DVB H Simulation

Final report

Simulation of video over DVB-H andQoE analyses

Author : Arulnambi Nandagoban1

Supervisors: Gerardo Rubino2, Kamal Singh3

Project Advisor :Antoine Chevreuil4

Laboratory : INRIA, RennesTeam: INRIA/DIONYSOS5

Master of science - Telecommunication and Signal processingESIEE,Paris

30 June, 2011

[email protected], INRIA, [email protected], INRIA, [email protected], INRIA, [email protected]://ralyx.inria.fr/2007/Raweb/dionysos/uid0.html

Abstract

The Personal Mobile Television is a commercial reality in many countries, and is becoming areality in other places. The Quality of Experience (QoE)actually delivered to end users will bea key criteria for the adoption of this new service. In order to evaluate it, a QoE analysingtool called Pseudo Subjective Quality Assessment (PSQA) is used in this project to evaluate thevideo quality with the simulation of video over DVB-H.

DVB-H transmitter and receiver in MATLAB are implemented in this project. The imple-mented DVB-H transmitter and receiver include Forward Error Correction (FEC)in the linklayer and Orthogonal Frequency Division Multiplexing (OFDM) in the physical layer as the im-portant parameters. A channel model is implemented in the physical layer to represent the realenvironment with the parameters like distance,speed of the terminal and type of environment.The simulation studies is made to optimize the Digital video broadcasting for handheld terminal(DVB-H) parameter with a novel approach of QoE using PSQA.

Contents

1 Introduction 41.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.2 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.3 Internship Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 DVB-H system 62.1 Link layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.1.1 Forward Error Correction mechanism . . . . . . . . . . . . . . . . . . . . 72.1.2 Time slicing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2.2 Physical layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.2.1 OFDM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.2.2 Channel model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

3 Quality of Experience and PSQA 113.1 Quality of Experience (QoE) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113.2 PSQA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

4 Simulink and Matlab 134.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134.2 Other features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

5 DVB-H Simulator model 145.1 Simulator overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145.2 Simulator model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145.3 Simulator implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155.4 Physical layer transmitter model . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

5.4.1 External encoder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165.4.2 External interleaver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165.4.3 Internal encoder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165.4.4 Internal interleaver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165.4.5 Mapper . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185.4.6 Pilot signals insertion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185.4.7 OFDM and Guard interval insertion . . . . . . . . . . . . . . . . . . . . . 18

5.5 Physical layer receiver model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195.5.1 Channel estimation and equalization . . . . . . . . . . . . . . . . . . . . . 19

5.6 Link layer model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205.6.1 Decapsulation methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

5.7 Channel model implemented . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

1

5.8 Problem faced and assumptions made . . . . . . . . . . . . . . . . . . . . . . . . 225.9 Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

6 Simulation results and analysis 246.1 Simulation parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246.2 Simulation results analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

6.2.1 Bit Error Rate analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246.2.2 Packet errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

7 Conclusion 29

8 Suggestions for the future work 30

9 Glossary 31Appendix A. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

Appendix A.1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32Appendix A.2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32Appendix A.3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

2

List of Figures

2.1 Functional block diagram of DVB-H system . . . . . . . . . . . . . . . . . . . . . 62.2 Link layer encapsulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72.3 Time slicing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.4 DVB-H transmission scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82.5 Guard Interval and Cyclic Prefix . . . . . . . . . . . . . . . . . . . . . . . . . . . 92.6 QPSK,16QAM and 64 QAM mapping . . . . . . . . . . . . . . . . . . . . . . . . 102.7 Urban and rural channel environments . . . . . . . . . . . . . . . . . . . . . . . . 10

3.1 QoE and QoS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113.2 PSQA training . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123.3 PSQA utilization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

4.1 Simulink example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

5.1 Global simulator model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145.2 Software hierarchiel model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155.3 Physical layer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175.4 Continual pilot tone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185.5 Scattered pilot tone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195.6 Section header . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215.7 TS header . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215.8 Section erasure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215.9 TS erasure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225.10 Typical urban profile (TU6) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225.11 Bit Error Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

6.1 Bit Error Rate after viterbi 16 QAM CR=3/4 . . . . . . . . . . . . . . . . . . . . 256.2 Bit Error Rate after viterbi 16 QAM CR=2/3 . . . . . . . . . . . . . . . . . . . . 256.3 Bit Error Rate after viterbi 64 QAM CR=2/3 . . . . . . . . . . . . . . . . . . . . 266.4 Bit Error Rate after viterbi QPSK CR=2/3 . . . . . . . . . . . . . . . . . . . . . 266.5 Bit Error Rate after RS decoder 16 QAM CR=3/4 . . . . . . . . . . . . . . . . . 266.6 Bit Error Rate after RS decoder 16 QAM CR=2/3 . . . . . . . . . . . . . . . . . 276.7 Bit Error Rate after RS decoder 64 QAM CR=2/3 . . . . . . . . . . . . . . . . . 276.8 Bit Error Rate after RS decoder QPSK CR=2/3 . . . . . . . . . . . . . . . . . . 286.9 IP packet Error Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

