People s Democratic Republic of Algeria Ministry of Higher Education and Scientific Research University M Hamed BOUGARA Boumerdes

Transcription

1 People s Democratic Republic of Algeria Ministry of Higher Education and Scientific Research University M Hamed BOUGARA Boumerdes Institute of Electrical and Electronic Engineering Department of Electronics Final Year Project Report Presented in Partial Fulfilment of the Requirements for the Degree of MASTER In Telecommunication Option: Telecommunications Title: Study and Simulation of Turbo Coding Technique in LTE Systems Presented by: - TEMMAR Mohamed Nasr Eddine - BABAAMER DJELMAM Mohammed Supervisor: A. ZITOUNI Registration Number:.../2017

2 Abstract To examine the performance and the efficiency of turbo coding technique in the Physical layer (PHY) of the Long Term Evolution (LTE) standard, we implement MATLAB programs that gradually build and simulate the Transport Channel (TrCh) processing steps combined with modulation and scrambling based on predefined MATLAB functions that describe each involved processing step. The functions employ System Objects belonging to the Communication System Toolbox of MATLAB. The performance of each program is measured using graphs representing the BER as a function of Energy by bit (E b /N 0 ), by considering two different channel models: Additive White Gaussian Noise (AWGN) and Rayleigh. It is found that the performance over the AWGN channel model is better than the Rayleigh one, yet the later is considered practically to represent better the real system. Further, the performance depends mainly on a set of parameters: the modulation scheme, the number of decoding iterations, the E b /N 0 level, the coding rate and block size. Hence, a trade-off exists between the computational complexity and the performance of the system, and between energy efficient and bandwidth efficient code rates. I

3 Dedication We dedicate this work to our parents, our brothers and sisters, and our teachers. II

4 Acknowledgments: All praises and thanks to Almighty ALLAH, the most beneficent and the most merciful, who gave us the all abilities and helped us to complete this research work. We would like to express our sincere gratitude to our supervisor Dr. Abdelkader Zitouni for his support and guidance during the course of this work. His encouragement and guidance has always been a source of motivation for us to explore various aspects of the topic. Discussions with him have always been instructive and insightful and helped us to identify our ideas. Finally, we are very grateful to our parents, brothers and sisters for their sacrifices, unremitting motivation and everlasting love and their continuous support during our study. III

5 Table of contents: Abstract... I Dedication... II Acknowledgments:... III Table of contents:... IV List of tables... VI List of figures... VII List of abbreviations... IX Introduction... 2 Chapter I: Introduction to LTE Evolution of mobile communication: st Generation (1G) nd Generation (2G) rd Generation (3G) th Generation (4G) LTE technology: LTE requirements: LTE network architecture LTE protocol stack LTE enabling technologies: OFDM SC-FDM MIMO Turbo coding Link adaptation...14 Chapter II: Turbo coding in LTE Mobile communication channel: Distance-based Path loss: Fading: Interference: Channel models: Added White Gaussian Noise AWGN channel model: Rayleigh channel model: LTE DLSCH and PDSCH processing chain Transport block segmentation Channel encoding...19 Convolutional Encoder...19 Turbo encoder Rate matching for turbo coded transport channels Block concatenation Scrambling Modulation Demodulation Descrambling Rate Dematching Channel decoding...26 IV

6 Turbo decoding Viterbi decoding:...28 Chapter III: Simulation Results and Discussion Simulation approach: Simulation functions: Simulation results Signal Modulation Scrambling Turbo coding Early-Termination Mechanism Rate matching LTE DLSCH and PDSCH Processing Conclusion: References Appendices V

7 List of tables Table 1 QPSK modulation mapping 25 Table 2 transceiver computation time for QPSK with different CPU for each channel model 39 Table 3 Typical execution time with and without early-termination mechanism for different modulation modes. 45 Table 4 SNR level required for BER values: 10-2, 10-3, and 10-4 for QPSK, QAM16, and QAM64 with different coding rates with three decoding iterations for AWGN channel. 55 VI

8 List of figures Figure 1 LTE network architecture... 7 Figure 2 The E-UTRAN architecture... 8 Figure 3 The EPC architecture... 9 Figure 4 Layer architecture in a LTE RAN Figure 5 Rate 1/3 tail biting convolutional encoder Figure 6 Structure of rate 1/3 turbo encoder (dotted lines apply for trellis termination only) Figure 7 Rate matching for turbo coded transport channels Figure 8 Block diagram of a turbo decoder Figure 9 the block diagram of the Viterbi decoder Figure 10 presentation of all MATLAB test benshes Figure 11 Signal modulation test bench Figure 12 BER performance with QPSK, QAM16, and QAM64 for (a) AWGN channel and (b) Rayleigh channel Figure 13 Scrambling test bench Figure 14 BER performance with the presence of scrambling for QPSK, QAM16, and QAM64 for (a) AWGN channel and (b) Rayleigh channel Figure 15 Turbo coding test bench Figure 16 Convolutional encoder/viterbi decoder test bench Figure 17 BER performance: Turbo coding vs convolutional\viterbi coding for AWGN channel for (a) QPSK and (b) QAM Figure 18 BER performance: Turbo coding vs convolutional\viterbi coding for Rayleigh channel for (a) QPSK and (b) QAM Figure 19 Part from profiling results for turbo coding with 5 decoding iterations for AWGN channel Figure 20 Part from profiling results for turbo coding with 1 decoding iteration for Rayleigh channel Figure 21 Turbo coding with early termination mechanism test bench Figure 22 Comparison of BER results with and without CRC-based early termination for QPSK for AWGN channel with (a) 3 decoding iterations (b) 6 decoding iterations (c) 10 decoding iterations and (d) 15 decoding iterations VII

9 Figure 23 Comparison of BER results with and without CRC-based early termination for QAM16 for AWGN channel with (a) 3 decoding iterations (b) 6 decoding iterations (c) 10 decoding iterations and (d) 15 decoding iterations Figure 24 Comparison of BER results with and without CRC-based early termination for QPSK for Rayleigh channel with (a) 1 decoding iteration (b) 6 decoding iterations (c) 10 decoding iterations and (d) 15 decoding iterations Figure 25 Rate matching test bench Figure 26 Effect of rate matching on turbo coding BER performance for AWGN channel with (a) QPSK (b) QAM16 and (c) QAM Figure 27 Effect of rate matching on turbo coding BER performance for Rayleigh channel with (a) QPSK (b) QAM16 and (c) QAM Figure 28 LTE DLSCH and PDSCH test bench Figure 29 BER performance: TrCh processing with increasing the number of decoding iterations for AWGN channel with (a) QPSK (b) QAM16 and (c) QAM Figure 30 BER performance: TrCh processing with increasing the number of decoding iterations for Rayleigh channel with (a) QPSK (b) QAM16 and (c) QAM Figure 31 BER performance: TrCh processing for QPSK modulation for AWGN channel with different coding rates (a) 1/2 (b) 2/ Figure 32 BER performance: TrCh processing for QPSK modulation for Rayleigh channel with different coding rates (a) 1/2 (b) 2/ Figure 33 BER performance: TrCh processing for QAM16 modulation with 1/2 coding rate for (a) AWGN channel and (b) Rayleigh channel Figure 34 BER performance: TrCh processing for QAM64 modulation for AWGN channel with different coding rates (a) 1/2 (b) 2/ Figure 35 BER performance: TrCh processing forqam64 modulation for Rayleigh channel with different coding rates (a) 1/2 (b) 2/ Figure 36 BER performance: TrCh processing for QPSK modulation with different code block size for (a) AWGN channel and (b) Rayleigh channel Figure 37 Trellis diagram for error free decoding with hamming distance for branch metric calculations...64 VIII

