16QAM; and by playing with the coding rates and the number of codes the total dynamic range can reach 30 dB. • Hybrid ARQ : HSDPA incorporates a physical layer retransmission functionality that adds robustness against link adaptation errors and improves the performance significantly. The Hybrid ARQ functionality consists of a two stage matching functionality which allows tuning two different retransmission types. These 17
two Hybrid ARQ strategies are: (1) identical retransmissions (also called soft combining) or (2) non-identical retransmissions (or so-called ’incremental redundancy’). The Hybrid ARQ technique is fundamentally different from the WCDMA retransmissions because the UE decoder combines the soft information of multiple transmissions of a transport block at bit level. Let us go a little bit more into detail: Soft combining : as proposed in [5] every retransmission is simply a replica of the coded first transmission. The same bits after rate matching operation are sent, for every retransmission of the same packet. The decoder at the receiver combines these multiple replicas of the transmitted packet weighted by the received SNR prior to decoding (so called ”soft combining”). This technique requires some memory on the mobile terminal, which must store the soft information of unsuccessfully decoded transmissions. The delay in the retransmissions and memory required shell be as small as possible. Incremental Redundancy (IR): it requires even more memory in the receiver user equipment capabilities. The retransmissions include additional redundant information that is incrementally transmitted if the decoding fails on the first attempt. That causes the effective coding rate to be increased with the number of retransmissions. Incremental Redundancy can be further classified in Partial IR and Full IR. Partial IR includes the systematic bits in every coded word, which implies that every retransmission is self-decodable, whereas Full IR only includes parity bits, and therefore its retransmissions are not self-decodable. If due to a signalling error that could fill the buffer with undesired data, due to a low coverage, or due to a change of the serving HSDSCH cell, the number of physical layer retransmissions exceeds the maximum or the retransmissions fail, the radio link layer will handle further retransmissions. High-speed shared control channel (HS-SCCH) The HSDPA concept includes a Shared Control Channel (HS-SCCH) to signal the users when they are going to be served as well as the necessary information for the decoding process. Compared with the HS-DSCH, the HS-SCCH has two slots offset, as shown if Figure 2.9. This enables the HSSCCH to carry time-critical signalling information which allows the terminal to demodulate the correct codes. A spreading factor of 128 allows 40 bits 18
per slot to be carried (with 4QAM modulation). The phase reference does not change when using HS-DSCH due to the lack of pilots or power control bits on the HS-SCCH.
Figure 2.9: Relative timing between HS-SCCH and HS-DSCH, [2].
The HS-SCCH carries the following information, [6]: • UE Id Mask: to identify the user to be served in the next TTI. • Transport Format Related Information: specifies the set of channelization codes, and the modulation. The actual coding rate is derived from the transport block size and other transport format parameters [2]. • Hybrid ARQ Related Information: such as if the next transmission is new or related to an earlier transmitted packet, and if it should be combined, the associated ARQ process, and information about the redundancy version, [7]. This control information solely applies to the UE to be served in the next TTI, which permits this signalling channel to be a shared one. The RNC can specify the recommended power of the HS-SCCH (offset relative to the pilots bits of the associated DPCH), [8]. The timing between the HS-SCCH and the HS-DSCH allows the terminal to have one slot time to receive the information which codes have to despread and with which to modulate. For the remaining parameters, a slot processing time is needed before a new 2-ms TTI starts.
19
When HSDPA is operated using the time multiplexing principle, then only one HS- SCCH can be configured. In this case only one user receives data at a time. When there is code multiplexing, then more than one HS-SCCH is needed. A single terminal may consider at most four HS-SCCHs; the system itself could configure even more. The use of code multi-plexing is not necessarily needed either when the carrier is shared with DCH traffic, or when there is a desire to have HSDPA data users operating with reasonable data rates -in the order of 384 kbps or more. In general, the data rate available for each user in different cases will depend on power allocation, the environment and the type of terminal being used. The channel coding is one-third convolutional coding (as turbo-coding does not make sense with such a small amount of information). In the second part there is a cyclic redundancy check (CRC) to make sure that there is no corruption of the information. A signalling error with, say, an HARQ process number would cause problems as it would cause buffer corruption; thus, a 16-bit CRC is used to ensure sufficient reliability, [2].
2.5.2
High-speed dedicated physical control channel
An uplink High Speed Dedicated Physical Control Channel (HS-DPCCH) carries the necessary control information in the uplink, namely, the ARQ acknowledgements, and the Channel Quality Indicator (CQI) reports. The CQI reports are deeply described in Section 5. To aid the power control operation of the HS-DPCCH an associated Dedicated Physical Channel (DPCH) is run for every user. This information from the terminal to the base station allows for the link adaptation and physical layer retransmissions. According to [8], the RNC may set the maximum transmission power on all the codes of the HS-DSCH and HS-SCCH channels in the cell. Likewise, the RNC determines the maximum number of channelization codes to be used by the HS-DSCH channel. By keeping the existing uplink DPCCH and DPDCH unchanged the active set can also accommodate Release 4 based base stations. The initial uplink DPCCH transmit power is set by higher layers. Subsequently the uplink transmit power control procedure simultaneously controls the power of a DPCCH and its corresponding DPDCHs (if present). The relative transmit power offset between DPCCH and DPDCHs is determined by the network. Any change in the uplink DPCCH transmit power shall take place immediately before the start of the pilot field on the DPCCH. The change in DPCCH 20
power with respect to its previous value is derived by the UE. The previous value of DPCCH power shall be that used in the previous slot, except in the event of an interruption in transmission due to the use of compressed mode, when the previous value shall be that used in the last slot before the transmission gap. During the operation of the uplink power control procedure the UE transmit power shall not exceed a maximum allowed value which is the lower out of the maximum output power of the terminal power class and a value which may be set by higher layer signalling. Uplink power control shall be performed while the UE transmit power is below the maximum allowed output power, [9]. As already mentioned, the uplink feedback information is carried on the HSDPCCH. The HARQ feedback informs the base station whether the packet was decoded correctly or not. The CQI, respectively, tells the base station scheduler the data rate the terminal expects to be able to receive at a given point in time. Fractional DPCCH For Release 6, further optimization took place for the situation where only packet services are active in the downlink other than the signalling radio bearer (SRB). In such a case, especially with lower data rates, the downlink DCH introduces too much overhead and can also consume too much code space if looking for a large number of users using a low data rate service (like VoIP). The solution was to use an F-DPCH, which is basically a strippeddown version of DPCH that handles the power control. The code resource is time-shared, thus several users can share the same code space for power control information. Each user sees only the channel which has one symbol per slot for transmission power control (TPC) information and assumes there is no transmission in the rest of the symbols. With several users, the network configures each user having the same code but different frame timing and, thus, users can be transmitted on the single code source. Up to ten users can share one SF 256 code, thus reducing code space utilization for the associated DCH for users with all services mapped to the HS-DSCH,[2].
21
Chapter 3 HSDPA Simulator In this chapter the initial simulator on which this bachelor thesis is based and all the work that was developed is explained. It contains the information about the initial simulator created by M.Wrulich, et al., in which I had to implement some new functionalities for the enhancement of the simulator. The chapter is organized as follows: Section 3.1 briefly sketchs the purpose and context of the simulator; a short description of the system-level model for the investigation is included in Section 3.2 and Section 3.3 explains the structure of the simulator.
3.1
Introduction
Let me briefly introduce the work environment and goals of the initial SISO HSDPA Simulator. In the program, written in MATLAB, a mixed UMTS and HSDPA network is simulated. As described in Chapter 2, one of the advantages in a WCDMA network is that HSDPA can coexist within an existing 5MHz band of Release 4, allowing for sharing the power amplifier and spreading codes at the Node-B between the HS-DSCH and the Release4 dedicated channels. Even if HSDPA is widely installed, a mixed carrier operation is more cost-efficient for cells that are not fully loaded, [10]. One of the goals of the initial simulator was to deduce the optimum Node-B power split within the mixed scenario, by means of snapshot based network simulations, in order to maximize the overall cell throughput. The obtained results can be used for the cell operation planning by the network operators.