3

Chapter 1

Introduction

1.1 Introduction

The development of wideband telecommunications system is a modern trend in the currenttelecommunication industry. The multimedia services like video streaming and television ser-vices require high bandwidth. DVB-H [1] is one of the system which satisfies this requirement.It achieves various telecommunication challenges such as achieving high data rates in wirelessnetworks, providing the power saving options to the mobile receivers and the design of bandwidth-efficient Single frequency networks terrestrial (SFN). DVB-H is an evolution of DVB-T [2]. Theintention of its design is to improve performance in mobile environments, to add flexibility innetwork planning and to enable efficient power control in handheld receivers. Since the data iscarried in Internet Protocol (IP) packets, it is easy to adopt various services other than videostreaming. The television services are transmitted over DVB-H using the existing cellular net-work as a downlink. The challenge faced by the service providers is to find a best combination ofDVB-H parameters to achieve robust transmission over the channel. A lot of research was donewith different approaches to find the best combination of DVB-H parameters. Our approach is tooptimize the parameters with QoE. The QoE is analysed based on human perception of differentservices like video,audio etc. Here in this project, Pseudo Subjective Quality Assessment is usedto analyse the video quality and to provide the feedback in real time.

In this report, the focus will be on presenting the link layer and physical layer parameters ofDVB-H. Chapter 2 gives a brief overview of DVB-H system, presenting its link layer, physicallayer parameters and the new features added in these layers.The encapsulation methods anddifferent channel model for DVB-H are discussed in detail. Chapter 3 describes about the QoEand PSQA. Chapter 4 gives an overview of Matlab and Simulink. In chapter 5, the simulatordesign and its software implementation are elucidated in detail. The importance of simulatingDVB-H and problem faced are also discussed. Simulation results are interpreted and analysed inthe chapter 6. Finally, conclusions and possibilities for further research are presented in chapters7 and 8.

1.2 Motivation

In order to achieve success for DVB-H, or video broadcasting in general, the system parametersneed to be adapted for the QoE to be optimised. Different coding, interleaving and modulationmakes the amount of parameter combinations huge and these have to be critically evaluated

4

before building the networks. There are around 144000 combinations possible with these param-eters. At the physical layer one of the most important research objectives is to find appropriatechannel models to be abale to analyse different parameters and help in network planning.Tooptimize these parameters different approaches were taken into account and the one,which isproposed by DIONYSOS1 is to adapt these parameters based on QoE feedback.

1.3 Internship Objective

Since the combinations of physical and link layer parameter is extremely large, to analyse theseparameter is made easy by building a computer simulator model of DVB-H system. The ob-jective of this project work is to build a simulator, validate it with specification standard andfinally to study the video streaming over this simulator for QoE analyses using PSQA tool. Threephases of work are proposed to have a best strategy to finish off this project. The phase 1 is todedicated to establish state of the art and development of this software model. The phase 2 isplanned for validation and simultion work and the phase 3 is to work on the results obtained forQoE analyses. This work also gives the chance to evaluate the Forward Error Correction (FEC)performance of link layer and power saving mechanisim.

The main objective of this internship is to build a DVB-H simulator and after, study thevideo quality over it.

1A team of INRIA,Rennes which works on different problems related to the design and the analysis of com-munication services.

5

Chapter 2

DVB-H system

DVB-H is an ETSI specification for delivering broadcast services to battery-powered handheldreceivers[1]. DVB-H is a evolution of DVB-T specification for digital terrestrial television. Com-pared to DVB-T, television services to handheld devices are more complex from technical pointof view. The constraints starts right from the size of antenna to power capability of handhelddevices. The major constraints in handheld devices are mobility and its reception environment.They expect the reception in all types of environment like indoor,outdoor etc. The other problemis the power consumption which restricts the handheld capability for video services.

Figure 2.1: Functional block diagram of DVB-H system

The Figure 2.1 shows the functional block diagram of the DVB-H system. The DVB-H standard addresses the above stated problems by adding a number of features to DVB-Tstandard.DVB-H adds functional changes in the link and physical layers while it is backwardcompatible with DVB-T. In the link layer, DVB-H has two new features with respect to DVB-T:Time Slicing and Multiprotocol Encapsulation Forward Error Correction (MPE-FEC). In thephysical layer, DVB-H also adds some features DVB-T: DVB-H Signaling and 4K-Mode OFDM.A detailed overview of these parameters is presented in the following sections.