10 List of abbreviations 1G: First Generation 2G: Second Generation 3G: Third Generation 4G: Forth Generation 3GPP: 3rd Generation Partnership Project A APP: A Posteriori Probability AWGN: Additive White Gaussian Noise ACS unit: Add Compare and Select unit B BER: Bit Error Rate BCJR: Bahl, Cocke, Jelinek, Raviv BMU: Branch Metric Unit C D E F G H I CDMA: code division multiple access CDMA2000: code division multiple access 2000 CRC: Cyclic Redundancy Check DFT: Discrete Fourier Transform DL: Downlink DL-SCH: Downlink Shared Channel EPC: Evolved Packet Core E-UTRAN: Evolved Universal Terrestrial Radio Access Network EPS: Evolved Packet System enb: evolved NodeB EDGE: Enhanced Data rates for GSM Evolution FDD: Frequency Division Duplex FM: frequency modulation FDMA: frequency division multiple access FEC: Forward Error Correction GSM: global system for mobile communication GPRS: general packet radio service HARQ: hybrid automatic repeat request HSPA: High Speed Packet Access inter-rat: inter Radio Access Technology IX

11 IP: Internet Protocol ITU: International Telecommunication Union ITU-R: International Telecommunication Union Radio communications Sector IMT-Advanced: International Mobile Telecommunications-Advanced L M N O P LLR: log-likelihood ratios LTE: long term evolution MAC: Media Access Control MME: Mobility Management Entity MIMO: multiple-input multiple-output MMS: Multimedia Messaging Service MAP: Maximum A posteriori Probability NAS protocol: Non Access Stratum protocol OFDM: orthogonal frequency division multiplexing PDN: Packet Data Network P-GW: Packet data network Gateway PDSCH: Physical downlink shared channel PUSCH: Physical uplink shared channel PHY: physical layer PCCC: Parallel Concatenated Convolutional Code Q R S T QoS: Quality of Service QPSK: Quadrature Phase Shift Keying QAM: Quadrature Amplitude Modulation QPSK: Quadrature Phase Shift Keying RRM: Radio Resource Management RRC: Radio Resource Control SIM: Subscriber Identity Module SNR: Signal-to-noise ratio SISO: soft-in soft-out S-GW: Serving gateway SMS: Short Message Service SC-FDM: Single-carrier frequency division multiplexing TDMA: Time Division Multiple Access TrCh: Transport Channel X

12 TBU: Trace Back Unit U V W UMTS: Universal Mobile Telecommunications Service USIM: Universal Subscriber Identity Module UE: User Equipment UL: uplink UL-SCH: Uplink Shared channel VoIP: Voice over IP VA: Viterbi Algorithm WiMAX: Worldwide Interoperability for Microwave Access XI

13 INTRODUCTIONTO LTE Introduction Mobile communication systems have evolved rapidly in few decades. This evolution is driven by the fast growth in the number of mobile subscribers and the invention of new mobile devices: smartphones, and tablets. With these devices, customers ask for faster data rates enabling them to run their applications and services such as video calls and high speed internet access. Such applications involve the transmission of a huge amount of data over the wireless channel that causes changes to the transmitted signals in an unwanted manner. To overcome this effect an appropriate error control coding should be employed enabling the detection or even the correction of channel errors improving the reliability of transmission. The LTE standard employs turbo coding mechanism which is a powerful forward error correction coding as it achieves performance near Shannon limit. Turbo coding is considered as one of the enabling technologies of the LTE standard that contributes to its remarkable performance. It is the only channel coding mechanism used to process the user data. The LTE standard introduces many improvements to the turbo coding mechanism to reach an optimal efficiency. In this report we study and simulate turbo coding mechanism in the physical layer of the LTE standard verifying its performance in terms of BER versus E b /N 0 levels. The report contains the following: Chapter1: brief description of the evolution of mobile communication technologies, and a presentation of the LTE standard technology treating its requirements, its network architecture, its protocol stack and its enabling technologies. Chapter2: introduction to the wireless channel characteristics followed by presentation of the AWGN and the Rayleigh channel models, then description of the TrCh processing steps including: code-block segmentation and code-block CRC attachment, turbo coding, rate matching, code-block concatenation, in addition to modulation and scrambling. 2

14 INTRODUCTIONTO LTE Chapter3: simulation results shown as graphs of BER versus E b /N 0 levels and discussions of the TrCh processing steps with modulation and scrambling over AWGN and Rayleigh channel models. Conclusion: general conclusion and further work. 3

15 INTRODUCTIONTO LTE Chapter I: Introduction to LTE Chapter I: Introduction to LTE Long Term Evolution (LTE) is a high-speed broadband mobile communication standard developed by the 3rd Generation Partnership Project (3GPP) organization to meet the increasing demands of subscribers for higher data rates with low latency. This chapter presents the LTE technology, but before describing LTE, we start first by presenting the evolution of mobile communication from the 1 st generation to the 4 th one. 4

16 INTRODUCTIONTO LTE 1.1. Evolution of mobile communication: In order to see the improvements brought by the LTE standard we are going to present briefly the evolution of mobile communication which is structured in successive generations as described below: st Generation (1G) The first generation established the foundations of cellular mobile communication. 1G supported only voice traffic based on the analog frequency modulation (FM), and frequency division duplex (FDD) as duplex method. 1G operated on the frequency band around 900 MHz The available bandwidth was split using frequency division multiple access (FDMA) technique into channels of 20 KHz that can support only one user per channel. This explains the low capacity of 1G technology to support mobile users nd Generation (2G) 2G system is operating on the 900 MHz and the 1800 MHz bands. It has more capacity than 1G since the available spectrum is split using FDMA into channels of 200 KHz divided using time division multiple access (TDMA) to 8 time slots that can handle 8 different users. 2G transmission is based on digital modulation, and provides some data services: Short Message Service (SMS) and Multimedia Messaging Service (MMS), in addition to the voice traffic. Data rate is about 9.6 kbps in the Global System for Mobile communication standard (GSM) which reaches kbps for the General Packet Radio Service standard (GPRS) and 473 kbps for the Enhanced Data rates for GSM Evolution standard (EDGE) rd Generation (3G) 3G technology provides more broadband data services to mobile users such as: Web browsing, , TV streaming, video conferencing with higher data rates up to 2 Mbps. Most of 3G systems such as the Universal Mobile Telecommunications Service (UMTS) are based on the wide-band code division multiple access (CDMA) technology. In CDMA technology, the 1.25 MHz channels are shared among different users; however, each user is assigned a unique code. High Speed Packet Access (HSPA) is another 3G technology with data rates up to 14 Mbps and it is regarded as an upgrade of UMTS. HSPA+ standard offers the higher 3G data rate with 84 Mbps. 5

17 INTRODUCTIONTO LTE th Generation (4G) 4G technologies have been developed by the International Telecommunication Union Radio communications Sector (ITU-R) as the International Mobile Telecommunications Advanced (IMT-Advanced) specification to support the increasing demands of mobile users for high data rates and low latency or packet delay required for their mobile applications. 4G systems are characterized by an all-ip network infrastructure relying only on packet switching technique, distributed radio network architecture, use of advanced antenna techniques, large radio channel bandwidths, high-order modulation schemes, stringent Quality of Service (QoS) control. The Worldwide Interoperability for Microwave Access (WiMAX) and LTE are the two 4G system contenders LTE technology: LTE was started as a project in 2004 by the telecommunication body 3GPP with the purpose of identifying the LTE specifications required to meet IMT-Advanced requirements. LTE is considered as an evolution from UMTS. The following sections present LTE requirements, LTE enabling technologies, LTE network architecture, as well as the LTE protocol stack LTE requirements: LTE requirements have been specified by the 3GPP to achieve the following goals: Spectrum flexibility: LTE allocates channel bandwidths of different sizes: 1.4, 3, 5, 10, 15 and 20 MHz for uplink (UL) and downlink (DL). Peak Data Rate: 100 Mbps for DL and 50 Mbps for UL for 20 MHz channel bandwidth, which increased the spectral efficiency to 5 bps/hz for DL and to 2.5 bps/hz for UL. Latency: For both user and control planes User plane latency: The one-way transit time for a packet to travel from the IP layer in the user equipment (UE) or in the Evolved Universal Terrestrial Radio Access 6