22
3.2
System model
According to [10], the HSDPA performance is done with a system-level simulator in which the following aspects are modelled: • Channel modeling: the channel coefficient used for the simulation allows us to model the radio propagation and it considers the macroscale pathloss, that depends on the the distance between the base station and the user equipment, the shadow fading and the fast fading with multiple paths and no time correlation. • HSDPA modeling: this modeling represents the HSDPA transmission performance in an accurate way by using a simplified system level description in which they have modelled the channel quality observed by the user equipment and the bit error/decoding performance. The models are the so-called link-measurement model and link-performance model respectively. In the link-measurement model, the signal-to-noiseand-interference ration (SINR) is evaluated after Rake-combining and despreading for each user equipment in the cell. The link-performance model aims at an analytical approximation of the block error ratio (BLER), where it is assumed that the scheduler in the Node B decides to serve a specific user with an MCS as specified by the CQI mapping table [11]. Furthermore, it is assumed that the desired HSDPA user always gets the full available transmission power and there is enough data to transmit (full buffer assumption). • Release 4 modeling: the Release 4 traffic is only coarsly modelled, since the main goal of the simulator was the prediction of the achievable user data rates with HSDPA within the existing Release 4 network. In [10] further information and the estimation of the number of DCH users that can be served can be found. • Power split: the total intra-cell transmission power depends on the transmission power of the DCH and the HSDPA traffic, and in each cell, the total available transmit power is shared between DCH and HSDPA users. The power of HSDPA can be allocated in the Base station downlink power budget by means of two possibilities: – By sending Node-B application part (NBAP) messages to the base station, the RNC can dynamically allocate the HSDPA power. This is kept at a fixed level by the base station, and the DCH power varies accordingly to the fast closed loop power control. 23
– The base station is allowed to allocate all unused power for HSDPA, instead of sending NBAP messages. The total intra-cell transmission power is calculated as folows:
Pintra = PDCH + PHS−DSCH + Pother
(3.1)
Pother incorporates the power from the common pilot channel and other needed common channels. This explains the fact that the total intracell power depends on the DCH and HSDPA transmission power.
3.3
Simulation process
After explaining briefly the system model of the simulator, now the initial SISO HSDPA Simulator itself is going to be treated in detail. Figure 3.1 illustrates the three main steps: • Load settings: before starting the process, the function ”load settings” is called. This settings file allows for the specification of the simulation, i.e. the kind of network, channel and user equipment. • Precalculations: after the settings are loaded and before the simulation starts, there are some precalculations which allow us creating some necessary elements like user and Node-B positions, all the pathlosses for every users and PDP for serving links. • Simulation loop: the last step is the simulation loop, which calculates different average data rates by means of multiple independent snapshots.
3.3.1
Load settings
The settings are divided in: network, channel, user equipment and simulator, and by changing the parameters we can specify the kind of simulation. • Network settings: the mixed traffic network is modelled according to the parameters:
24
Figure 3.1: Three main steps in the simulator: load settings, precalculations (i.e Node-B and users positions in order to prepare the network) and simulation loop to obtain the HSDPA data rate. – R’99: contains the needed parameteres for Release ’99, like the bandwith (5 MHz), the chiprate, the UMTS load in percent and the UMTS required Eb/No for the requested UMTS DCH bearer. – Node-B: here, some variables like the distance between Node-Bs, the power level of each Node-B (the maximum power, the CPICH power and the common power), and the power distribution of the Node-B are specified. – Power distribution: determines the power distribution among the neighboring Node-Bs, thus specifying the intercell-interference structure. – HSDPA: in this part the HSDPA network is specified, thus the number of HSDPA users, the spreading factor of HSDPA transmissions (fixed at 16), the number of codes, the absolute HSDPA 25
power and the TTI value (usually 2 ms) are assigned. – MAC-hs: this is used in an enhanced version of the simulator for scheduling variables. – Other: variables like the grid density, which determines the number of grid points for user positioning within the cell, the G factor of the network or the other cells interferences are chosen here. – Network structure: the number of base stations (7 or 19), the number of sectors for each base station (1 or 3) and the antenna gain pattern are specified. • Channel settings: the channel modeling is done through three different fadings that can attenuate the signals in the communications between the base station and the users. – Deterministic fading: first the model (COST231, Berger, fixed, exponent, tr25848 or none) is selected, and accordinglyly the necessary variables like frequency or antenna height are assigned. – Shadow fading: there are two possibilities for the shadow fading: the lognormal model or the lognormal moving model, and accordinglyly different parameters can be selected. – Fast fading: it can be modeled with a Rayleigh model or it can be omitted. – Power Delay Profile: the oversampling factor, the model (pedestrian A or B, vehicular A or B, or none) and the chiprate are selected here. • User Equipment settings: the parameters needed to model the user equipment are defined here. – General: the user category class and different noise powers seen in the receiver, like the receiver noise figure or the thermal noise density are defined. – Movement: contains the setting for speed of the user. – Receiver: the reciver type is specified, and in case of a Rake receiver the number of fingers is also determined. – Traffic: will be used in further developements of the simulator. • Simulator settings: now, some parameters of the simulator are defined, although they do not belong to the communication scheme itself. 26
– Simulation type: the kind of simulation can be chosen here. It is possible to simulate all the possible users positions (also called grid positions) uniformly, or it can be done by means of multiple simulations based on multiple snapshot, in which we simulate as many positions as number of users we have set. – Link performance model: the COST290 model, where a simple link performance model based on cost 290 or no model can be used, but a model is needed to performance the link and simulate a real situation with a BLER value different to zero. – R’99 datarate model: a simple data rate model described in [10]. – Power distribution: settings for the step-size of the power loop in the simulator. – Display/save results: backup options.
3.3.2
Precalculations
As already said, when the parameters which determine the kind of simulation and the variables are loaded, some precalculations have to be done to prepare the network before the simulation loop starts. The Node-B positions, user positions, PDP for links in serving site, and the pathlosses for every users are created here. Some of these calculations have been enhanced in the improved version of the simulator developed in this thesis. 1. Node-B positions: the network consists of one central hexagonal cell and six or eighteen hexagons more around it, depending on the desired number of interfering Node-Bs. The base stations are situated in the center of each hexagon, and the serving Node-B is in the central hexagon, so it is in the middle of the network. This configuration is described as cellular layout 2 in [12]. Both configurations are shown in Figure 3.2. 2. Users positions: the simulator determines the users postions by using a grid of points inside the main sector (it is considered the first sector in the serving node), as we can see in Figure 3.3. We can observe in this figure that some part in the bottom of the sector is missing. It corresponds to a 10 square meters area around the base station which is considered as a limit where no users are positioned. The users are positioned randomly in the grid positions. The number of grid points and users is detetermined in the HSDPA traffic simulation settings. This is shown in Figure 3.4. 27
Figure 3.2: Network layout with 7 and 19 base stations.
Figure 3.3: Example of the grid positions generation in the serving cell. 3. Power Delay Profile (PDP): gives the intensity of a signal received through a multipath channel as a function of time delay. The time delay is the difference in travel time between multipath arrivals. In this part, the PDP profile of a given ITU model and for all links in the serving site is generated. 28
Figure 3.4: Example of the users position in the serving cell. 4. Users pathloss: in this part, the pathlosses from all Node-Bs in the simulated network structure to the given users are generated. As briefly explained in the channel modeling and channel settings, the radio propagation model used in the simulator considers three different elements: macro-scale pathloss, shadow fading, and small-scale fading with multiple paths and no correlation in time, since this initial simulator is snapshot based with no correlations in between, [10]. The channel coefficient which models the radio propragation between the base station and the user equipment can thus be written as:
h(τ ) = d · s ·
L X √
pl · fl · δ(τ − τl )
(3.2)
t=1
where d denotes the deterministic pathloss, s the shadow fading, pl and τl are the relative power and delay of the multipath components, fl represents L independent Rayleigh fading processes at fixed time slots and δ denotes the Dirac function. The macro-scale pathloss depends on the distance between the BS and the UE, which is modelled accordingly to the COST231 model [13], and it depends also on the antenna gain pattern if a sectorized model is used. The shadow fading is modelled by a lognormal random variable with zero mean and σs = 8 dB, with no correlation in time. [10] 29
5. Noise power: settings needed to evaluate the noise power level.
3.3.3
Simulation loop
After the settings are loaded and the precalculations are done, the system is prepared to start the simulation of multiple independent snapshots. A configuration with three sectors model has been used, and the simulations are done in the first sector of the serving Node-B, the so-called target sector. The results are getting after the average of all individual calculations. Figure 3.5 illustrates an overview of the HSDPA calculations explained below, where we can observe that the SINR evaluation is the first step, then the CQI is mapped as function of the SINR, and using both, the block error rate (BLER) and the transport block size (TBS), the HSDPA and Release 4 data rate are estimated.