6

2.1 Link layer

2.1.1 Forward Error Correction mechanism

Forward Error Correction is used in the link layer for correcting the errors which occur due to theradio environment and its interference. The link layer packet encoding, encapsulation, decodingand decapsulation are illustrated in Figure 2.2. The IP datagrams are arranged column-wise intothe Application data table (ADT) of the MPE-FEC frame. To create MPE-FEC, IP packetsare filled into an (NX191) matrix where each cell of the matrix hosts one byte of information.According to the standard the number of rows in the matrix, i.e. N can be selected from thefollowing values: 256, 512, 768 or 1024. The ADT is then encoded row-wise with Reed-SolomonRS(255,191) and concatenated such that the final size of the matrix is of size (NX255). TheADT need not be completely filled. The unfilled part of the ADT is called padding. The paddingallows rate control and preventing fragmentation of IP packet between two MPE-FEC frames.All the 64 columns of RSDT need not be transmitted, i.e. the RS data table (RSDT) may bepunctured. This allows control of code rate[1]. The frame is divided into MPE-sections, with oneIP datagram as payload, and MPE-FEC-sections with one RS column as payload. Each sectionis protected with a CRC-32. The sections are transmitted in a MPEG-2 transport stream (TS).The decoding is suggested to be erasure correction based on reliability information provided bythe CRC. If the CRC fails, the bytes of the section are marked as unreliable. The reliability ofeach symbol is indicated in the Erasure Info Table (EIT). As erasure correction is twice as strongas conventional error correction, the RS(255,191) is able to correct 64 erasures on each row ofthe MPE-FEC frame.

Figure 2.2: Link layer encapsulation

2.1.2 Time slicing

Time slicing is a mechanism which reduces average power consumption of a DVB-H receiversignificantly based on time division multiplexing. To reduce the power consumption in mobilehandheld terminals, the service data is time-sliced (i.e segmented in time) and then sent throughthe channel as bursts at a significantly higher bit rate compared to the bit rate of the audio-visual service. Time-slicing enables a receiver to stay active only for a small fraction of the time,

7

while receiving bursts of a requested service. It significantly reduces the power consumption usedfor radio reception parts.

Figure 2.3: Time slicing

Figures 2.3 compares the services provided in typical DVB-T and DVB-H channels. It alsoshows the backward compatibility of DVB-H with DVB-T system. Time-slicing also supports asmooth seamless handover by accomplishing the changing of the reception from one transportstream to another during the off-time between bursts.

2.2 Physical layer

The physical layer of DVB-T is adopted for DVB-H with some additional features like 4K mode inOrthogonal Frequency Division Multiplexing (OFDM),Transmission Parameter Signalling(TPS)etc.,The transmission scheme of DVB-H is shown in Figure 2.4.

Figure 2.4: DVB-H transmission scheme

2.2.1 OFDM

OFDM concept is to segment the bandwidth into several sub-channels such that these narrowsub-channels can have flat fading. The feature of orthogonal sub-channels makes OFDM havea high spectral efficiency. Cyclic extension is a copy of the last or the forward part of eachOFDM symbol. It prevents inter symbol interference (ISI) and inter carrier interference (ICI),and makes the transmitted signal periodic. The new 4k mode [1] is added in the DVB-T standardfor OFDM transmission. DVB-T already has 2k and 8k mode. The 2k,4k and 8k refers to the

8

number of carriers used in the OFDM transmission. The diiference between 2k and 8k is thetransmitter distance and speed of the mobile terminal for a SFN. The 2k mode of OFDM allowsreception for short distance for fast moving terminal, whereas 8k mode allows long distance forslow moving mobile terminals.The 4k mode provides the tradeoff between these two modes ofOFDM by allowing a considerable distance for fast moving mobile terminals by giving moreflexibility in the transmission of DVB content.[3]

Guard Interval and Cyclic Prefix

While using OFDM, the distortion caused by ISI and ICI.To solve this problem, an empty Guardinterval (GI)between two consecutive symbols is added. If the length of GI is longer than thedelay spread of channel response, the next symbol doesnt interfere with the previous one. Butif symbol boundary estimation doesnt precisely locate the symbol, the empty GI destroys theorthogonality and introduces ICI. In order to prevent this situation, a mechanism is proposed tocopy the last part of an OFDM symbol into the empty GI, which is so called Cyclic Prefix (CP)as shown in Figure 2.5

Figure 2.5: Guard Interval and Cyclic Prefix

Other features

Randomiser disperses the energy in order to obtain an evenly distributed energy within the chan-nel and transport multiplex adaptation using the DVB randomisation polynomial 1 + x14 + x15

and disperses except the sync byte (0x47) of the TS packet[1].Reed-solomon encoder generates reed solomon packets based on the RS(204,188) code with

code generator polynomial galois field and field generator polynomialp(x) = x8 +x4 +x3 +x2 +1.It adds 16 parity bytes in the end of each information data and it can correct upto 8 error bytes[1].

Outer and inner interleaver performs the DVB outer interleaving function with depth I=12and bit interleaver(inner interleaver) concatenated with a symbol interleaver in a two-step pro-cess. If a hierarchical system is selected then the two streams are merged at this point using amodified interleaving equation. If required, DVB-H provides in-depth interleaving as specified bythe DVB standard. Both interleaver are used to reduce the impact of burst errors by spreadingthem in the channel[1].

QAM mapper which allows to perform QAM constellation mapping using the mapping schemespecified by DVB for QPSK, QAM16 or QAM64. It outputs I/Q QAM values to the FFT Fastfourier transform core as given in Figure 2.6.

9

Figure 2.6: QPSK,16QAM and 64 QAM mapping

2.2.2 Channel model

Figure 2.7: Urban and rural channel environments

The users of mobile TV Television may not watch the video for more than ten minutescontinuously. There are many places where these users watch TV on bus or the train (receptionin fast moving terminals), bars or restaurants (reception indoor)[4]. The proposed channel modelfor DVB-H are urban, rural and indoor environments as shown in the Figure 2.7. These channelsare modeled using rayleigh and rician distribution as per the type of environment.

