Modeling and Analysis of Synchronization Schemes for the TDMA Based Satellite Communication System

Transcription

1 A Dissetation entitled Modeling and Analysis of Synchonization Schemes fo the TDMA Based Satellite Communication System by Chong Wang Submitted to the Gaduate Faculty as patial fulfillment of the equiements fo the Docto of Philosophy Degee in Engineeing D. Junghwan Kim, Committee Chai D. Paticia R. Komunieci, Dean College of Gaduate Studies The Univesity of Toledo Decembe

2

3 An Abstact of Modeling and Analysis of Synchonization Schemes fo the TDMA Based Satellite Communication System by Chong Wang Submitted to the Gaduate Faculty as patial fulfillment of the equiements fo the Docto of Philosophy Degee in Engineeing The Univesity of Toledo Decembe As a multiple access technique, time division multiple access TDMA is becoming moe widespead in satellite communications. Although single-caie SC modulation techniques has been mostly employed in the TDMA satellite communication system, multi-caie modulation techniques, lie othogonal fequency division multiplexing OFDM, ae also pomising candidates fo the futue high ate satellite communication due to its high bandwidth efficiency and obustness to fading. As is nown, accuate synchonization is a ey equiement fo efficient data tansmission in TDMA systems. Because of the channel impaiment o popagation delay, the eceive may expeience the caie fequency offset CFO o symbol timing eo STE that may esult in difficulty of signal detection and estimation. Compaing with SC systems, OFDM systems ae moe sensitive to synchonization eos, because these eos will lead to inte-symbol intefeence ISI and inte-caie intefeence ICI. Although many synchonization algoithms have been disclosed in liteatue as the effective ecovey methods to educe detection eo in the pesence of CFO as well as STE, tadeoffs still have to be made between the estimation accuacy and estimation ange o iii

4 implementation complexity. This motivates futhe eseach to impove the pefomances of synchonization methods fo the TDMA based satellite communication system. The main objectives of this eseach ae modeling and mathematical analyses of the synchonization methods unde vaious channel impaiment scenaios towads the pactical applications in TDMA satellite communication systems. To achieve these objectives, this eseach fistly investigates the pefomance degadation due to CFO fo both SC and OFDM systems. The pefomance of SC system impaied by CFO is mathematically analyzed and compute simulation is pefomed unde additive white Gaussian noise AWG. Fo OFDM system in the pesence of CFO, the pefomance of binay phase shift eyingquadatue phase shift eying BPSKQPSK-OFDM impaied by CFO is evaluated and simulated unde AWG, Rayleigh flat fading, and fequency selective fading. Because of the lac of analyses in highe ode modulation o modulation with non-ectangula decision egion, this eseach deives the closed fom of symbol eo pobability expessions fo 8-ay phase shift eying 8PSK-OFDM and 6- ay quadatue amplitude modulation 6-QAM-OFDM impaied by CFO unde AWG. All of the coesponding simulation esults validate the coectness of the theoetical analyses and symbol eo pobability deivations. In ode to compensate the effect of CFO to SC systems, numeous methods have been poposed as the effective achitectue fo CFR. Thus, compaison between vaious existing synchonization algoithms unde cetain condition has been pefomed based on modeling and mathematical analysis. By the esults, the majo impact factos of diffeent CFR algoithms have been identified. Then, a novel CFR algoithm, which is deived iv

5 based on least squae estimation method, is poposed and evaluated unde AWG. Finally, the poposed algoithm is simulated and compaed with othe exiting algoithms with espect to estimation ange, estimation accuacy and computational complexity. The simulation esults show that the poposed algoithm has the benefit of educed computational complexity and high estimation accuacy. In the case of OFDM systems, two novel data-aided DA caie fequency and symbol timing ecovey algoithms ae poposed. These maximum lielihood ML based algoithms jointly estimate the CFO and STE fo the OFDM systems. The pefomance of the poposed algoithms is evaluated and compaed with existing estimation algoithms in tems of estimation ange and accuacy unde AWG, Rayleigh flat fading and fequency-selective multipath fading, espectively. Simulation esults confim the effectiveness of the poposed algoithms unde vaious levels of signal to noise atios SRs and the obustness of diffeent lengths of cyclic pefix CP of the peamble. v