18 INTRODUCTIONTO LTE Network (E-UTRAN) to the IP layer in the E-UTRAN or in the UE shall be less than 30 ms. Control plane latency: the time it takes for the UE to switch from a passive connection with the network (IDLE state) to an active connection (CONNECTED state) shall be less than 100 ms. Mobility: The system should be optimized for low mobile speed (0-15 km/h), should support higher mobile speeds be as well as high speed train environment as special case. Coverage: Full performance for 5 km cells; Slight degradation 5 km to 30 km cells; Cells up to 100 km should not be precluded. Quality of Service: End-to-end QoS and Voice over IP (VoIP) shall be supported LTE network architecture The high-level network architecture of LTE is shown in figure 1 Figure 1 LTE network architecture[1] It is composed of following main components: The User Equipment (UE) A LTE compatible mobile device associated with a Subscriber Identity Module (SIM) card that runs the Universal SIM (USIM) application which stores user-specific data about the user's phone number, home network identity and security keys. [1] 7

19 INTRODUCTIONTO LTE The Evolved UMTS Terrestrial Radio Access Network (E-UTRAN) The architecture of E-UTRAN is illustrated in figure 2 Figure 2 The E-UTRAN architecture[1] It is composed of evolved base stations, called enodeb or enb. Each enb controls mobiles in one or more cells. LTE Mobiles can communicate with just one enb known as its serving enb via Uu interface which is simply the air interface. The enbs are interconnected with each other by means of X2 interface to exchange information related to the coordination of transmissions in adjacent cells (for intercell interference reduction). The E-UTRAN handles the radio communications between UE and Evolved Packet Core (EPC). Its main functions are: Radio resource management (RRM): radio mobility control, scheduling and dynamic allocation of resources to UEs in both uplink and downlink. Security: All data sent over the radio interface is encrypted. Connectivity to the EPC: This consists of the signaling toward the Mobility Management Entity (MME) and the bearer path toward the Serving Gateway (S- GW). [2] 8

20 INTRODUCTIONTO LTE The Evolved Packet Core (EPC) Figure 3 the EPC architecture[18] It is the core network of the LTE standard which is responsible for communication with Packet Data Networks (PDN) in the outside world such as the internet and for the establishment of the bearers. A bearer is an IP packet flow with a defined quality of service (QoS) between the gateway and the UE. The main nodes of the EPC are: Packet Data Network Gateway (P-GW) It communicates with the outside world i.e. PDN using SGi interface. Packet data network Gateway (P-GW) is responsible for all the IP packet based operations: IP address allocation for UEs, filtering of downlink user IP packets into the different QoSbased bearers, UL and DL rate enforcement. It also serves as the mobility tunnel for interworking with non-3gpp technologies such as CDMA2000 and WiMAX networks. Serving gateway (S-GW) It forwards data between enbs and P-GW. S-GW acts as a local mobility entity for inter enb handovers. It also serves as the mobility tunnel for interworking with other 3GPP technologies such as GPRS and UMTS. S-GW is responsible for packet routing and forwarding of downlink packets. Mobility management entity (MME) It is the control node that processes control plane operations: signaling between the UE and the EPC by means of Non Access Stratum (NAS) protocols. The main functions supported by the MME are: Functions related to bearer management: The establishment, maintenance and release of the bearers and is handled by the session management layer in the NAS protocol. 9

21 INTRODUCTIONTO LTE Functions related to connection management: The establishment of the connection and security between the network and UE and is handled by the mobility management layer in the NAS protocol layer. [2] LTE protocol stack Communication between LTE entities is based on different protocols organized in protocol stack composed of different layers. In this section we present the protocol layers for the E-UTRAN radio access network. A general protocol architecture of E-UTRAN splits the radio interface into three layers, each layer is characterized by the services provided to the higher layers and the functions it supports. The three layers are as follows: Physical layer (PHY): The PHY carries information from the transport channels of the Medium Access Control layer (MAC) over the air interface. It is responsible for modulation, multipleinput multiple-output (MIMO) implementations and orthogonal frequency division multiplexing processing (OFDM) for the DL and single-carrier frequency division multiplexing processing (SC-FDM) for the UL transmissions. Medium Access Control layer (MAC): The MAC offers a set of logical channels to the Radio Resource Control layer (RRC), it multiplexes data for different services ( and web-browsing) into a single packet for the physical layer's further processing, it also manages the hybrid automatic repeat request (HARQ) error correction, handles the prioritization of the logical channels for the same UE and the dynamic scheduling between UEs to meet QoS requirements. Radio Resource Control layer (RRC): The RRC takes care of handling RRC procedures over the air interface, including the following: Broadcasting system information to all UEs attached to the cell, establishment and release of the RRC connection between the UE and E-UTRAN, Security functions including key management, Mobility control (inter Radio Access Technology handover (inter-rat)), QoS control. It is also responsible for lower layer configuration, including: MIMO Transmission mode, Uplink power control, Bandwidth configuration. 10

22 INTRODUCTIONTO LTE Figure 4 Layer architecture in a LTE RAN[3] Information flows between the three layers through three different channels: Logical channels Represent the data transfers between the RRC layer and the MAC layer. They specify the type of information transferred. Two types of logical channels exist in LTE: Traffic logical channel: transfers user-plane data Control logical channels transfer the control-plane information. Transport Channels Connect the MAC layer to the PHY layer. They are defined by how and with what characteristic the information within each logical channel is transmitted over the radio interface, so they are responsible for: the structure for passing data to/from higher layers and the mechanism by which higher layers can configure the PHY. Physical channels Carry information from higher layers of the protocol stack for uplink and downlink transmission of data on the air interface. Compared to previous standards, LTE has reduced the use of dedicated channels and relies more on shared channels where many different types of logical and transport channels converge on the shared physical channels. In this report we focus on the following channels: For the DL transmission: Transport channel: Downlink Shared Channel (DL-SCH) Physical channel: Physical Downlink Shared Channel (PDSCH) For the UL transmission: Transport channel: Uplink Shared channel (UL-SCH) Physical channel: Physical Uplink Shared Channel (PUSCH) 11

23 LTE enabling technologies: INTRODUCTIONTO LTE The key enabling technologies in the physical layer that lead to the good performance of LTE technology are: OFDM, SC-FDM, MIMO, Turbo coding, and link adaptation mechanisms OFDM OFDM is a multicarrier transmission technique used for the DL transmission. It divides the available spectrum into many orthogonal subcarriers, where user data streams are divided into sub-streams sent among these subcarriers in parallel, and the combination of the parallel subcarriers at the destination provide high data rates transmission. OFDM has the following advantages: Robustness to the multipath fading channel: When the frequency spacing between subcarriers is sufficiently small, an OFDM transmission scheme can represent a frequency-selective fading channel as a collection of narrowband flat fading subchannels. High spectral efficiency due to the efficient use of the available spectrum. OFDM helps to estimate the channel frequency response based on transmitting reference data. With a good estimate of the channel response at the receiver, we can then recover the best estimate of the transmitted signal using a low-complexity frequency domain equalizer that inverts the channel frequency response at each subcarrier. [3] SC-FDM One of the drawbacks of OFDM multicarrier transmission is the large variations in the instantaneous transmit power [3]. This is because the subcarriers carry different data substreams in parallel. In UL transmission the power consumption is very crucial and the design of complex power amplifiers is especially challenging. As a result, a variant of the OFDM transmission known as SC-FDM is selected in the LTE standard for uplink transmission. SC-FDM is implemented by combining a regular OFDM system with a precoding based on Discrete Fourier Transform (DFT) [1].By applying a DFT-based precoding, SC-FDM substantially reduces fluctuations of the transmit power. The resulting uplink transmission scheme can still feature most of the benefits associated with OFDM, such as low-complexity frequency-domain equalization and 12