10

Chapter 3

Quality of Experience and PSQA

3.1 Quality of Experience (QoE)

Figure 3.1: QoE and QoS

“Quality of Experience (QoE) is the overall acceptability of an application or service,as perceived subjectively by the end-user ”[5]. It also noted that QoE includes complete end-to-end system effects (client, terminal, network, services infrastructure etc) and is influnced by thecontext and user’s expectations. QoE related to the QoS in many context but also differs insome ways. Mostly QoS depends on the objective analysis of the service like delay, errors andbandwidth of a network when it is for a network system. But QoE deals with the user experienceas a function of network parameters, content and type of delivery of the service. It is interestingto note that more and more video service providers taking attention of the user experience toimprove their service. For mobile TV transmission, the signals degrade with respect to thetransmission channel by introducing errors into the signal. Moreover, the video compressionmade used for reducing the transmission bandwidth are the major aspects which affects the QoEof a end user.

11

3.2 PSQA

Figure 3.2: PSQA training

Figure 3.3: PSQA utilization

The assessment made by objective video quality analysing tools are not same as the humanperceived vedio quality. Whereas the subjective quality assessments are expensive in both costand time. A novel method proposed to assess the video quality is PSQA. It uses Random NeuralNetwork (RNN) for learning the examples obtained from the subjective tests. It is cost effectiveand simple to use. The methodology for the PSQA design and usage is given in Figure 3.2 and3.3. The idea behind its design is modelling a relation between the network parameters whichcause the impact on the video quality with the QoE results obtained from the human observerswho evaluate the sample videos. After that, the trained PSQA can be used in many real timeapplications to measure the subjective video quality.

12

Chapter 4

Simulink and Matlab

4.1 Overview

Simulink is a plaftform for multidomain simulation and model-based design of dynamic sys-tems. It provides an interactive Graphical User Interface (GUI) and lets you choose between thebuilt-in libraries to design, simulate, implement, and test various system like signal processing,communication and mechanical systems. It is integrated with MATLAB.Matlab1 is a programming language which allows to perform various numerical computations,data analysis and access. A simple example of Minimum shift keying (MSK) modulator modelis shown in Figure 4.1

Figure 4.1: Simulink example

4.2 Other features

Simulink has a lot of desirable features as compared to many other simulation tools. Someof its features are used in this project. It allows to store the values of the variable in theworkspace which can be retrieved for the later use. It uses subsystem methodology which allowsto encapsulate a group of blocks into a single block for reducing the model managing complexity.Moreover, the functions from different programming languages like C, C++, Python can becalled using user defined function block of simulink. In this project work,the link layer part isdesigned using Matlab code and the physical layer part is modeled using Simulink.

1www.mathworks.com

13

Chapter 5

DVB-H Simulator model

This describes about the simulator modelling with simulink and Matlab code. In section 5.1, aglobal architecture of the simulator is described. In the following sections, the developed unusedsimulink model is presented. In section 5.4, the physical layer is explained with respective matlabcode.

5.1 Simulator overview

DVB-H simulator is developed to evaluate the network parameters which affect the video qualityin DVB-H.It is developed using simulink and matlab.The physical layer part is modeled usingsimulink and link layer functions are developed using matlab codes.The existing communicationblocks of simulink are enhanced to adopt DVB-H standard and some new functions are developedto achieve the full functionality of DVB-H.In the following sections,the simulator model designstrategies and its implementations are discussed in detail.

5.2 Simulator model

Figure 5.1: Global simulator model

The global model for the simulator is divided into mainly two parts namely DVB-H simulatorand integration of PSQA module as shown in Figure 5.1. As for DVB-H simulator, a simple and

14

effiencient model is developed by including all the features of link layer and physical layer includ-ing the time slicing in link layer, randomiser and pilot insertion in physical layer. Consideringthe main objective, the focus is on the link layer for analysing the link layer parameters. Theparameters needed to evaluate the video quality are the frame errors and IP packet loss in thelink layer.

5.3 Simulator implementation

Figure 5.2: Software hierarchiel model

The software implentation is made as shown in Figure 5.2. Except the link layer part of thetrnsmitter and error calculation part, all other parts iterates for each burst in the chain.

5.4 Physical layer transmitter model

During the development phase of the work, the communication toolbox of simulink is used tomodel the physical layer of DVB-H. This model supports 3 types of modulations, 3 modes ofOFDM, all values of code rate in convolutional coding and all proportion of guard intervalinserion. During validation, due to problems like data syncronization and delay compensationof simulink blocks, a new model is coded in Matlab. So this model is used for the simulationstudies. Some example codes for blocks like RS encoder and decoder, convolutional ineterleavercan be found in Appendix.

The simulator has all the DVB-H processing functionality and it consists of several commu-nication and signal processing blocks as shown in Figure 5.3.