Figure 3.5: Overview of the basic steps in the simulator. • HSDPA System Level Modelling: Let us now go a little bit into detail onto the HSDPA system-level modeling. 30
1. SINR: The Signal-to-noise-interference ratio (SINR) is used to evaluate the channel quality as observed by the receiver, where a standard single antenna Rake is used because it is the most common in the SISO HSDPA terminals. The SINR is calculated after Rakecombining and despreading for every users in the sector, according to the expression:
SIN Ru =
NF X i=1
PHS−DSCH · |hi |2 γ Pintra,residual + Pinter + Pnoise SF ·
(3.3)
where u is ther user, SF represents the spreading factor, PHS−DSCH denotes the power used for the HS-DSCH, Pintra,residual is the residual intracell interference in the downlink, Pinter denotes the transmitted interfering power from the neighbouring base stations, Pnoise is the noise power seen at the receiver, and γ represents the number of assigned spreading codes. The residual intracell interference arriving at the receiver fromPthe serving base station is given by [14], Pintra,residual = Pintra · Ll=1 |hl |2 , where L is the total number of taps of the current realisation, denoted by hl , and Pintra denotes the total power transmitted in the serving cell, [10]. The structure of the Rake reciver is implicitly shown in Equation 3.3, since in the numerator the useful signal power is added up, which is cancelled out from the interference power in the denominator for all the NF available fingers consecutively. The receiver weights and the location of the fingers can be chosen perfectly since perfect channel state information (CSI) is assumed at the receiver. Accordingly, only the squared absolute values of the channel coefficients (for each tap), |hi |2 occur in the equation. It is also assumed that the transmission power of the HS-DSCH is divided equally among all HS-PDSCH.
31
2. SINR to CQI: The next step in the simulator is to calculate the CQI value for a given SINR. This is done via a linear mapping, i.e. SIN R ≤ −3.5 0, CQI = SIN R[dB] + 3.5, −3.5 < SIN R < 26.5 30, SIN R ≥ 26.5
(3.4)
The CQI values (ranging from 0 to 30) are used by the link adaptation algorithm at the Node-B. Each value represents a specific combination of the transport block size (TBS), the number of codes and the modulation type. Thus, each value indicates the maximum TBS that can be correctly received with 90% probability.
3. CQI to TBS: The bit-error/decoding performance, also called link-performance model, can be described once the channel quality is known. The simulator uses the link performance modelling for the transport formats of each mobile category class, given by the range of possible CQI values. The tables for each category, used to determine the number of codes as a function of the CQI and the Transport Block Size (TBS) can be found in [11]. In the simulator we used the UE categories 1 to 6, and the table is shown in Chapter 5.
4. BLER: The Block Error Ratio (BLER) is calculated according to an analytical model specified in [15]. Due to the snapshot based simulation approach, no HARQ retransmissions gains are modelled. The BLER is considered directly in the evaluation of the throughput. According to [16], the BLER under AWGN conditions and utilizing a standard Rake receiver together with turbo coding, can analitically be well approximated by: 32
√R−1.03CQI+5.26 2 SIN
BLER = [10
3−log10 (CQI)
−1
+ 1] 0.7
(3.5)
5. HSDPA data rate: After all the steps done in the simulation, the HSDPA user data rate can be calculated. It is important to note that the TBS denotes the maximum amount of data that can be transmitted via the network in one TTI of 2 ms to the UE without exceeding a BLER of 0.1 in average, and accordingly the HSDPA user data rate is calculated with Equation 3.6, Ru = T BS ·
1 · (1 − BLER) 2ms
(3.6)
which is consecutively averaged over fading realizations and over all the individual snapshot simulations to get the average HSDPA cell and user data rate. • Release 4: As explained before, the investigations conducted so far were focused on the prediction of the achievable HSDPA user data rates in dependence of a given Release 4 DCH load in the cell, and because of that the Release 4 traffic is modelled only coarsly. This prevents an exact evaluation, and the possibility of predicting the R’4 cell throughput at the actual power level is very limited. The simulator just roughly estimates the total DCH cell throughput in order to be able to predict the overall cell throughput as function of the power distribution. Further information can be found in [10].
3.3.4
Results
A short description of the simulation results concludes the inicial simulator explanation. The investigation identifies the optimun power setting in order to maximize the total cell throughput. 33
Figure 3.6 shows the achievable average cell throughput on DCH and HSDPA, and the total cell throughput (the sum of the throughput on DCH and HSDPA) versus the amount of power that is allocated for HSDPA traffic. The simulation parameters are as follows: Release 4 load of 20%, Node-B distance of 0.5 km, Pedestrian A model, 10 codes used for HS-DSCH and the user category 6. The HSDPA cell throughput increases when more power is allocated to HSDPA, while the DCH throughput decreases as there will be less power for the transmission of these channels. It is observed that the HSDPA power allocation that maximizes the total cell throughput is around 4 W, which results in the total cell throughput of 735 kbps, with around 600 kbps being carried on HSDPA with PHS−DSCH = 6.078 W. Note that the HSDPA user data rate starts softly decreasing when the power reaches 6 W because an increase in power results also in an increment of the intercell interference. The R’99 cell data rate is always decreasing, due to the continuous increasing interference seen by the transmissions.
Figure 3.6: Average data rates with RNC power control of the HS-DSCH.
34
In practice, the optimal HSDPA power setting also depends on the offered traffic in the cell and the mixture of DCH- and HSDPA-capable UEs. The gain in the cell throughput from introducing HSDPA mainly comes from a higher spectral efficiency for the HS-DSCH over the DCH by using Hybrid ARQ and adaptive modulation and coding, multiuser diversity gain from using fast PF scheduling, and better utilization of the available cell transmission power. Further information about the optimizations can be found in [10].
35
Chapter 4 Real Network Layouts One of the main goals of the work was to extend the SISO HSDPA Simulator in order to handle real network layout data that the Mobilkom Austria AG provided by means of measured pathloss matrices of a HSDPA cluster, and parameters like the height of the antennas and the Nodes-B positions. It was necessary to convert this data in a useful form to be analyzable in the simulator. Furthermore, the source-code was enabled to use the data files after the necessary modifications. The conversion of the real data matrices into suitable information for the simulator, the overall process and necessary new functionalities and modifications in the initial source-code for handling the new real data layout are described in this chapter.
4.1
Introduction
This chapter leads with the use of geographical information in mobile radio communications from a propagation perspective. Some interest in the COST231 project has focused on the types, resolution and accuracy of digital terrain databases required for propagation modelling. Despite it is not possible in this thesis to list the contents of the files (the accurate location, tilt and azimuth values of the antennas) since this is confidential information, a brief description from [17] about acquisition of geographical data is depicted below. Traditionally, geographical information has been obtained from paper maps. In the last decade increasing use has been made of high resolution remote sensing (aerial and satellite) for acquisition and of digital sto-rage and distribution methods. The generation of Digital Elevation Models (DEM) and the efficient and accurate extraction of radial data from them is reviewed in 36
[18]. Indoor propagation modelling possess the heaviest reliance on high resolution geographical information. For urban propagation, it is essential to have accurate information at least about the average height of individual buildings, when modelling larger cells or performing interference calculations and when terminals are operating close to roof-top height. The incorporation of information about clutter, particularly vegetation, is very important since propagation characteristics are quite sensitive to scatterers around terminals. Aerial stereo photography provides a means of obtaining quite accurate data on the heights and outlines (resolution of the order 1m) of building and terrain features, the location of vegetation, etc. The wealth of data can, in principle, be extracted from these sources. However, the extraction of data is a quite intensive labour. Information about the building cladding, windows, etc. is more difficult to obtain. It appears that accurate geographical information of this nature must be obtained using video cameras to capture data, for example. Wall properties are particularly important for estimating building penetration. For outdoor propagation, in practice we may only need this level of detail around potential BTS sites. Considering the influence of database information on prediction accuracy it is noted in [19] that prediction errors can be attributed to database inaccuracies arising from the omission of vegetation data and the poor resolution of terrain height data. Probably one of the most important causes of the effect of database errors in field estimations is the effect of database inaccuracies on model evaluation. Complex models which visually correspond to the measured data often display a large error standard-deviation with respect to empirical models because of spatial offsets. These offsets can arise if there are small database or measurement location errors. The existence of these offsets between predictions and measurements is not necessarily an indication of a poor model. This problem is typically addressed by separately comparing the locally averaged model prediction and the statistics of the faster variations separately with the measured data.