24 INTRODUCTIONTO LTE frequency-domain scheduling, with less stringent requirements on the power amplifier design. [3] MIMO MIMO techniques bring to bear the advantages of using multiple antennas in order to meet the ambitious requirements of the LTE standard in terms of peak data rates and throughput. MIMO methods can improve mobile communication in two different ways: by boosting the overall data rates and by increasing the reliability of the communication link. The MIMO algorithms used in the LTE standard can be divided into four broad categories: receive diversity, transmit diversity, beamforming, and spatial multiplexing. In transmit diversity and beamforming, we transmit redundant information on different antennas. As such, these methods do not contribute to any boost in the achievable data rates but rather make the communications link more robust. In spatial multiplexing, however, the system transmits independent (non-redundant) information on different antennas. This type of MIMO scheme can substantially boost the data rate of a given link. The extent to which data rates can be improved may be linearly proportional to the number of transmit antennas. In order to accommodate this, the LTE standard provides multiple transmit configurations of up to four transmit antennas in its downlink specification. [3] Turbo coding Turbo coding is the forward error correcting coding used in LTE standard. It is based on the convolutional coding technology. The LTE turbo coders come with many improvements, aimed at making them more efficient in their implementation. For example, by appending a Cyclic Redundancy Check (CRC) checking syndrome to the input of the turbo encoder, LTE turbo decoders can take advantage of an early termination mechanism if the quality of the code is deemed acceptable. Instead of following through with a fixed number of decoding iterations, the decoding can be stopped early when the CRC check indicates that no errors are detected. This very simple solution allows the computational complexity of the LTE turbo decoders to be reduced without severely penalizing their performance. [3] 13

25 INTRODUCTIONTO LTE Link adaptation Link adaptation is defined as a collection of techniques for matching and adapting the transmission parameters of a mobile communication system with the dynamic nature of the communication channel. In LTE there exist different link adaptation techniques: Adaptive modulation and coding Adaptive MIMO Adaptive bandwidth Closely related to link adaptation is channel-dependent scheduling in a mobile communication system. Scheduling deals with the question of how to share the radio resources between different users in order to achieve more efficient resource utilizations. Typically, we need to either minimize the amount of resources allocated to each user or match the resources to the type and priority of the user data. Channeldependent scheduling aims to accommodate as many users as possible, while satisfying the best quality-of-service requirements that may exist based on the instantaneous channel condition. [3] 14

26 Chapter II: Turbo coding in LTE Chapter II: Turbo coding in LTE The LTE (Long Term Evolution) downlink PHY (Physical Layer) chain can be viewed as the combination of processing applied to the Downlink Shared Channel (DLSCH) and Physical Downlink Shared Channel (PDSCH). DLSCH processing at the level of the transmitter includes these steps: Transport-block segmentation and code-block CRC attachment. Turbo coding based on a one-third rate(1/3). Rate matching to handle any requested coding rates. Code-block concatenation to generate code words Whereas PDSCH processing takes the outputs of the DLSCH processing to perform: Scrambling Modulation of scrambled bits to generate complex-valued modulation symbols. [3] 15

27 Note that the LTE, like other mobile standards, does not make any recommendations concerning the operations performed in the receiver. [3] Hence, all the receiver specifications presented here can be considered typical inverse operations of operations specified in the transmitter as follows: 1. Inverse physical channel processing (inverse PDSCH processing): Demodulation to generate the estimated bits Descrambling 2. Inverse transport block processing (inverse DLSCH processing): Iteration over each code block Rate dematching (from target rate to 1/3 rate) Code block 1/3-rate turbo decoding with early termination based on CRC These proposed inverse operations represent best efforts to recover the estimate of the transmitted bits. Although not specified in the standard, it is necessary to include these receiver-side inverse operations in order to evaluate the accuracy and performance of the system in the next chapter. This chapter introduces first the characteristic of mobile communication channel and AWGN with Rayleigh channel models, then it presents DLSCH processing in addition to scrambling and modulation that are involved in PDSCH processing Mobile communication channel: The wireless channel over which mobile communication occur is a major factor in the evolution of mobile communications as it poses a severe challenge which restrict the range, data rate, and the reliability of the communication. As a signal travels from the transmitter to the receiver, its strength suffers from variations over time and over frequency. This is mainly due to three factors: Distance-based Path loss: Expressed as the ratio of the transmitted-signal power to the received-signal power, on a given path. It is a function of the propagation distance, frequency used, and the nature of the terrain. Different models describe path loss (free-space model; two-ray model; Okumura-Hata model) and they generally assume that path loss is the same at a given transmit-receive distance (no shadowing effects are considered). 16

28 Fading: Defined as the variation in the signal strength at the receiver. It can be classified into two types: Slow fading (large scale fading): It is typically frequency independent, and it occurs when large obstacles (buildings and hills) that partially absorb the transmission lie between the transmitter and the receiver causing random variations in the received signal power although the distance between the transmitter and receiver remains the same. Variations may last for multiple seconds or minutes hence the word slow. Fast fading (small scale fading): It is frequency dependent, and it refers to the rapid changes in the signal strength as the receiver moves a small distance (on the order of the wavelength or even a fraction of the wavelength). These changes are due to the constructive and destructive interference between multiple versions of the same transmitted signal arriving at the receiver at slightly different times. The multipath propagation of the transmitted signal, which causes fast fading, is because of the three propagation mechanisms: Reflection: Occurs when the signal encounters a large solid surface, whose size is much larger than the wavelength of the signal. Diffraction: Occurs when the signal encounters an edge or a corner, whose size is larger than the wavelength of the signal. Scattering: Occurs when the signal encounters small objects of size smaller than the wavelength of the signal Interference: Wireless transmissions have to counter interference from a wide variety of sources. The main forms of interference are: Adjacent channel interference: Signals in nearby frequencies have components outside their allocated ranges, and these components may interfere with on-going transmission in the adjacent frequencies. It can be avoided by carefully introducing guard bands between the allocated frequency ranges. [4] Co-channel interference: It is due to other nearby systems using the same transmission frequency. [4] 17

29 2.2. Channel models: SIMULATION RESULTS AND DISCUSSION In order to measure the performance of wireless communication, it is necessary to describe the wireless channel with statistical models. In our report we consider two channel models: Added White Gaussian Noise AWGN channel model: As its name indicates, the AWGN Channel is a transmission channel, where a Gaussian white noise is added. This kind of noise has an identical power spectral density for all frequencies and follows a normal distribution of zero mean and defined variance. The noise n (t) affects the transmitted signal x (t) with the following equation: (2.1) Rayleigh channel model: In mobile radio channels, the Rayleigh distribution is commonly used to describe the statistical time varying nature of the received envelope of a flat fading signal. The Rayleigh channel model can be expressed as : (2.2) Where the amplitude factor g (t) is the summed effect of many reflectors, it may be regarded (by the central limit theorem) as a complex Gaussian random variable. The transmitted signal at certain frequency reaches the receiver via a number of paths. Signal weakening can cause the main component not to be noticed among the multipath components, originating Rayleigh model. Rayleigh fading is a reasonable model when there are many objects in the environment that scatter the radio signal before it arrives at the receiver. [5] 2.3. LTE DLSCH and PDSCH processing chain Transport block segmentation The input stream to the code block segmentation is denoted by b 0, b1, b2, b3,..., b B 1. If B is larger than Z=6144, which is the maximum supported code block size, the input stream is segmented and an additional CRC sequence bits is attached to each code block with length 18

30 Total number of code blocks is (2.3) If B is less than Z and is not equal to one of the supported code block sizes shown in the appendix 1 it is padded with zeroes to reach the closest supported size. The bits output from code block segmentation, for C 0, are denoted byc r0, cr1, cr2, cr3,..., cr 1 bits for the code block number r. [6] K r Channel encoding, where r is the code block number, and K r is the number of Channel coding comprises error detection and error correction. With error detection using the CRC (Cyclic Redundancy Check) detector, the receiver can request the repeat of a transmission, in what is known as an automatic repeat request. Forwarderror-correction coding allows errors to be corrected based on the redundancy bits that are included with the transmitted signal. A hybrid of error detection and forward error correction known as HARQ (Hybrid Automatic Repeat Request) forms an integral part of most 3G standards and is also used in the LTE standards. [3] The bit sequence input for a given code block to channel encoding is denoted byc 0, c1, c2, c3,..., c K 1, where K is the number of bits to encode. After encoding the ( i) ( i) ( i) ( i) bits are denoted by d 0, d1, d 2, d3,..., d ( i) D 1, where D is the number of encoded bits per output stream and i indexes the encoder output stream. The relation between c k and (i) d k and between K and D is dependent on the channel coding scheme. Convolutional Encoder Convolutional codes have memory, which means the coded output bit depends not only on the current bit but also on the m previous bits, where m is the number of registers in the convolutional encoder. A convolutional encoder C (k, n, m) consists of a shift register with m stages, whose outputs are added (XOR) together to build the output bit. Per k bit input information bits, n output bits are produced from the encoder. [7] The configuration of the convolutional encoder in LTE with constraint length 7 and coding rate 1/3 is presented in figure 5. [6] 19