1. External encoder (RS encoder)

2. External interleaver (Convolutional interleaver,I=12)

15

3. Internal encoder(Punctured convolutional code)

4. Internal interleaver

5. Mapper

6. OFDM and Guard Interval Insertion

5.4.1 External encoder

The encoding and decoding is done with Reed-solomon coding. RS(204,188,T=8). It adds 16parity bytes at the end of each information byte of the TS packets. It is noted in [2] that inputBER1 required in the input of this coder is 2x10−4. In average this coder can correct upto 20million errors for each error it fails to correct. The following lines of code which are used toimplement RS encoder and decoder found inside the function rsendec.m.

5.4.2 External interleaver

Transmission errors corrupt many bits in the data stream. A convolutional interleaver rearrangesthe transmitted packets with the aim to increase the efficiency of the ReedSolomon decoding byspreading the burst errors introduced by the channel over a longer time. External interleaverinserts 11 bytes from other TS packets between bytes from the same TS packet (at the input).This allows burst errors of maximum 12x8=96 bytes to be corrected because only eight or fewerbyte error per TS packet are obtained after the deinterleaver in the DVB receiver/decoder. Theseerror can be corrected by the RS decoder with the efficiency of correcting 8 bytes.

5.4.3 Internal encoder

Internal encoder uses convolutional coding in addition with the ReedSolomon coder and externalinterleaver to improve the transmission effiency against errors. It is based on a mother convo-lutional code of rate 1/2 with 64 states (generator polynomials of the mother code are G1 =171OCT for X output and G2 = 133OCT for Y output). The puncturing is made if requiredand it reduces the redundancy of the mother code. It results the code rates 2/3, 3/4, 5/6 and7/8. For example for 2/3, puncturing for each 3 bit input , it outputs only 2 bits. It is a tradeoffbetween the bandwidth efficiency and redundancy of the system.

5.4.4 Internal interleaver

Two separate interleaving processes are used to reduce the bad impact of burst errors, oneoperating on bits (bit interleaver) and another on groups of bits (symbol interleaver). Dependingon the modulation mode QPSK, 16QAM or 64QAM the bit interleaver comprises two, four orsix paths. An input stream is demultiplexed into v sub-streams, where v = 2 for QPSK, v = 4for 16-QAM, and v = 6 for 64-QAM [1]. Three modes are defined for the COFDM multicarriermethod: 2K with 1705 carriers, 4k with 3405 and 8K with 6817 carriers. The use of the symbolinterleaver is to map v bit words to each OFDM symbol. The interleaver output data words arebrought together in 12 blocks of 126 bits in 2K mode,in 24 blocks of 126 bits in 4k mode and in48 blocks of 126 bits in 8K mode. The symbol interleaver processes the bit groups to generateCOFDM symbols.

1Bit Error Rate

16

Figure 5.3: Physical layer

17

5.4.5 Mapper

The mapper block used here can perform three types of modultion QPSK, 16QAM, 64QAM. Ituses the rectangular QAM block and in the demodulation, it uses the variance obtained fromthe modulator block to make decision.

5.4.6 Pilot signals insertion

There are two types of pilot tone found in the DVB-H. They are continual and scattered pilots.The continual pilots are found in Table.7 of [2]. The indices of pilot tones in OFDM can be seenin the Figure 5.4. Continual refers to the pilot tones which appear in the every OFDM symbols.In addition to continual pilot tones, scattered pilot tones appear in the OFDM symbol as wellin the pattern given in the Figure 5.5.

Figure 5.4: Continual pilot tone

5.4.7 OFDM and Guard interval insertion

OFDM modulation consists of N closely spaced orthogonal carriers of duration T0 (each oneis modulated with a conventional modulation scheme like QPSK, 16-QAM or 64QAM), with aspacing of 1/T0 between two consecutive carriers. Increasing the number of carriers does notmodify the payload bit rate, which remains constant. In DVB-T OFDM uses 2048,4096 or 8192carriers (2K ,4k and8K mode). Every OFDM block is added with cyclic prefix in the end of theblock. The cyclic prefix serves as a guard interval and eliminates the intersymbol interferencefrom the previous symbol. Insertion of the guard interval extends symbol duration by 1/4,1/8, 1/16 or 1/32 to give the total symbol duration TS . By the end of the guard interval, allechoes caused by multipath reception, reception of other transmitters in the SFN (SFN, SingleFrequency Network) or Doppler effects in mobile reception, i.e. all fading effects, must havesettled or decayed.

18

Figure 5.5: Scattered pilot tone

5.5 Physical layer receiver model

The blocks in the transmitter are reversed in the receiver part. The channel estimation andviterbi decoder parts are added in addition to the transmitter model in the receiver. Harddecision is choosen for the moment in the receiver for demodulations. The delay is calculated foreach combination of the above stated parameter for the viterbi decoding and error calculations.The BER calculations are made in two places of the receiver. One after the viterbi decoding andone at the end of the transmitter-receiver chain. These values are used for the validation of thesimulator.