37
4.2
Files structure and information
The package provided by Mobilkom AG consists on twelve ”prediction files”; each one corresponding to a specific Base Station situated in the city center of Vienna. All of them represent a map of a central area of the city, with the respective Node-B situated in the center, and the generated pathloss around it for different distances. The files are divided in header and pathlosses values. Each value of the pathloss that is read in the file corresponds to a pixel in the map. Establishing this correspondence is possible by using the information in the header of the files, since Mobilkom AG also provided the necessary information to manage it, such as the lenght in bytes of all the interesting parameters. Figure 4.1 shows the useful information that we can extract from the header.
Figure 4.1: Information extracted from the header of the files. By using this information we are able to include the real pathloss in the simulator as well as the Node-B position, grid position and users positions accordingly to the extracted parameters. Besides the mentioned files, there are two files more including information about the base stations identifier and streets where are situated, and the antennas identification (class, tilt, azimuth and power). These files are not used so far in the simulation, although 38
it is possible to read them and extract information by means of specific functions that can be enabled in the load settings. Once we have read all the information, we are able to represent the real pathloss extracted from the files. Figure 4.2 shows the read pathloss values from one of the prediction files. This figure gives an idea about the kind of information attached in the files. By using the new implemented functions, we are able to convert ASCII data into suitable information that we can plot. The figure represents the variation of the pathloss with the distance in the city of Vienna, with the Base Station located in the center of the map. As observed with the different colors in the figure, in the surroundings of the Base Station the pathloss is low and increasing with the distance.
Figure 4.2: Read pathloss from data files given by Mobilkom AG.
4.3
Network setting based on real layout data
The overall simulation process for setting a suitable network in the case that the real data is used is explained below. To achieve a global source code 39
which involves both functionalities (handling real data layout or theorical pathloss calculation), the structure of the initial MATLAB code has been kept. Besides some necessary new functions, the functionalities already existing but adapted for the new purpose have been integrated in the original source code. Thus, the simulation mode can easily be chosen just by setting the according variable at the beginning. Let us now list the main steps of the initial procedures of the simulator: 1. Reading data: as mentioned before, the option for reading the data files can be enabled in the load settings function. In such case, the function xf read out data works on the prediction files and extracts the header of the files, with the information before explained, and the information about the pathloss. In the case that reading also the two left files is desirable, this function calls also read bts csv and read site csv in order to extract their respective information. For the overall process, each parameter is stored in a structure after calling these functions in order to be used next by the necessary, pertinent functions during all the simulation. 2. Node-B positions: following the same process than the initial simulator, the function which locates the Node-Bs is called first. In this function, the base stations are situated accordingly to the stored information extracted from the pedestrian files, or it can be done in the original way (hexagonal layout). Figure 4.3 shows the positions of the twelve Node-Bs belong to the twelve prediction files. Notice that some coordenates are taken by two stations; this is because the position is the same but not the height. 3. Provisional grid positions: the next step is to generate the grid positions where the users will be situated. As occured with the NodeBs, the new grid is calculated in the original function accordingly to the active option. If the real data are being used, we generate a provisional grid around the center of the global map, where the strongest Node-B (which is the one providing the lowest pathloss possible) is situated. This is explained graphically in next steps. 4. Pathloss: in the initial simulator, the pathloss was theorically calculated distinguishing between deterministic pathloss as function of the UE-NodeB distance and as function of the antenna gain pattern (calculated accordingly to the COST 231 Hata model [13]), and stochastic pathloss, based on shadow and fast fading. 40
Figure 4.3: Node-Bs positions and respective prediction files.
In this step of the simulation, only the deterministic pathloss as function of the antenna gain is calculated since the rest of the contributions are included in the prediction files. It is generated for each one of the three sectors as in the first version of the simulator. If the pathloss is set to be read-in from the prediction files, we extract it from the prediction files by using a function which extracts the information according to the structure of the files. This function reads and stores the pathloss for every pixels in each one of the twelve virtual maps. Next, we extract the pathlosses corresponding to the provisional grid created before, and it is summed up to the antenna gain (in dB) calculated in order to obtain a final pathloss in all the grid positions. Finally a 3D-matrix with the final pathloss for every grid position, sector, and Node-B is generated. 5. Find main BS and sector: once we have the pathloss matrix, the next step is to find the main sector and Node-B, which means the NodeB and sector where the lowest pathloss occured (within the region of 41
interest), and that way we can set a Node-B and pathloss information for the simulation process. For this purpose, there is a function which reads the value of every position in the 3D-matrix, and extracts for every pixel which one of the twelve files (Nodes-B) and sector generates the minimum pathloss. This information is kept in a new matrix and then we read every position of it and we count which Node-B and sector is repeated more times in that matrix (these will be the main BS and sector for the simulation). 6. Generate final grid positions: this step fixes the grid positions that can be generated. The new grid consists of those positions of the provisional grid, where the lowest pathloss was generated by the main BS (target sector) simultaneously. Figure 4.4 shows the situation so far, where the provisional grid in blue, the main Node-B and sector and finally the definitive grid in green (which defines the sector shape) are illustrated.
Figure 4.4: Provisional grid, main BS and sector shape. 7. Generate users positions: this is the last step in the new simulation process. The final goal of the process is setting up a complete network 42
suitable for simulation. Once the grid positions are defined, users are located randomly (uniformly) using the same functions as utilized in the initial version. An example in which a zoom of the previous figure after the users are generated (plotted in black) is shown in Figure 4.5.
Figure 4.5: Last step: user positions located randomly and uniformly within the limits of the generated sector (green points). For a better visualization, Figure 4.6 shows a general overview of the network setting process when the real data layouts are handled, summarizing the steps before explained in a diagram.
43
Figure 4.6: Diagram of the overall handling real layouts data process.
44
Chapter 5 CQI Mapping Optimizations This chapter explains more deeply the CQI mapping in HSDPA performance. Section 5.1 includes an introduction and the CQI basis. In Section 5.2 the mapping proposal by the 3GPP specifications is explained, and Section 5.3 describes the enhancements in the SISO HSDPA Simulator, where the object of this part of the work and the achieved results are explained.
5.1
CQI Basis in HSDPA
The Channel Quality Indicator (CQI) is the key indicator for HSDPA downlink channel quality. The CQI value that the terminal reports is a function of: multipath environment, terminal receiver type, ratio of the interference of the own base station compared with others and expected BTS HSDPA power availability. Thus, the value reported does not just correspond to the signal to interference ratio (SIR) that the terminal is experiencing. When a UE reports a particular CQI, it is reporting that under the current radio conditions (including the current power settings), the UE is able to receive the reported CQI and lower CQI’s, at single-transmission BLER no greater than 0.1. User equipments report CQI and HARQ-ACK feedback information to the Node-B through the uplink HS-DPCCH. The clear benefit of the approach is that the solution defined will automatically accommodate the various possible receiver implementations and environment variations and, thus, gives an indication of the best data rates needed by the terminal to cope with the environment in question. This removes the need from the network end to have to consider the delay profile characteristics of the cell/sector, [2]. Possible implementatios of the CQI evolution can be found in [20]. 45
The HS-DPCCH carries uplink feedback signalling related to the downlink HS-DSCH transmission. This signalling consists of HARQ-ACK and CQI as shown in Figure 5.1. Each 2-ms subframe, like those of the downlink physical channels, consists of three slots, each with 2560 chips. The HARQ-ACK is carried in the first slot of the HS-DPCCH subframe and the CQI in the second and third slots. A system parameter k controls the CQI transmission frequency, and also control repetition for both slots is controlled by a separate parameter. Repetition over multiple 2-ms periods is needed in some cases, like cell edge operation when the available power would not ensure sufficient quality for feedback reception, [2].