31 c k D D D D D D (0) d k G 0 = 133 (octal) (1) d k G 1 = 171 (octal) (2) d k G 2 = 165 (octal) Figure 5 Rate 1/3 tail biting convolutional encoder In order to decode a convolutionally encoded sequence, the initial and final states should be known to the receiver. One method to do this is Tail-biting where the initial value of the shift register of the encoder shall be set to the values corresponding to the last 6 information bits in the input stream so that the initial and final states of the shift register are the same. Therefore, denoting the shift register of the encoder by s 0, s1, s2,..., s5, then the initial value of the shift register shall be set to s i c (2.4) 1 K i The encoder output streams d, (0) k (1) d k and and third parity streams, respectively as shown in Figure 5. (2) d k correspond to the first, second Convolutional codes are often preferred in practice over block codes, because they provide excellent performance when compared with block codes of comparable encode/decode complexity. [8] Turbo encoder Shannon s channel coding theorem implies strong coding behaviour for random codes as the code block length increases, but increasing block length typically implies an exponentially increasing decoding complexity. In 1993, an approach to error correction coding was introduced which provided for very long code words with only (relatively) modest decoding complexity. These codes were termed turbo codes by their inventors. They have also been termed parallel concatenated codes. [8] The scheme of turbo encoder is a Parallel Concatenated Convolutional Code (PCCC) with two 8-state constituent encoders and one turbo code internal interleaver. The coding rate of turbo encoder is 1/3. The structure of turbo encoder is illustrated in figure 6. [6] 20

32 x k 1st constituent encoder z k c k D D D Output Input Turbo code internal interleaver Output 2nd constituent encoder z k c k D D D x k Figure 6 Structure of rate 1/3 turbo encoder (dotted lines apply for trellis termination only) The transfer function of the 8-state constituent code for the PCCC is: (2.5) The initial value of the shift registers of the 8-state constituent encoders shall be all zeroes when starting to encode the input bits. The output from the turbo encoder is (2.6) The bits input to the turbo encoder are denoted by c 0, c1, c2, c3,..., c K 1, and the bits output from the first and second 8-state constituent encoders are denoted by z0, z1, z 2, z3,..., z K 1 and z 0, z1, z 2, z3,..., z K 1, respectively. The bits output from the turbo code internal interleaver are denoted by c 0, c 1,..., c K 1, and these bits are to be the input to the second 8-state constituent encoder. [6] 21

33 Trellis termination for turbo encoder In order for the decoder to know the initial and final states of the shift register, both constituent encoders start in the all-zeroes state and they are forced back the allzeroes state after they have encoded the entire input. Turbo encoder cannot be brought into the all-zeroes state from some other state merely with a sequence of zeroes. However, if the feedback bit were to be used as the encoder input, then the XOR of these two bits will be zero, and thus the encoder will return to the all-zeroes state after three clock cycles (i.e., the output of the leftmost XOR gate in Figure 6 will be zero, since the two inputs are the same). Therefore, the encoder can be brought back to the all-zeroes state by inputting the three feedback bits generated immediately after the k-bit code word has been encoded. This is achieved by moving the switches from the up position to the down position after the k-bit input has been encoded. Note that because of the interleaver, the state of the two encoders are likely to be different, and thus the tails required for each encoder will also be different. [9] The first three tail bits shall be used to terminate the first constituent encoder (upper switch of figure 6 in lower position) while the second constituent encoder is disabled. The last three tail bits shall be used to terminate the second constituent encoder (lower switch of figure 6 in lower position) while the first constituent encoder is disabled. The transmitted bits for trellis termination shall then be: d (0) d, d K x K (0) (0) K x K, K 1 z K 1 (0) 2, d K 3 z K 1 d (1) d x, d K z K (1) (1) K z K, K 1 K 2 (1) 2, d K 3 x K 2 (2.7) (2) (2) d K x K 1, K 1 z K 2 Turbo code internal interleaver (2) (2) d, d K 2 x K 1, d K 3 z K 2 The interleaver permutes the input sequence so that input patterns giving low weight words from the first encoder are interleaved to patterns giving words with high weight for the second encoder yielding to better coding performance. The bits input to the turbo code internal interleaver are denoted by c 0, c1,..., c K 1, where K is the number of input bits. The bits output from the turbo code internal interleaver are denoted by c 0, c 1,..., c K 1. 22

34 The relationship between the input and output bits is as follows: Where the relationship between the output index i and the input index (i) satisfies the following quadratic form: [6] appendix 1. (2.8) (2.9) The parameters f 1 and f 2 depend on the block size K and are summarized in Rate matching for turbo coded transport channels Rate matching is instrumental in the implementation of adaptive coding, an important feature of modern communications standards. It helps augment the throughput based in the channel conditions. [3] The rate matching for turbo coded transport channels is defined per coded block and consists of interleaving the three information bit streams d, d and d, followed by the collection of bits and the generation of a circular buffer as depicted in Figure 7. [6] The bits output from the Rate matching module are denoted by e 0,e 1,...,e E-1 (0) k (1) k (2) k (0) d k Sub-block interleaver (0) v k (1) d k Sub-block interleaver (1) v k Bit collection virtual circular buffer w k Bit selection and pruning e k (2) d k Sub-block interleaver (2) v k Figure 7 Rate matching for turbo coded transport channels Block concatenation The input bit sequence for the code block concatenation block are the sequences e rk, for r 0,..., C 1and k 0,..., E 1. The output bit sequence from the code block concatenation block is the sequence f k for k 0,..., G 1. The code block concatenation consists of sequentially concatenating the rate matching outputs for the different code blocks. [6] r 23

35 Scrambling In LTE downlink processing, the code word bits generated as the outputs of the channel coding operation are scrambled by a bit-level scrambling sequence. Different scrambling sequences are used in neighbouring cells to ensure that the interference is randomized and that transmissions from different cells are separated prior to decoding. [3] Scrambling is composed of two parts: pseudo-random sequence generation and bit-level multiplication. The pseudo-random sequences are generated by a Gold sequence with the length set to 31. The output sequence is defined as the output of an exclusive-or operation applied to a specified pair of sequences. The polynomials specifying this pair of sequences are as follows: (2.10) The initialization value of the first sequence is specified with a unit impulse function of length 31. The initialization value for the second random sequence depends on such parameters as the cell identity, number of codewords, and subframe index. Finally, the bit-level multiplication is implemented as an exclusive-or operation between the input bits and the Gold sequence bits. [3] ( q) ( q) ( q) For each code word q, the block of bits b (0),..., b ( M 1), where M is the number of bits in code word q transmitted on the physical channel in one subframe, ~ ( q) ~ ( q) (q) is scrambled resulting in a block b (0),..., b ( M 1) according to: [10] bit bit ( q) bit Where: ( q) ( q b ( i) c ( i) mod 2 ~ ( q) ) b ( i) (2.11) (2.12) Modulation ~ ( q) ~ ( q) (q) For each code word q, the block of scrambled bits b (0),..., b ( M 1) shall be modulated using one of the modulation schemes: QPSK, 16QAM, 64QAM, ( q) ( q) (q) resulting in a block of complex-valued modulation symbols d (0),..., d ( M 1). [10] bit symb 24