5.5.1 Channel estimation and equalization

It is well known that the wireless channel causes an arbitrary time dispersion, attenuation, andphase shift in the received signal. the use of OFDM and a cyclic prefix mitigates the effect oftime dispersion. The channel estimation methods proposed here will make use of pilot signal toestimate the time-variant channel response in the frequency domain. Here in this simulator, twotypes of channel estimation techniques are used. As we know, the received signal is a filteredand noise-corrupted version of the transmitted sequence:

rk = sk ⊗ ck + nk (5.1)

The multipath channel causes frequency selectivity and inter symbol interference. Equalizationcan reduce the inter symbol interference and noise effects for the better demodulations. Thepurpose of equalization is to remove the effect of multipath effect from the received signals. It isdone by estimating the channel ck and dividing the received signal with the estimated channelresponse cke. This simple equalizer is implemented in the model. The estimation is performedusing the boosted pilot signals rkp.

Method 1:

The steps of channel estimation and equalization using method 1 are given by

• Taking IFFT of rkp

19

• Filling zeros in the place of useful carriers

• Taking FFT

• Divide received signal in frequency domain with interpolated rkp

Method 2:

This method uses both the continual and scattered pilots for the channel estimation i.e channelestimation is done in both frequency and time axis. For this method atleast 10 OFDM symbolsare used for interpolation. The steps of channel estimation and equalization using method 2 aregiven by

• Taking IFFT of rkp

• Linear interpolation in time and frequency axis

• Taking FFT

• Divide received signal in frequency domain with interpolated rkp

Both methods are implemented but the method 1 is used for simulation studies.

5.6 Link layer model

The link layer model is developed using matlab code. It performs the MPE encapsulation totill TS encapsulation. The headers for MPE sections and TS packets are made as per thespecifications [1]. The MPE encapsulation is made as per the Figure 2.2. For the decoding,sectionerasure chosen.

Burst formation

In this simulator model, many assumption were made with respect to the burst formation. Here,the duration of one burst is in the range of 0.5 seconds. For the given IP packets, 0.5 secondscorresponds to 30 IP packets. A simple study is made on FEC percentage with respect touncorrected IP packets. Since, the objective of this project has less interest in time slicing, weconsider this assumption.

MPE and TS encapsulation

To make the decoder more efficient, many useful information are sent through headers of sectionand TS headers. The Figures 5.6 and 5.7 show the informations contained in the section andTS headers. The section header informations like section length, index information and synchro-nization byte are used to place the IP and FEC bytes in the corresponding positions in ADT andRS table. The same way, TS header contains informations like syncronization bytes and sectionstart indicator helps to decapsulate TS packets to MPE sections.

5.6.1 Decapsulation methods

There are two decapsulation and decoding methods that are proposed during the course of thisproject. This methods are found more efficient with respect to [6]. A brief description andimplementation are discussed in the section erasure and TS erasure sections.

20

Figure 5.6: Section header

Figure 5.7: TS header

Section erasure

Figure 5.8: Section erasure

In the decoding process, section erasure method is considered for the moment. The processworks as given in Figure 5.8. Section erasure is based on the CRC. A section is consider as lostor erased when CRC fails. This is marked in EIT as unreliable symbols. Later this EIT is usedduring the RS decoding process (in link layer) for the better decoding. The disadvantage of thismethod is that it marks many reliable bytes as unreliable considering the CRC. There is othermethod proposed by researcher to overcome this problem is TS erasure[6]. It will be discussedin the following part of the this topic.

21

TS packet erasure

Another method in the decoding process which is implemented with respect to section erasureis TS packet erasure. But for the moment, it is not used during the course of simulation studies.The algorithm works as shown in the Figure 5.9. TS earsure is based on the RS decoder in thephysical layer. If RS decoder fails to retrieve the data bytes, then the corresponding TS packetis erased and replaced with zeros. This is given as input for the proceedings steps. This methodis performed without using the CRC.

Figure 5.9: TS erasure

5.7 Channel model implemented

A profile which represents terrestrial propagation in an urban area is implemented in this simu-lator. It has been defined by COST 207 as a typical urban (TU6) profile and is made of 6 pathshaving wide dispersion in delay and relatively strong power. The profile parameters are shownin the Figure 5.10. This channel was used to validate the simulator.

Figure 5.10: Typical urban profile (TU6)

5.8 Problem faced and assumptions made

Variable size IP packets

DVB-H works with variable size IP packets and the objective is to replicate the real demodulator.This complicate the design of decapsulator. As we know from previous sections, most of theaddressing informations are carried by the header of the section. During transmission, if someaddressing informations are erased due to the radio channel, then it makes complication in placingthe IP datagrams in the ADT table and also in retrieving the reliability information for decoding.So, a assumption is made which states that the section length information in the section headeris received without any errors.

22

Multipath channel model

A multipath channel model is constructed corresponding to TU6 profile but it is not used duringthe course of simulations due to high frequency selectivity behaviour. A further work will devotedto find a solution for this problem.

Long simulation time problem

During the second phase of this project, a simulink model was developed for the physical layer.During the validation period, the problem related to data synchronization, compiler and sim-ulation time were faced. Due to time constraint, instead of solving the problem in simulink, amodel is developed using Matlab code.