Figure 5.1: High-Speed Dedicated Physical Control Channel that carries the uplink.
Furthermore, we can explain some cases in which the CQI is used for the performance of HSDPA, say: • Link adaptation: as already said, the CQI is used for the dynamic HS-DSCH link adaptation, in addition to the scheduling decision. The user equipment sends every 2-ms TTI a CQI to the serving HS-DSCH cell on the uplink HS-DPCCH. By means of the CQI, the maximum 46
TBS that can be received correctly with at least 90% probability is indicated. The range of CQI values is 0-30, and each step corresponds appoximately to a 1-dB step in HS-DSCH SINR, [9]. The HS-DSCH link adaptation algorithm at the Node B adjusts the transmit bit rate on the HS-DSCH every TTI when a user is scheduled for transmission. Ideally, the HS-DSCH transmit bit rate should be adjusted as a function of the per-TTI HS-DSCH signal-to-interference plus noise ratio (SINR) experienced at the user end. Figure 5.2 illustrates the general principle HS-DSCH link adaptation, [2].
Figure 5.2: HS-DSCH link adaptation principle: (1) the UE reports lowquality channel information and the Node B allocates a low bit rate; (2) the UE reports high-quality channel information and the Node B allocates a high bit rate, [2]. As shown in Figure 5.3, several factors contribute to variance of HSDSCH SINR although the HS-DSCH transmit power is supossed constant. Due to the transmission of power controlled DCHs, the total transmit power from the serving HS-DSCH cell is time variant, like the downlink radio channel in case than the user is moving, and the other cell interference experienced at the user end is also time variant. The signaling delay on the uplink is used for the case that the SINR at the user has changed from the time when the CQI was estimated until the actual transmission on the HS-DSCH. In such a case, the use of the recommended transmission format by the CQI report for the 47
HS-DSCH transmission does not necessarily guarantee that the BLEP is lower than 0.1, [4].
Figure 5.3: Block diagram showing the received signal at the HSDPA user and report of the CQI to the serving HS-DSCH cell, [4]. A simple link adaptation algorithm would directly follow the CQI values reported by the UE, but there may be a need to adjust the UE-reported CQI by adding an offset because the HS-DSCH transmit power from the Node B to the user might be different from the assumed HS-DSCH transmit power by the UE at the time it derived the CQI report; besides, the UE assumes a received HS-PDSCH power level PHS−P DSCH derived from the observed P/S-CPICH power level PCP ICH according to PHS−P DSCH = PCP ICH + G, where G is a power offset parameter signalled to the UE via RRC signalling from the RNC. The effect of feedback delays on link adaptation performance has previously been addressed in [21], [22]. In order to further adjust the CQI index received from a user before applying it for adjustment of the HS-DSCH transmission format, these studies indicate a need for an outer loop HS-DSCH link adaptation algorithm, which can be based on ACKs/NACKs from past transmissions. The algorithm adjusts the offset values to ascertain the average targetted retransmission probability. If too many retransmissions occur, an unnecessary delay is added, whereas too few 48
retransmissions indicate that the transport block sizes used are not large enough, unnecessarily lowering throughput, [2]. • HS-SCCH power control: sufficient power should be allocated to transmission of the HS-SCCH to ensure reliable reception quality of the HSSCCH, which is important since the transport block on the HS-DSCH can only be decoded if the HS-SCCH has been correctly received. Furthermore, reduce HS-SCCH transmission power is also desirable, in order to decrease the interference levels in the network. Hence, it is generally recommended to control the HS-SCCH power every TTI. The 3GPP specifications do not explicitly specify any power control mechanism for the HS-SCCH. The HS-SCCH transmit power can be adjusted relative to the transmit power of the associated downlink DPCCH, or as a function of the CQI report received from the user. This is possible if there is an internal table at the Node B expressing a power offset between each CQI index and the required HS-SCCH power. In both cases it is possible to implement a pseudo closed loop power control scheme for the HS-SCCH, relying on either feedback information from the user about the reception quality of the associated DPCH or the HS-DSCH (CQI). Common to both approaches is the Node B need for a priori knowledge of a power offset parameter before it can adjust HS-SCCH transmit power as a function of either DPCCH power or CQI.
5.2
CQI mapping proposal
The CQI from the user expresses the recommended transmission format, signaling via a CQI index n [0-30], where the CQI integer index number n can be regarded as a pointer to a vector, i.e., fCQI (n) = [An , Mn , Xn , 4n ] ∀ n [0 − 30]
(5.1)
where An is the recommended number of bits in a TTI, (TBS), Mn denotes the recommended number of HS-PDSCH codes, Xn [4QAM, 16QAM] specifies if the user recommends the Node-B to use 4QAM or 16QAM modulation, and 4n is a recommended power offset for the nth CQI index. Notice that the CQI mapping tables for function fCQI (n) specified by 3GPP in [23], are organized, so that An ≤ An+1 . The CQI index estimated by the user satisfies n = arg max {An |BLEP < 0.1} n
(5.2)
in the 2-ms interval that ends one slot before the CQI is sent, where BLEP is the per-TTI block error probability. Note that the CQI index in (5.2) depends 49
on the experienced HS-DSCH user SINR since the BLEP for different values of An depends on the SINR. The received CQI index n is adjusted by the scaling factor A, so CQI index n ˆ is used instead. The latter one is possible by assuming that the Node-B has a table with a priori knowledge of the relative HS-DSCH SINR difference in decibels between the entries in the CQI table, which is denoted by fSIN R (n). Note that the CQI tables have the following approximate property: fSIN R (n) ' 1·n dB ∀ n [0−30], i.e., the resolution of the CQI mapping table is approximately 1 dB. Given this a priori knowledge and scaling factor A, the used CQI index for HS-DSCH link adaptation is obtained as follows, [4] n ˆ = arg min {fSIN R (ˆ n) − (fSIN R (n) − An )} n ˆ
5.2.1
(5.3)
CQI Table
The CQI is based on the 5-bit CQI mapping table shown in Table 5.1. For each CQI, the table lists an associated turbo-encoder information word size expressed as Transport Block Size (TBS), modulation type and number of HS-PDSCH. These parameters describe the reference, single-transmission modulation and coding scheme associated with each CQI value. It is assumed that the reference turbo-codeword comprises all of the systematic bits, plus the parity bits identified by assuming that the number of soft metric locations in the UE is at least equal to the number of unpunctured coded bits before rate matching. The CQI table has been designed to uniformly and linearly quantize the UE SNR observation space, and offers Node-B and UE manufacturers flexibility in respectively implementing the Scheduler and the UE signal processing required to generate the CQI report, [24].
5.3
CQI optimizations in the Simulator
One of the main goals of the work is to optimizate the CQI mapping in order to maximize the overall cell throughput. In the initial version of the simulator, the CQI mapping was implemented accordingly to Equation 3.4. As mentioned, for the purpose of HS-DSCH link adaptation, the UE periodically sends a CQI to the serving HS-DSCH cell on the uplink HS-DPCCH. The CQI indicates the maximum transport block size that can be received correctly with at least 90% probability. This information is signalled via a CQI index in the range from 0 to 30, where each step corresponds appoximately to a 1-dB step in HS-DSCH SINR [9]. A simple link adaptation algorithm would directly follow the CQI values reported by the UE. However, there may be a need to adjust the UE-reported CQI by adding an offset. 50
CQI value 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Transport Block Size N/A 137 173 233 317 377 461 650 792 931 1262 1483 1742 2279 2583 3319 3565 4189 4664 5287 5887 6554 7168 7168 7168 7168 7168 7168 7168 7168 7168
Nr of HS-PDSCH 1 1 1 1 1 1 2 2 2 3 3 3 4 4 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5
Modulation 4QAM 4QAM 4QAM 4QAM 4QAM 4QAM 4QAM 4QAM 4QAM 4QAM 4QAM 4QAM 4QAM 4QAM 4QAM 4QAM 16QAM 16QAM 16QAM 16QAM 16QAM 16QAM 16QAM 16QAM 16QAM 16QAM 16QAM 16QAM 16QAM 16QAM 16QAM
Table 5.1: CQI mapping table for UE categories 1 to 6, [11].