36 Modulation mapper The modulation mapper takes binary digits, 0 or 1, as input and produces complex-valued modulation symbols, x as output. QPSK In case of QPSK modulation, pairs of bits, b ( i), b( i 1) are mapped to complexvalued modulation symbols x according to Table 1 where x I jq. [10] Table 1 QPSK modulation mapping 16QAM b( i), b( i 1) I Q In case of 16QAM modulation, quadruplets of bits, b ( i), b( i 1), b( i 2), b( i 3) are mapped to complex-valued modulation symbols x according to appendix 2 where x I jq. 64QAM In case of 64QAM modulation, hex-tuplets of bits, b ( i), b( i 1), b( i 2), b( i 3), b( i 4), b( i 5) are mapped to complex-valued modulation symbols x according to appendix 3 where x I jq Demodulation Demodulation is the inverse operation for modulation that depends on the modulation mode, it process the input symbols to generate the demodulated output. Demodulation can be based on either hard-decision decoding or soft-decision decoding. In hard-decision decoding, the input symbols of a demodulator are mapped to estimated bits, whereas in soft-decision decoding the output is a vector of log-likelihood ratios (LLRs) Descrambling. The descrambling operation inverts the operations performed by the scrambler. The same pseudo-random sequence generator is used. However, there is a difference between bit-level scrambling and bit-level descrambling. Descrambling operations can be implemented in one of two ways. If, prior to the descrambling operation, a harddecision demodulation is performed, the input to the scrambler is represented by bits. 25

37 In this case, an exclusive-or operation between the input bits and the Gold-sequence bits will generate the descrambler output. On the other hand, if a soft-decision demodulation is performed prior to descrambling, the input signal is no longer composed of bits but rather of LLRs. In that case, descrambling is performed as a multiplication operation between the input log-likelihood values and Gold sequence bits transformed to coefficient values. A zero-valued Gold-sequence bit is mapped to 1 and a 1-valued bit is mapped to -1. [3] Rate Dematching. The inverse operations to those in the rate matching. We create a vector composed of dummy padded Systematic and Parity bits, place the available samples of the input vectors in the vector, and by de-interlacing and de-interleaving create the right number of LLR samples to become inputs to the 1/3 turbo decoder. [3] Channel decoding Turbo decoding. A turbo decoder is based on the use of two A Posteriori Probability (APP) decoders and two interleavers in a feedback loop. The same trellis structure found in the turbo encoder is used in the APP decoder, as is the same interleaver. The difference is that turbo decoding is an iterative operation. The performance and the computational complexity of a turbo decoder directly relate to the number of iterations performed. [3] Figure 8 Block diagram of a turbo decoder The heart of the decoding algorithm is a soft-decision decoding algorithm which provides estimates of the posterior probabilities of each input bit. The algorithm most commonly used for the soft-decision decoding algorithm is the MAP algorithm, also commonly known as the BCJR algorithm. [8] 26

38 The decoders are based on the MAP (maximum aposteriori probability) algorithm and output soft decision information learned from the noisy parity bits. Initially decoder 1 starts without initialization information (apriori estimates are set to zero). In subsequent iterations, the soft decision information of one decoder is used to initialize the other decoder. The decoder information is cycled around the loop until the soft decisions converge on a stable set of values. The latter soft decisions are then sliced to recover the original binary sequence. [11] The maximum a posteriori (MAP) decoding algorithm suitable for estimating bit and/or state probabilities for a finite-state Markov system is frequently referred to as the BCJR algorithm, after Bahl, Cock, Jelenik and Raviv who proposed it originally in. [8] The BCJR algorithm computes the posterior probability of symbols from Markov sources transmitted through discrete memoryless channels. Since the output of a convolutional coder passed through a memoryless channel (such as an AWGN channel or a BSC) forms a Markov source, the BCJR algorithm can be used for maximum aposteriori probability decoding of convolutional codes. [8] BCJR algorithm computes the a posteriori log-likelihood ratio, a real number defined by the ratio: [12] (2.13) Where is the corresponding information or message input bit to the symbol generated by the encoder at time k. and y is the received sequence to the decoder. The a posteriori log-likelihood ratio for BCJR algorithm is written in the following form: [12] (2.14) Where: At time k. The corresponding state is = s and the previous state is = s and means the summation is computed over all the state transitions from s to s that are due to message bits = +1 and R2 due to message bits = -1 The joint probability P (s, s, y): [12] (2.15) 27

39 Where: (2.16) Besides that, the probabilities α, γ and β can be expressed as follow: (2.17) For AWGN channel (memoryless): [12] (2.18) Where is a quantity that will cancel out when we compute the conditional LLR and is the channel reliability measure. The probabilities α and β are computed recursively. The respective recursive formulas are: [12] (2.19) Viterbi decoding: In 1967 by Andrew Viterbi was proposed the Viterbi algorithm (VA) and is used to decoding a bit stream that has been encoded using FEC code. [13] Viterbi algorithm can be explained by a trellis diagram it requires which comprises of minimum path and minimum distance Calculation and retracing the path. Fig-9 shows the block diagram of the Viterbi decoder. It consists of following blocks: [14] a. Branch Metric Unit (BMU) b. Path metric calculation 28

40 c. Add Compare and Select Unit (ACS) d. Trace Back Unit (TBU) SIMULATION RESULTS AND DISCUSSION Figure 9 the block diagram of the Viterbi decoder From the encoder output through the channel the BMU receives input data and computes a metric for each state and each input bit. There are two types of methods to calculate the metric. The metric which is used to calculate the Hard-decision encoded data is the Hamming distance and the metric is the Euclidian distance for Soft-decision encoded data. The metric is calculated for the entire path. The ACS unit is based on minimum distance calculations that are obtained from the previous row values. The Trace-back unit restore maximum likelihood path from the decisions made by BMU. This is the final stage of the Viterbi decoder where the input that was transmitted by using the convolution encoder is once again retrieved. [14] Viterbi algorithm is called optimum algorithm since it minimizes the probability of error. The Viterbi algorithm can be explained briefly with the following three steps. 1. Weigh the trellis; that is, calculate the branch metrics. 2. Recursively compute the shortest paths to time n, in terms of the shortest paths to time n-1. In this step, decisions are used to recursively update the survivor path of the signal. This is known as add-compare-select (ACS) recursion. 3. Recursively find the shortest path leading to each trellis state using the decisions from Step 2. The shortest path is called the survivor path for that state and the process is referred to as survivor path decode. Finally, if all survivor paths are traced back in time, they merge into a unique path, which is the most likely signal path. An example about the process of Conventional Decoding (Viterbi decoding algorithm) is in appendix 4. [18] The primary difference between the hard decision and soft decision Viterbi algorithm is that the soft decision algorithm cannot use hamming distance metric because of its limited resolution. A distance metric with needed resolution is the Euclidean distance. [15] 29

41 Chapter III: Simulation Results and Discussion Chapter III: Simulation Results and Discussion In this chapter we examine both transport block (DLSCH) processing and physical channel (PDSCH) processing (excluding the MIMO and OFDM operations) within two different channel models: an Additive White Gaussian Noise (AWGN) channel model and a Rayleigh channel model. 30

42 3.1 Simulation approach: Using MATLAB, we create functions for each of the processing steps, and then we create MATLAB test benches presented in figure 10 that implement DLSCH and PDSCH processing steps in a progressive manner in order to see the effect of each step on the performance of the system. Each program that includes a function at the transmitter should include its inverse function at the receiver with the channel model in between. The performance is represented with graphs of BER as a function of Eb/No which is a measure of the relative power of noise. Figure 10 presentation of all MATLAB test benshes 3.2 Simulation functions: The functions used in the presented test benches are described below: Modulator: This function computes the modulated symbols based on two inputs: the input bit stream (u) and a parameter specifying the modulation type (mode). The function implements QPSK using the comm.pskmodulatorsystem object by setting the modulation order to 4 whereas QAM 16 and QAM64 are implemented using comm.rectangularqammodulator System object with the modulation order set to 16 and 64 respectively. Demodulator: Since demodulation depends on the type of decoding we have two distinct demodulation functions: DemodulatorHard: associated with hard-decision decoding, its inputs are the received modulated symbols (u) and the modulation mode (mode), and its output are the demodulated bits. 31