5.9 Validation

The model developed was validated before the simulation studies as per as DVB-H standardrequirements. From the Figure 5.11, it has been shown that the simulator respects the recom-mandations of DVB-H specifications. The expectation for 16-QAM,CR=3/4 with CNR=12 dB,DVB-T QEF or MFER 5% is achieved. Here , we can notice that the first condition is achieved.

Figure 5.11: Bit Error Rate

23

Chapter 6

Simulation results and analysis

6.1 Simulation parameters

The simulation is done with including the link layer decoding and decapsulation of 30 secondsvideo. This video flux is in the form of IP packets. These IP packets are given as input tothe simulator. The important system parameters are defined before the start of the simulation.Once the simulation is started, it takes approximately 40 minutes to complete transmitter-receiver chain for a burst. The error analysis is made in the physical layer to confirm the properadaptation of the simulator to the specification and packet error statistics are calculated in thelink layer for QoE analysis. The data of this video was simulated by varying the SNR in theGaussian channel and with different modulation and code rate. Here, the simulation are madewith following parameters

Number of rows 256,512,768,1024Inner code 2/3, 3/4Modulation QPSK, 16 QAM, 64 QAMOFDM mode 2K

6.2 Simulation results analysis

6.2.1 Bit Error Rate analysis

In DVB-H receiver chain, there are two places where BER is measured. One is after the viterbidecoder and other one is after RS decoder. All of the measurement given in this section followsthe above statement. The impact of AWGN noise can be clearly seen in the Figures 6.1 to 6.8 .On analysing the BER after RS decoder, we can determine the performance of the decapsulation.Since the simulator uses the section erasure as decapsulation method, BER of 10−11 is expectedafter RS decoder for error free reception of IP packets. The results shown in the following figureproves the proper working of the simulator. But,the results which are obtained cannot be usedfor the QoE analyses.

24

Figure 6.1: Bit Error Rate after viterbi 16 QAM CR=3/4

Figure 6.2: Bit Error Rate after viterbi 16 QAM CR=2/3

6.2.2 Packet errors

A packet error analysis was made based on the errors in IP packets is shown in Figure 6.9. Itis visible in this curve that higer valued SNR produces the lesser IP packet error rates. Thisresult is used to achieve the final goal of the project. Since, PSQA is works with function of IPpacket error rate, it is interesting to analyse this result. Due to time constraint, we stopped thesimulation only with QPSK and CR=2/3. In the future, a elaborate study will be made to havethis kind of results.

25

Figure 6.3: Bit Error Rate after viterbi 64 QAM CR=2/3

Figure 6.4: Bit Error Rate after viterbi QPSK CR=2/3

Figure 6.5: Bit Error Rate after RS decoder 16 QAM CR=3/4

26

Figure 6.6: Bit Error Rate after RS decoder 16 QAM CR=2/3

Figure 6.7: Bit Error Rate after RS decoder 64 QAM CR=2/3

27

Figure 6.8: Bit Error Rate after RS decoder QPSK CR=2/3

Figure 6.9: IP packet Error Rate

28

Chapter 7

Conclusion

The main objective of this thesis work is to develop a DVB-H simulator. In order to develop thissimulator, Matlab has been proposed as a software platform, section erasure has been proposed asdecapsulation and decoding method. The goal of this project is achieved by sucessfully designingthe simulator. The problems like getting higher BER values when using urban model couldn’tbe resolved due to time constraint to finish the project.

The application presented allows simulation of the DVB-H transmission in OFDM mode2k,4k and 8k with QPSK,16QAM and 64QAM along with error corrections. Moreover, it willcontribute to the video quality analyses based on the QoE using PSQA tool.

The BER dependencies after viterbi and after RS decoder for the AWGN channel was comparedwith referenced measured values. The variation between the referenced measured value and theactual simualtion results turned out to be good.

29

Chapter 8

Suggestions for the future work

The section erasure method adds too much unnecessary erasures, resulting in failure of decodingmore often than necessary. Most of the symbols, marked as erasure by CRC decoder are actuallyreceived correctly [6]. So, TS erasure method can be build in the link layer of the receiver for thecomparison with the section erasure using PSQA. It is interesting to study the different typesequalizer and its impact on the quality of video. A discussion on different decoding methods andstudy can be made for the best analysis of QoE.

30

Chapter 9

Glossary

• QoE → Quality of Experience

• PSQA → Pseudo Subjective Quality Assessment

• FEC → Forword Error Correction

• OFDM → Orthogonal Frequency Division Multiplexing

• DVB-H → Digital Video Broadcasting for handheld device

• SFN → Single Frequency Network

• DVB-T → Digital Video Broadcasting terresrial

• MPE → Multiprotocol Encapsulation

• ADT → Application Data Table

• RSDT → Reed Solomon Data Table

• CRC → Cyclic redundancy check

• TS → Transport Stream

• EIT → Erasure Info Table

• TPS → Transmission Parameter Signalling

• ICI → Inter Carrier Interference

• GI → Guard Interval

• CP → Cyclic Prefix

• QAM → Quadrature Amplitude Modulation

• QPSK → Quaddrature Phase Shift Keying

• FFT → Fast Fourier Transform

• TV → Television

31

• RNN → Random Neural Network

• GUI → Graphical User Interface

• MSK → Minimum Shift Keying

• BER → Bit Error Rate

• SNR → Signal to Noise Ratio

• dB → Decibel

• QEF → Quasi Error-Free

• AWGN → Additive White Gaussian Noise

• SER → Symbol error rate

• MFER → MPE-FEC Frame Error Rate

Appendix A

Appendix A.1

The codes for RS encoder and decoder can be found in the following lines:

f unc t i on r sout=rsendec ( datain , txrx )i f ( txrx==10)