The CQI mapping is a very important issue for the optimization of the network and HSDPA performace. A poorly implemented CQI algorithm on one model of UE can have a significant adverse impact on the efficient allocation 51
of network resources, which will affect all other UEs on the network. When the terminal is close to the base station and assumes high HS-DSCH power allocation (based on the value given by the network), a high CQI value is reported. Respectively, when the terminal is closer to the cell edge, then the reported CQI is much lower, especially if the expected HSDPA Node-B power allocation is low as well, [2]. The initial formula for the mapping used previously in the Simulator already explained, has been modified. For the purpose of achieving the new mapping and visualization, we represent the CQI values as function of the SINR, and the possible modifications in the original formula are: • Slope: by adding a scale factor is possible to change the slope of the straight line. It is important to find an optimum slope, otherwise the results could be very negative. • Shift: the shift value is used like an offset, indicating in which value of the SINR the CQI starts to have values different to cero. As in the case of the slope, the selection of this offset is also important for the optimization. The new formula results as follows:
CQI = (SIN R − shif t) · slope + 3.5,
(5.4)
where the SINR (in dB), the slope and the shift are inputs of the function and the output is the CQI value. The slope and shift values are chosen in the loading settings, and also a range of values is possible, in order to evaluate the different results. Note that the CQI values under 0 and over 30 are mapping like 0 and 30 respectively. Figure 5.4 illustrates an example in which two possibilities are shown.
5.3.1
Source Code Enhancements
Concerning to the new CQI evaluation, some changes in the initial source code have been necessary. Thus, the main function in the Simulator and the function which calculates one SISO HSDPA network snapshot have been modificated in order to handle the new functionalities. Moreover, the new necessary functions for the purpose of CQI mapping according to maximize 52
Figure 5.4: CQI mapping. the overall cell throughput by changing the slope and shift have been implemented. Figure 5.5 illustrates a coarse view of the simulation process in which the new functionalities have been included. The first step is to load the simulator settings, where it is possible to fix the range of values for the slope and shift we want to make the simulation with. These variables are found in the UE settings. Second, the precalculations are done by means of the main function. Then the simulation loop starts with a determined value of the slope and the shift, the function for the SINR to CQI mapping is called and all the calculations are done, so finally there is a throughput value for those shift and slope values. The simulation is done once again and it lasts while the range of values is not finished. At the end of the process the throughput as function of the slope and shift is achieved and represented in a figure.
5.3.2
Consequences
For the purpose of the optimization of the network and HSDPA performace, the CQI needs to be mapped in the most efficient way. In the model of the UE, the implementation of a poorly CQI algorithm can result in an impor-
53
Figure 5.5: Coarse view of the simulation including slope and shift. tant adverse impact on the efficient allocation of network resources, which will affect every user equipments on the network. Remember the object of the CQI reporting, when the terminal is close to the base station and assumes high HS-DSCH power allocation based on the value given by the network, a high CQI value is reported and respectively, when the UE is closer to the cell edge and most of the interference comes from other cells, then the reported CQI is much lower. Because of this great impact, during the work many different values for the slope and shift have been used to evaluate the best option, by using the throughput evaluation. It is important to observe how the choice of wrong values can lead to significant adverse consequences. With this object, two examples with different extreme slope and shift values are shown. Figure 5.6 illustrates the effect of choosing a very low or high value of the slope. Both of the choices have unfavourable consequences in the CQI value assignment. In the case that the slope is very high, like showing the red line, the range fo SINR values for that the CQI is not zero or thirty is very short; consequently, many users are getting a CQI value of zero, and in the other side a very high CQI value is assigned for many users with a not very good SINR. This is not optimum because the users are getting more CQI than they are able to handle, their SINR is not so high so the BLER will increase and the throughput decrease. On the other hand, the blue line shows the effects of a very low slope, this is a more conservative situation, not many users are getting the highest CQI value although they present a good SINR value, and the CQI values are more distributed for every users.
54
Figure 5.6: Different values of the slope for the CQI mapping.
Figure 5.7: Different values of the shift for the CQI mapping.
In Figure 5.7 the effect of two extreme shift values is illustrated. As before, the choice is important for the mapping optimization and the consequences 55
are the same. A too low shift value results in users with a low SINR getting CQI values different to zero, when it is impossible they receive any throughput because of the low signal quality; and respectively if the shift value is too high, the users need to have also high SINR values to get a good channel quality indicator, this leads in the situation that user equipments with good signal qualities are loosing their capacity to obtain a good service.
5.3.3
Results
Once the source code had been enhanced and adapted for the new CQI mapping, and the necessary functions were implemented, we may start with the investigations for the optimization of the final throughput. First of all it was necessary to find the optimum range of values for the slope and shift in which we can achieve high values of the throughput, and once the two ranges are found, we have to find the optimum values. Figures 5.8 and 5.9 illustrate two first evaluations of the effect that different values for the slope and the shift have in the overall cell throughput achieved by the SISO HSDPA Simulator. For the evaluations, one of the two parameters have been fixed and the other one has a big range of values to observe the effect of each single variable. The results show how important is the choice of these values, as explained in subsection 5.3.2. The throghput values in [bps] are calculated for different slopes and shifts, and we can observe a bit peak in the curve in values around 1 for the slope and 0 in case of the shift. When using very distant values there is a big drop off in the evaluated throughput.
56
Figure 5.8: Throughput as function of the slope value with shift = 0.
Figure 5.9: Throughput as function of the shift value with slope = 1. Finally, we obtain a 3-dimensional figure in which the throughput as function of the two parameters is represented. Thus, we can observe the effect of the two variables changing during the simulation, and we can find an optimum value for the slope and shift which provide the optimum overall cell throughput. For this figure, after finding the suitable delimited range of 57
slope and shift values, the investigations have been done by using low steps from the minimun to the maximum value, so that it is not probable to miss any optimum couple of values.
Figure 5.10: Throughput as function of the slope and shift.
We concludes the chapter with Figure 5.10, which shows how the throughput is reaching different values and the maximum is achieved for a slope value of 1 and a shift value of -0.4, according to the results that we observed in the two previously figures.
58
Chapter 6 Miscellaneous This chapter includes other changes in the initial simulator and investigations during this thesis that really do not fit into the previous chapters. Section 6.1 explains new scenarios for the positioning of the users. Section 6.2 includes some improvements for calculating the macro-scale pathloss by considering the three antenna radiations of every Nodes-B, and further enhancements such as the obtention of the outage users and plots of the average throughput are treated in 6.3.
6.1
New scenarios
In the new version of the simulator, it is possible to choose four different situations concerning the users positions. By using a new variable in the load settings function, the type of simulation is chosen, and accordingly the users positions are established in the function xf User positions.m. Figure 6.1 illustrates the four possibilities. Figure 6.2 illustrates the first two scenarios in greater detail. Snapshot means the users are uniformly and randomly situated according to the grid positions, whereas exhaustive snapshot refers the user positions are unformly situated but fixed at all possible grid positions. It is importat to remember that the users are situated in the main sector of the serving Node-B. In Figure 6.3 we can observe an explample with the remaining two possibilities. Fixed angle and fixed distance denote that the angle between the user and the radiation directions or the distance to the main Node-B, respectively, are fixed in the load settings. In this explamples, the angle is fixed at 90o , which means that the users are situated in the same direction 59
Figure 6.1: Four possible user positions configuration.
Figure 6.2: Snapshot and exhaustive snapshot scenarios. than the main radiation, and the distance is changing. In the other scenario, the distance between the users and the Node-B is 100 meters, but the angle changes.
60
Figure 6.3: Fixed angle and fixed distance scenarios.
6.2
More efficient implementation of the pathloss generation
In the initial version of the simulator, the macro-scale pathloss for a given user was calculated by using the three antennas in the main Node-B and the most influent antenna of the neighbors nodes (evaluated by geometrical means). In the new function, based on the initial function xf user pathloss v2, we calculate the macro-scale pathloss by considering the three antenna radiations of every Nodes-B, including the principal and the neighboring, in order to sum the overall cell interference for each user. Depending on the angle between the user and the antenna, the signal can strongly reach the user equipment or not. For implementing this, it was necessary to create a function which calculates the antenna gain for each user for every node and every sector. Figure 6.4 illustrates the network scheme used in the simulation, although an scenario with 7 cells is also possible. We can observe the user position in the target sector, the main base station in the center and the three radiation directions (90o , 210o and 330o ). By utilizing the new scenarios it is possible to represent the user pathloss as function of the distance with fixed angle, and as function of the angle with fixed distance. Figure 6.5 illustrates how the deterministic pathloss be61
Figure 6.4: Network with 19 base stations and 3 sectors model. haves when increasing the distance between the user and the Node-B. When the user is near the BS, there is a low pathloss, and it is increasing with the distance. In Figure 6.6 the pathloss as function of the angle in degrees is represented. As before, if the UE is situated in such position that the antenna is radiating directly to it (which is the case when the angle is 0o ), the macro-scale pathloss is very low, whereas when the angle is changing the pathloss increases.