43 DemodulatorSoft: associated with soft-decision decoding, its inputs are the received modulated symbols (u), the modulation mode (mode), and the estimate of the noise variance in the current subframe, and its output are LLRs. AWGNChannel: It adds AWGN noise to its input signal (u) using the comm.awgnchannelsystem object. It affects (u) according to the range of the Eb/No. RayleighChannel: It affects the input signal (u) with Rayleigh noise model according to the range of the Eb/No. Scrambler: It computes the scrambled version of the input bit stream (u) according to a parameter specifying the subframe index (ns) in the current frame. The function uses comm.goldsequencesystem object to generate the pseudo-random sequences according to the specifications of the LTE standard. The scrambled bits are result of an XOR operation between the input bits and the Gold sequence bits. Descrambler: It inverts the operations performed by the scrambler using the same comm.goldsequencesystem object with the same parameters as the scrambler; however, since there are two decoding modes, there exist two descrambler functions: DescramblerHard: associated with hard-decision demodulation which outputs bits that enter the DescramblerHard which produces the output bits as a result of the XOR operation between the input bits and the Gold-sequence bits. DescramblerSoft: associated with soft-decision demodulation which outputs LLR values that enter the DescramblerSoft. Descrambling is performed as a multiplication operation between the input LLRs values and Gold sequence bits transformed to coefficient values where a 0 bit is mapped to 1 and a 1 bit is mapped to 1. TurboEncoder: Performs turbo coding on the input stream (u), and it has a second input (intrlvrindices) to specify the interleaver indices. This function uses comm.turboencoder system object which describes the trellis structure of the constituent convolutional encoders. 32

44 TurboDecoder: Performs turbo decoding on the input stream (u). It has two additional inputs: (intrlvrindices) which specifies the interleaver indices as in the encoder, and (maxiter) which sets the maximum number of decoding iterations. This function uses comm.turbodecoder system object describing the same trellis structure of the TurboEncoder function. lteintrlvrindices: This function generates turbo code internal interleaver indices according to the block length specified in the input (blklen). It uses thegetf1f2 command to find the f1 and f2 parameter values based on BLKLEN. ConvolutionalEncoder: It generates the convolutional code for the input data stream (u) using comm.convolutionalencodersystem object that defines the trellis structure of the encoder. ViterbiDecoder: It accepts the input data stream (u) and decodes it using Viterbi algorithm by the comm.viterbidecodersystem object that specifies the trellis structure which should be similar to the one defined for the convolutional encoder. TurboDecoder_crc: Performs turbo decoding on the input stream (u) based on the CRC early termination mechanism. It has two inputs:(u), and(intrlvrindices)which specifies the interleaver indices as in the encoder. This function uses commlteturbodecodersystem object that is able to decode and to check the CRC. CbCRCGenerator: Generates the CRC bits for each input data frame (u) and it appends them to the input frame (u) using comm.crcgeneratorsystem object which specifies the generator polynomials defined in the LTE standard to be used by the CbCRCGenerator. CbCRCDetector: It accepts the input data stream (u) and computes its checksum to detect errors using comm.crcdetectorsystem object with the same generator polynomial defined for the CbCRCGenerator function. The final output is the received data stream without the checksum. 33

45 RateMatcher: It receives the input coded block (u) with the length of the code block (kplus) and performs the rate matching for the coded block with turbo Encoder function with the desired coding rate (rate). RateDematcher: it performs the inverse of RateMatcher function, by receiving the descrambled stream (in) which is coded with the chosen rate and the used length of the code block (Kplus) to provide turbo decoder with coded stream (1/3 code rate). TbChannelCoding: it carry out download shard channel processing or the transport block channel coding where it receives the transport block stream padded with transport block CRC (in) and perform subblock segmentation, turbo coding and rate matching using modulation mode as an input (modulationmode) and the desired coding rate (coding rate).it provides the chain of the system with coded blocks with the desired rate (out) and the number of code blocks (c) with their length (Kplus). TbChannelDecoding: it performs exactly the inverse of TbChannelCoding function where it receives the descrambled stream (in) and parameter that are imposed in the transmitting side including the number of code blocks for the transport block (C) and code block length (Kplus) with coding rate (CodingRate) and Modulation Mode (ModulationMode). Also it receives the maximum decoding iterations that can be executed (maxiter).this function gives the actual decoding iterations and the transmitted transport block padded with crc. 3.3 Simulation results Signal Modulation In this section we examine the effect of the three modulation schemes deployed by the LTE standard on the performance. The system is composed of a modulator, an AWGN channel, and a demodulator based on hard decision demodulation as shown in figure

46 Figure 11 Signal modulation test bench The result of the simulation is shown below: (a) (b) Figure 12 BER performance with QPSK, QAM16, and QAM64 for (a) AWGN channel and (b) Rayleigh channel For both channels, the obtained result shows that increasing the E b /N 0 ratio results in decreasing bit error rate for the three modulation schemes; however, to achieve transmission with a certain bit error rate,i.e. with a given quality of transmission, we require higher E b /N 0 as we move from QPSK to QAM16 to QAM64 modulation schemes progressively. This suggests that for a noisy channel, QPSK is used to provide robust communication at the expense of transmitting at low data rate, and as the channel becomes cleaner, QAM16 and QAM 64 are used to provide higher data rates. We notice that the bit error rates for Rayleigh channel model are much greater than for AWGN channel and this is due to fact that Rayleigh channel model is more close to real channel (reflections and scatterings). 35

47 3.3.2 Scrambling In this part we add to the previous system a scrambler and a descrambler before the modulator and after the demodulator respectively. Here we use soft-decision demodulation and the corresponding descrambling function. The system is shown in figure 13. Figure 13 Scrambling test bench The results of simulation are shown below: (a) (b) Figure 14 BER performance with the presence of scrambling for QPSK, QAM16, and QAM64 for (a) AWGN channel and (b) Rayleigh channel For both channel models, the results are similar to the ones obtained for the signal modulation part. This confirms that scrambling does not affect the quality of transmission in the channel since it is primarily used for identifying cells to randomize interference between neighboring ones. 36

48 3.3.3 Turbo coding In this part we implement a turbo-coding algorithm with a 1/3 coding rate to the previous system by adding a turbo encoder before the scrambler and a turbo decoder after the descrambler. Since turbo decoder input needs to be expressed in LLRs we use soft decision demodulator and descrambler. The system is shown in figure 15. Figure 15 Turbo coding test bench To demonstrate the efficiency of turbo coding algorithm we compare its performance when increasing the number of decoding iterations with a system consisting of an ordinary convolutional encoder and a Viterbi decoder. The system is shown in figure 16. Figure 16 Convolutional encoder/viterbi decoder test bench The results are shown below: 37

49 (a) (b) Figure 17 BER performance: Turbo coding vs convolutional\viterbi coding for AWGN channel for (a) QPSK and (b) QAM64 (a) (b) Figure 18 BER performance: Turbo coding vs convolutional\viterbi coding for Rayleigh channel for (a) QPSK and (b) QAM64 Comparing between convolutional coding and turbo coding, the result shows that the performance of turbo coding is better than convolutional coding for both AWGN and Rayleigh channels. Furthermore, for AWGN channel, and for any Eb/N0 value, the bit error rate becomes smaller as we increase the number of decoding iterations. This is also true for Rayleigh channel with QPSK modulation mode. For QAM 64 in Rayleigh channel, the performance decreases as the number of decoding iterations increases. This is due to the fact that the maximum a posteriori (MAP) decoding algorithm is suitable for estimating bit and/or state probabilities for a finite-state Markov system which is frequently referred to as the BCJR algorithm. It is 38

50 efficiently used to compute the posterior probability of the transmitted symbols through memoryless channels such as AWGN and BSC which is not the case for Rayleigh model. For Rayleigh model, the amplitude and the phase of the received signal can be seen as independent random variables; this is why different results are obtained for QPSK whose demodulation decision is made based on the phase of the received signal and for QAM whose demodulation decision is made based on the amplitude of the received signal. Transceiver computation time: Increasing the number of decoding iterations makes communication more robust for AWGN channel and QPSK of Rayleigh channel; however, this increases the computational complexity of turbo coding which can be described by the computation time it takes to complete the decoding operation. To illustrate this relation, the following table presents the elapsed time as we increase the number of turbo decoding iteration for QPSK modulation with EbNo=1 and max number of errors is equal to the maximum number of bits (10 6 ). Note that different CPUs have been use for these computations. Table 2 transceiver computation time for QPSK with different CPU for each channel model Maximum number of iteration in turbo coding Elapsed time (s) for AWGN channel Elapsed time (s) for Rayleigh channel To see which function contributes more in the complexity of the system, we execute a profiling script. The result is shown in figure 19 using QPSK with five decoding iteration for AWGN channel and figure 20 with one decoding iteration for Rayleigh channel: 39