% Create Reed−Solomon encoder ob j e c t .enc = f e c . r s enc ( 2 5 5 , 23 9 ) ;enc . ShortenedLength = 51 ;data in1=reshape ( datain , 188 , l ength ( data in ) / 1 8 8 ) ;r sout1 = encode ( enc , data in1 ) ;r sout=reshape ( rsout1 , numel ( r sout1 ) , 1 ) ;%rsout=reshape ( code ,204∗ sz ( 2 ) , 1 ) ;e l s e i f ( txrx==01)data in1=reshape ( datain , 204 , l ength ( data in ) / 2 0 4 ) ;dec=f e c . r sdec ( 2 5 5 , 23 9 ) ;dec . ShortenedLength = 51 ;[ code , cnumerr , ccode ] = decode ( dec , data in1 ) ;r sout=reshape ( code , numel ( code ) , 1 ) ;endend

Appendix A.2

The source code for internal interleaver can be found in the following lines:

f unc t i on [ out sym int zero pad ] =sym intv new (y , ra t e id , mode , sym table , txrx )i =1;j =0;out sym int = [ ] ;switch (mode)

32

case 2n car =1512;

case 4n car =3024;

case 8n car =6048;

endswitch ( r a t e i d )

case {1 ,2}m=2;

case {3 ,4 ,7}m=4;

case {5 ,6}m=6;

endn=m∗ n car ;

i f ( txrx==10)zero pad=n−rem ( length ( y ) , n ) ;

r =[y ; z e r o s (n−rem ( length ( y ) , n ) , 1 ) ] ;whi l e ( i<=length ( r )/n)

x=r ( j +1: i ∗n ) ;i n t o u t=i n t r l v (x , sym table ) ;out sym int =[ out sym int ; i n t o u t ] ;j=i ∗n ;i=i +1;

ende l s e i f ( txrx==01)

whi l e ( i<=length ( y )/n)x=y ( j +1: i ∗n ) ;i n t o u t=d e i n t r l v (x , sym table ) ;out sym int =[ out sym int ; i n t o u t ] ;zero pad =0;j=i ∗n ;i=i +1;

end%b i t i n t l v o u t=b i t i n t l v o u t 1 ( 1 : l ength ( b i t i n t l v o u t 1 )−b i t z e r o 1 ) ;

end

Appendix A.3

The codes shown below encapsultes IP packets with MPE header.

f unc t i on [ mpe f ec c rc2 ]= m p e f e c s e c t i o n ( f e c s e c , n , pad co l )m=0;p=1;

j j =1;mpe f ec c rc2 = [ ] ;

33

mpe fec c r c = [ ] ;s z f e c=length ( f e c s e c ) ;

whi l e ( (m<=s z f e c−n)&&(p<=(s z f e c /n)))&& ( j j<=length ( pad co l ) )f e c 1=f e c s e c (m+1:p∗n , : ) ;l =1;whi l e l<=64

mpe h=mpe fec header1 (n , l , 6 4 , pad co l ( j j ) ) ;mpe fec =[mpe h ’ ; f e c 1 ( : , l ) ] ;mpe f ec c rc1=crc32compute ( mpe fec ) ;mpe f ec c rc2 =[ mpe f ec c rc2 ; mpe f ec c rc1 ] ;l=l +1;

endm=p∗n ;p=p+1;j j=j j +1;

endend

34

Bibliography

[1] ETSI. Digital Video Broadcasting (DVB); Transmission System for Handheld Terminals(DVB-H). European Telecommunication Standard, November 2004.

[2] ETSI. Digital Video Broadcasting (DVB); Framing Structure, channel coding, and modulationfor digital terrestrial television. European Telecommunication Standard, January 2004.

[3] M.Kornfield and G.May. Dvb-h and ip datacastbroadcast to handheld devices. In Proceedingsof the 2010 Third International Conference on Software Testing, Verification, and ValidationWorkshops.

[4] H. Knoche and J. D. McCarthy. Mobile users needs and expectations of future multimediaservices. In WWRF12.

[5] ITU-T SG12. Definition of Quality of Experience. COM12 -LS 62 - E,TD 109rev2(PLEN/12),Geneva,Switzerland, 16-25 January 2007.

[6] Heidi Joki and Jussi Poikonen. Analysis and Simulation of DVB-H Link Layer. June 2005.

[7] Fisher.W. Digital video and audio broadcasting technology:A practical guide for engineers,2nded. 2008.

[8] Radim.S and Tomas.K. Influence of the transmission channel type and error correction ondvb-t error rates. In Proceedings of the 50th International symposium ELMAR-2008.

35