6.3
Further enhancements
Some other functionalities have been implemented in order to have an enhanced comprehension and visualization of some features developed for the simulator, namely:
6.3.1
Outage users
A new implemented function allows us to obtain the percentage of outage HSDPA users for a fixed power, so that we can calculate the cumulative dis62
Figure 6.5: Variation of the user pathloss with the distance.
Figure 6.6: Variation of the user pathloss with the angle. tribution function. Besides, it’s possible to plot the number of outage users as function of the power. As explained in [24], if the UE cannot support the minimum CQI (CQI = 1) at BLER = 0.1, then the UE reports CQI=0 which indicates an ”Out of range” (OOR) condition to the Node B. As we can observe in Figure 6.7 when a low power is used, the number of
63
outage users is very high; then, as the power is increasing, there is a big dropoff, and finally for the highest values of the power, the intercell interferences are also increasing and a small increment in the outage users is produced.
Figure 6.7: HSDPA outage users as function of the power.
6.3.2
Throughput figures
The key performance metric for HSDPA is data throughput, which is highly dependent on the multipath and interference environment experienced by the UE. As already said, adaptive modulation and coding allows HSDPA users near the base station to convert their high signal-to-noise ratio into higher data throughput. Since the scheduler, attempting to maximize average cell throughput, will in general favor such users, they significantly determine the overall cell throughput [14]. In the simulator, the overall cell throughput is calculated by
T hroughput = T BS · (1 − BLER),
(6.1)
thus, the average rate of messages succesfully delivered over the channel is estimated by means of the multiplication of the transport block size and the probability of having no errors. This throughput (in terms of bits that went through correctly) is consecutively averaged over 500 realizations per 64
user. In the new version of the simulator, we can plot the throughput of every users by means of a new function. The new user positioning explained before, allows us to represent the throughput over the different scenarios. Thus, we can observe the influence of the user positions, focusing our interest in the fixed angle and fixed distance scenarios. Figure 6.8 shows the average throughput as function of the angle when every user has the same distance to the serving Node-B. Accordingly the effect of the angle between the direction radiation and the user-position can be observed. The users positioned around a 90o angle are situated in the same direction that the main radiation of the antenna, so the achieved throughput is the highest possible, while as the angle is increasing or decreasing the throughput decrements until finally it is completely zero. Furthermore, we can observe in the figure the calculated average throughput under two different situations in order to apreciate the influence of the distance: picture in the left corresponds to an scenario in which a distance of 50 meters to the BS is fixed, so due to the closeness the throughput reachs values of 2.5 Mbps. Picture in the right represents the opposite situation, where the distance is 250 meters so the throughput is much lower, reaching a maximum value around 80 Kbps. Table 6.1 specifies some of the main simulation settings used for the generation of Figures 6.8 and 6.9.
Simulation settings Number of R’99 users 5 Number of HSDPA users 40 Node-B distance 0.25 Km Codes used for HS-DSCH 10 Maximum Node-B power 20 W HSDPA range power 0-14 W Table 6.1: Simulation settings used in the generation of Figures 6.8 and 6.9
In Figure 6.9 the throughput of the users is illustrated for the fixed angle scenario. Thus, the throughput as function of the distance can be observed, 65
Figure 6.8: Throughput as a function of the angle with two fixed BS - user positions distances (50 meters in the figure of the left and 250 meters in the right). and as before, the influence of the distance between the users and the main Node-B is evaluated. We can observe how the throughput is higher around the small distances and lower for langer distances. Here we also evaluate the throughput of two different scenarios. The picture in the left side has been obtained by using a fixed angle of 80o , so it is possible to achieve high values like a throughput of 2.5 Mbps, as before. As a difference, in the right side the angle used is 40o , so the antenna does not radiate with a proper direction and it is only possible to achieve maximum throughput values around 100 Kbps.
66
Figure 6.9: Throughput as a function of the distances with two fixed angles between the users and the BS (80o in the figure of the left and 40o in the right).
67
Chapter 7 Conclusions This bachelor thesis has investigated the HSDPA throughput performance in a mixed traffic network. HSDPA represents an evolution of UMTS/WCDMA, as a Third Generation mobile telephony communications protocol. The goal of 3G systems is to provide users with high quality video and images for person-to-person communication according to the new multimedia services. With the ever increasing commercial application of 3G worldwide, subscribers are showing a huge interest in high-speed data service. As a result, an exponential increase in data services and traffic is occurring. HSDPA can greatly improve the efficiency and data transmission rate of the radio network and decrease the transmission delay and cost. Besides, HSDPA technology allows more subscribers to enjoy high-speed data service at the same time. In sum, HSDPA means better service quality for subscribers. For our purpose we focused the investigation in two enhancements. First, the extension of the SISO HSDPA Simulator in order to handle real network layout data. Second, the CQI mapping optimizations in order to maximize the overall cell throughput. This concluding chapter a summarises this bachelor thesis as follows: In Section 7.1 a brief summary of the HSDPA simulator, especially the new functionalities implemented during the thesis, is presented. Section 7.2 includes a summary with the results obtained with the new developed functionalities, and Section 7.3 concludes this bachelor thesis describing some possible future enhancements in the simulator.
68
7.1
Simulator enhancement
This thesis has consisted in the development of new functionalities in the MATLAB simulator created by the Institute of Communications and RadioFrequency of the Vienna University of Technology. These functionalities are necessary for the investigation of the main enhancements explained before. The initial purpose of the simulator was to evaluate the performance of a mixed network in which both R’99 and HSDPA traffics are simulated, in order to find the optimum parameters to obtain the maximum overall cell throughput. One of the most important extensions in the SISO HSDPA simulator was the inplementation of new functions that allow us for the handling of real network layout data that the Mobilkom Austria AG provided by means of measured path-loss matrices of an HSDPA cluster. This data have been converted in an useful form to be analyzable in the simulator, and the source-code has been enabled to use the data files after the necessary modifications. The second extension concerns the CQI values reported by the user equipment. It consisted in the optimization of the CQI mapping in order to maximize the overall cell throughput. This mapping is important since a poorly algorithm on one model of UE can have an adverse effect on the efficient allocation of network resources. The initial formula used previously in the simulator has been modified. For the purpose of achieving the new mapping and visualization, we have represented the CQI values as function of the SINR, with two new parameters: a scale factor to change the slope of the mapping, and a shift value used like an offset. Finally, some other changes in the initial simulator have been implemented, namely: • The possibility of choosing four different situations concerning the users positions: snapshot, fixed distance, fixed angle and exhaustive snapshot. The type of simulation is chosen in the settings and accordingly the users positions are established. This enhances the simulation in the sense that it is possible to analize the user pathloss and throughput for every scenarios. • More efficient implementation of the macro-scale pathloss generation, by considering the three antenna radiations of every Nodes-B, including the principal and the neighboring, in order to sum the overall cell interference for each user. 69
• HSDPA outage users calculation as function of the power. • Analysis of the throughput of every users over the different scenarios explained before, in order to observe the influence of the user positions.
7.2
Results
This section presents a summary of the results obtained with the new developed functionalities in the simulator. First, table 7.1 shows the optimizations in the new CQI mapping in order to maximize the overall cell throughput, by changing two new parameters: slope and shift. The tables present some values of the slope and shift that cause significative throughputs. As a conclusion, the optimum pair of values are pointed out in yellow.
Shift = 0 Slope 0.5 0.75 1 1.2
Slope = 1 Shift -2 -1 -0.4 2
Throughput 0.6 Mbps 1 Mbps 1.5 Mbps 0.1 Mbps
Throughput 0.1 Mbps 0.5 Mbps 0.7 Mbps 0.55 Mbps
Table 7.1: Variation of the throughput as function of the slope and shift.