51 Figure 19 Part from profiling results for turbo coding with 5 decoding iterations for AWGN channel Figure 20 Part from profiling results for turbo coding with 1 decoding iteration for Rayleigh channel From the profiling result, we deduce that performing turbo decoding with a fixed number of iterations can be considered as a drawback in the LTE standard as it takes about 85% of the entire system simulation time for AWGN channel and 68% for Rayleigh channel Early-Termination Mechanism In this part we examine the CRC early-termination mechanism by implementing the CRC generator before the turbo encoder and the CRC detector after the turbo decoder. The system is shown in figure

52 Figure 21 Turbo coding with early termination mechanism test bench In this part, we examine both the BER results and the computational complexity of a system with early-termination mechanism and without early-termination mechanism which uses a fixed number of decoding iterations. The obtained results are shown in the next pages: 41

53 (a) (b) (c) (d) Figure 22 Comparison of BER results with and without CRC-based early termination for QPSK for AWGN channel with (a) 3 decoding iterations (b) 6 decoding iterations (c) 10 decoding iterations and (d) 15 decoding iterations 42

54 (a) (b) (c) (d) Figure 23 Comparison of BER results with and without CRC-based early termination for QAM16 for AWGN channel with (a) 3 decoding iterations (b) 6 decoding iterations (c) 10 decoding iterations and (d) 15 decoding iterations 43

55 (a) (b) (c) (d) Figure 24 Comparison of BER results with and without CRC-based early termination for QPSK for Rayleigh channel with (a) 1 decoding iteration (b) 6 decoding iterations (c) 10 decoding iterations and (d) 15 decoding iterations From the above graphs we have the following: 1. Using turbo decoder with CRC termination mechanism results in better performance compared to using turbo decoder with less than six (1 and 3 ) iterations. 2. The performance of the system with CRC termination mechanism is similar to the one with six decoding iterations. 3. Increasing the number of iterations to 10, then to 15 gives better performance than the system with CRC termination mechanism. To see the effect of adding CRC termination mechanism on the computational complexity, we compare the execution time with and without this mechanism using a 44

56 fixed number of decoding iterations equals 10. The measurements are obtained for E b /N 0 =1 db and for a total number of bits equal to The results are shown below for the three types of modulation Table 3 Typical execution time with and without early-termination mechanism for different modulation modes. AWGN channel Regular time Early crc time Regular BER Early crc BER QPSK QAM QAM Rayleigh channel Regular time Early crc time Regular BER Early crc BER QPSK QAM QAM For each modulation scheme, we see that the system employing turbo decoding without CRC attachment consumes more time than the system employing turbo decoding with CRC attachment. As a conclusion, employing the CRC termination mechanism results in reducing the consumed time taken when we use turbo decoding with fixed number of iterations. The drawback of this mechanism is that it prevents achieving the communication performance that can be achieved with a high fixed number of iterations for AWGN channel and QPSK modulation of Rayleigh channel Rate matching In this part, we add the rate-matching operation after the turbo encoder and the rate-dematching operation before the turbo decoder to see the effects of using any 45

57 coding rate different from 1/3 that was used in the previous simulations. The system is shown in figure 25. Figure 25 Rate matching test bench The results are illustrated in the next pages: 46

58 (a) (b) (c ) Figure 26 Effect of rate matching on turbo coding BER performance for AWGN channel with (a) QPSK (b) QAM16 and (c) QAM64. 47

59 (a) (b) (c) Figure 27 Effect of rate matching on turbo coding BER performance for Rayleigh channel with (a) QPSK (b) QAM16 and (c) QAM64. The results show that to achieve transmission with a certain bit error rate, i.e. with a given quality of transmission, we require higher E b /N 0 when we use 1/2 as a coding rate than when we use 1/3, so the performance of a 1/3-rate transceiver is better than that of a 1/2 transceiver; therefore, as the channel becomes noisy, lower coding rates are used in transmission with more redundant bits to ensure robust communication. When the channel becomes cleaner, higher coding rates are used reducing the redundant bits and providing higher throughput. 48

60 3.3.6 LTE DLSCH and PDSCH Processing In this part we examine the entire transport block (DLSCH) processing and physical channel (PDSCH) processing (excluding the MIMO and OFDM operations). The system is shown in figure 28. Figure 28 LTE DLSCH and PDSCH test bench We examine the performance of the system according to different parameters: According to the modulation scheme: In this section we see the effect of using different modulation schemes with increasing the number of decoding iterations. The coding rate is 1/3. The results are shown in the next pages: 49

61 (a) (b) (c) Figure 29 BER performance: TrCh processing with increasing the number of decoding iterations for AWGN channel with (a) QPSK (b) QAM16 and (c) QAM64. 50

62 (a) (b) (c) Figure 30 BER performance: TrCh processing with increasing the number of decoding iterations for Rayleigh channel with (a) QPSK (b) QAM16 and (c) QAM64. The obtained results for AWGN channel show the following: 1. comparing between the three modulation schemes with a fixed number of decoding iterations say 3, we see that in order to achieve transmission with a certain bit error rate say 10-2, it requires an SNR of 0.2 db for QPSK, 2.4 db for QAM16, and 4.3 db for QAM64 confirming that higher order modulation is used only with clean channels to provide higher data rate transmission, whereas in noisy channels, QPSK is used to offer robust communication. 2. For the three modulation schemes QPSK, QAM16, and QAM64, and for any E b /N 0 value, the bit error rate becomes smaller with increasing the number of decoding iterations. However, the E b /N 0 level at which this effect is seen differs between the three modulation schemes where QAM16 and QAM64 require higher E b /N 0 level. We conclude that the transmission quality using higher order modulation if the channel is 51

63 too noisy can be improved only with a sufficiently high number of turbo decoding iterations. 3. The gain at which the performance gets better becomes smaller with increasing the number of decoding iterations. This shows the importance of CRC early termination mechanism in improving the TrCH processing specified in the LTE standard by compromising between offering an acceptable robust communication with reducing computational complexity. For Rayleigh channel the obtained results show that: 1- For QPSK, the bit error rate becomes smaller with increasing the number of decoding iterations. 2- For QAM16and QAM64, the bit error rate becomes greater with increasing the number of decoding iterations as explained in section 3. According to the coding rate: In this section we see the effect of using different coding rates on the performance of the transmission for each modulation scheme. The results are shown below: (a) (b) Figure 31 BER performance: TrCh processing for QPSK modulation for AWGN channel with different coding rates (a) 1/2 (b) 2/3 52

64 (a) (b) Figure 32 BER performance: TrCh processing for QPSK modulation for Rayleigh channel with different coding rates (a) 1/2 (b) 2/3 (a) (b) Figure 33 BER performance: TrCh processing for QAM16 modulation with 1/2 coding rate for (a) AWGN channel and (b) Rayleigh channel. 53

65 (a) (b) Figure 34 BER performance: TrCh processing for QAM64 modulation for AWGN channel with different coding rates (a) 1/2 (b) 2/3. (a) (b) Figure 35 BER performance: TrCh processing forqam64 modulation for Rayleigh channel with different coding rates (a) 1/2 (b) 2/3. The following table gives the SNR level in db for each modulation scheme which is required to achieve three BER values: 10-2, 10-3, and The results are obtained for three decoding iterations. 54

66 Table 4 SNR level required for BER values: 10-2, 10-3, and 10-4 for QPSK, QAM16, and QAM64 with different coding rates with three decoding iterations for AWGN channel. Modulation scheme QPSK QAM16 QAM64 Coding rate BER 1/3 1/2 2/3 1/3 1/2 1/3 1/2 2/ > > >8 The results confirm that higher coding rates that improve the throughput can be used only in cleaner channels. According to code block size: In this section we see the effect of using different code block size on the performance of the transmission. The results are shown below: (a) (b) Figure 36 BER performance: TrCh processing for QPSK modulation with different code block size for (a) AWGN channel and (b) Rayleigh channel. 55