Table 7.2 illustrates the percentage of outage HSDPA users as function of the power. The number of outage users decreases with the increment of the power, except for the highest values of the power, when the intercell interferences are also increasing and a small increment in the outage users is produced.
70
Power 0.2 W 1W 4 - 11 W 11 - 14 W
Outage users 45% 10% 5% 6%
Table 7.2: Variation of the outage users as function of the range of HSDPA power.
In Tables 7.3 and 7.4 some values of the macro-scale pathloss for different distances and angles between the user and the main Node-B are shown, and Table 7.5 presents some significative values of the throughput calculated within different scenarios of the users possitions. We can observe the great influence of the user position in the pathloss and throughput generation.
Angle 0o 30o 60o 100o -180o
Pathloss -95 dB -100 dB -110 dB -133 dB
Table 7.3: Variation of the pathloss as function of the angle between the user and the Node-B.
Distance 20 m 100 m 250 m 500 m
Pathloss -55 dB -78 dB -93 dB -103 dB
Table 7.4: Variation of the pathloss as function of the distance between the user and the Node-B.
71
``` ``` ```Distance Orientation ``````
Good Medium Bad
Close
Medium
Far
2.5 Mbps 2 Mbps 100 Kbps
1.5 Mbps 900 Kpbs 5 Kpbs
90 Kbps 5 Kpbs 0 Kpbs
Table 7.5: Variation of the throughput as function of the angle of orientation and distance between the user and the Node-B.
7.3
Future enhancements
During the development of this bachelor thesis only some possibilities of the new simulator have been explored. A new CQI mapping and the handling of real data network layouts were introduced but more investigations can be developed, like some HSDPA technologies that still are not incorporated. Besides, a wide range of simulations is possible and the simulator is more complete and realistic thanks to the implementation of a time-based functionality and the HSDPA scheduling developed in other thesis. Further system level investigations on the simulator have already started, [25], [26], [27], [28], [29]. Besides these enhancements, let’s remember HSDPA Evolved or HSPA+, which is the next step in HSDPA and provides HSPA data rates up to 25 Mbps on the downlink and 11 Mbps on the uplink with MIMO technologies and higher order modulation. It also introduces an optional all-IP architecture for the network where base stations are directly connected to the internet. Australia’s Next G was the first network to enable some features of HSPA+ in June 2008, [30]. In this context, it could be interesting the investigation for adapting the simulator to new HSPA+ technologies.
72
Bibliography [1] HSPA Mobile Broadband Today (http://www.gsmworld.com) [2] Harri Holma and Antti Toskala. HSDPA/HSUPA for UMTS: High Speed Radio Access for Mobile Communications. John Wiley & Sons. June 2006. [3] HSDPA Channel (http://www.altera.com).
Coding
Implementation
Issues.
[4] Klaus Ingemann Pedersen, Member, IEEE, Frank Frederiksen, Troels Emil Kolding, Member, IEEE, Tako Freerk Lootsma, and Preben E. Mogensen, Member, IEEE ,”Performance of High-Speed Downlink Packet Access in Coexistence With Dedicated Channels”, in IEEE Transactions On Vehicular Technology, Vol. 56, No. 3, May 2007. [5] Chase D. Code Combining - A Maximum - Likelihood Decoding Approach for Combining an Arbitrary Number of Noise Packets. Communications, IEEE Transactions on, Volume 33, May 1985. pp. 441-448. [6] 3GPP. Technical Specification Group Radio Access Network. High Speed Downlink Packet Access(HSDPA); Overall Description. (3GPP TS 25.308 version 5.4.0). [7] Pablo Amiegeiras Guti´errez. ”Packet Scheduling And Quality of Service in HSDPA”. Ph. D. Thesis. October 2003. [8] 3GPP. Technical Specification Group Radio Access Network. UTRAN Iub Interface NBAP signalling. (3GPP TR 25.433 version 5.5.0). [9] 3GPP Technical Specification, TS 25.214, Physical Layer Procedures (FDD), version 6.4.0. [10] Martin Wrulich, Werner Weiler and Markus Rupp. ”HSDPA Performance in a Mixed Traffic Network”. Proc. 2008 IEEE 67th Vehicular Technology Conference. Singapore. May 2008. 73
[11] 3GPP. Technical Specification Group Radio Access Network. Physical layer procedures (FDD)(Release 7). (3GPP TS 25.214 version 7.6.0). [12] Members of 3GPP, ”Technical specification group radio access network; multiple input multiple output in utra,” 3GPP, Tech. Rep. v7.0.0, Mar. 2007. [13] Members of COST 231. ”Digital mobile radio towards future generation systems - COST 231 final report”. COST, Tech. Reo., 1999. [14] A.Seeger, M.Sikora, and A.Klein, ”Variable orthogonality factor: a simple interface between link and system level simulation for high speed downlink packet access,” in Proc. of the IEEE Vehicular Technology Conference 2003, Oct. 2003, pp. 2531-2534. [15] F.Brouwer, I. de Bruin, J.Silva, N.Souto, F.Cercas, and A.Correia. ”Usage of link-level performance indicators for HSDPA network-level simulations in E-UMTS”, in Proc. of the Eight Internationsl Symposium on Spread Spectrum Techniques and Applications, Sept. 2004. [16] R.Litjens, ”HSDPA flow level performance and the impact of terminal mobility,” in Proc. of the Wireless Communications and Networking Conference, Mar. 2005, pp. 1657-1663. [17] Dieter J. Cichon and Thomas K¨ urner, ”Propagation Prediction Models”, Chapter 4, COST 1998. [18] D. McGirl, P. C. Fannin, ”Digital Elevation Modelling,” COST 231 TD(94)75, Prague, Czech Republic, 1994. [19] B. E. Gschwendtner, ”Practical investigation using ray optical prediction techniques in microcells,” COST 231 TD(94)127, 1994 [20] Qun Hou, Dexiu Huang, ”Channel Quality Indication (CQI) Application in HSDPA Simulation”, in IEEE Wireless Communications, Networking and Mobile Computing, 2007. WiCom 2007. International Conference on, pp. 1200-1203, Sept. [21] H. Vishwanathan (1999), ”Capacity of Markov channels with receiver CSI and delayed feedback”, IEEE Trans. on Information Theory, 45(2), pp. 761-771, March. [22] A. Boariu (2003), ”Effect of delay commands on adaptive modulation scheme in fading channels”, IEEE Proc. Vehicular Technology Conference (VTC), October. 74
[23] 3rd Generation Partnership Project, Physical Layer Procedures, (Mar. 2004). TS 25.214, Version 6.1.0. [Online]. Available: www.3gpp.org [24] TSG-RAN Working Group 4, ”Revised HSDPA CQI proposal,” 3GPP, Tech. Rep. R4-020612, Apr. 2002. [25] M. Wrulich, C. Mehlf¨ uhrer and M. Rupp. ”Interference Aware MMSE Equalization for MIMO TxAA”. Proc. of IEEE 3rd International Symposium on Communications, Control and Signal Processing (ISCCSP). 1585-1589. 2008. [26] C. Mehlf¨ uhrer, M. Wrulich and M. Rupp. ”Intra-cell Interference Aware Equalization for TxAA HSDPA”. Proc. of IEEE 3rd International Symposium on Wireless Pervasive Computing. 406-409. 2008 [27] C. Mehlf¨ uhrer, S. Caban, M. Wrulich and M. Rupp. ”Joint Throughput Optimized CQI and Precoding Weight Calculation for MIMO HSDPA”. Conference Record of the Fourtysecond Asilomar on Signals, Systems and Computers. Oct. 2008. [28] M. Wrulich and M. Rupp. ”Efficient Link Measurement Model for System Level Simulations of Alamouti Encoded MIMO HSDPA Transmissions”. International (ITG) Workshop on Smart Antennas (WSA). Feb. 2008. [29] M. Wrulich, S. Eder, I. Viering and M. Rupp. ”Efficient Link-to-System Level Model for MIMO HSDPA”. Proc. of the 4th IEEE Broadband Wireless Access Workshop. 2008. [30] Ericsson world-first in delivering innovative 3G technology to Telstra (http://www.ericsson.com)
75