6 Acnowledgements I would lie to expess my deepest gatitude to my adviso, D. Junghwan Kim, fo his valuable guidance and advice thoughout my academic studies. I would also lie to than my committee membes, D. Mansoo Alam, D. Mohsin Jamali, D. Thomas Stuat and D. Dong-Shi Kim fo thei valuable advice, excellent suppot, and guidance duing my dissetation eseach. My sincee appeciation is extended to Macia King-Blandfod fo poviding me the oppotunity to wo at the povost office. Without he suppot, it would have been difficult fo me to complete my Ph.D study. It is eally a geat woing envionment whee I have leaned a lot. I would also lie to appeciate the povost staffs: Babaa Monge, D. Penny Poplin Gosetti, Magaet F. Taband, Donna Belofe fo thei ind help and encouagement. Special thans to my good fiends D. Mie Oa, D. Joe Moening, Xueliang Lu, Pooja Raoane, Ayoade Oguntade, and Mauli Oza who not only assist me with the eseach wo, but also help me quicly blend into Ameican cultue. It has been a geat pleasue to meet them all. Finally, I would lie to expess my deepest gatitude to my paent fo thei suppot and encouagement. Without thei love, I would have neve been possible to each hee. vi

7 Contents Abstact...vi Acnowledgements...vi Contents... vii List of Tables... xii List of Figues... xiii List of Abbeviations... xviii Chapte... Intoduction.... Motivation of Reseach.... Bacgound and Pevious Wos...5. Reseach Objective.... Dissetation Oganization... Chapte... Synchonization in TDMA based Satellite Communication Systems.... Intoduction of TDMA based Satellite Communication System...5. Model of TDMA based Mobile Satellite Channel...7. Identification of Synchonization Methods.... Evaluation Citeia fo Synchonization Methods...5 Chapte...9 vii

8 Caie Fequency Recovey CFR in Single-caie Systems...9. Modeling and Desciption of Synchonization in Single-caie TDMA System.... Pefomance Degadation due to Caie Fequency Offset CFO in Single-caie System...5. Conventional Caie Fequency Offset Estimation Algoithms Delay and Multiply-Based Algoithms Least Squae Based Algoithms Autocoelation Based Algoithms...9. Poposed Caie Fequency Offset Estimation Algoithms... Chapte...7 Pefomance Evaluation of Caie Fequency Recovey in Single-caie Systems...7. Pefomance Degadation Due to Caie Fequency Offset...7. Pefomance Evaluation of Conventional Caie Fequency Offset Estimation Algoithms Pefomance Evaluation of Delay and Multiply Based Algoithms Pefomance Evaluation of Least Squae Based Algoithms Pefomance Evaluation of Autocoelation Based Algoithms...5. Pefomance Evaluation of Poposed Caie Fequency Offset Estimation Algoithms Pefomance Evaluation of the Poposed Algoithm with espect to Estimation Range...6 viii

9 .. Pefomance Evaluation of Poposed Algoithm with espect to Estimation Accuacy...6. Pefomance Compaisons of Caie Fequency Offset Estimation Algoithms...66 Chapte Caie Fequency and Symbol Timing Recovey in Multi-caie Systems Model and Desciption of Synchonization in Multi-caie Systems Effect of Caie Fequency Offset on the Pefomance of OFDM Systems Effect of Caie Fequency Offset on the Pefomance of BPSKQPSK OFDM Effect of Caie Fequency Offset on the Pefomance of 8PSK OFDM unde AWG Effect of Caie Fequency Offset on the Pefomance of 6-QAM OFDM unde AWG Conventional Caie Fequency and Symbol Timing Recovey Methods Data-aided DA Caie Fequency and Symbol Timing Recovey Methods on-data-aided DA Caie Fequency and Symbol Timing Recovey Methods Poposed Joint Caie Fequency and Symbol Timing Recovey Methods Peamble Sequence and Its Popety Mathematical Model fo Symbol Timing Recovey Mathematical Model fo Caie Fequency Offset Estimation...6 ix

10 Chapte Pefomance Evaluation of Caie Fequency and Symbol Timing Recovey Methods in Multi-caie Systems Pefomance Degadation due to Caie Fequency Offset in Multi-Caie Systems Pefomance Degadation of BPSKQPSK OFDM due to Caie Fequency Offset Pefomance Degadation of 8PSK OFDM due to Caie Fequency Offset Pefomance Degadation of 6-QAM due to Caie Fequency Offset Pefomance Evaluation of Conventional Symbol Timing Recovey Methods Timing Metics MSE Pefomance ove AWG MSE Pefomance ove Rayleigh Flat Fading MSE Pefomance unde Multi-path Fequency Selective Rayleigh Fading Pefomance Evaluation of Conventional Caie Fequency Offset Estimation Methods S-cuve MSE Pefomance unde AWG MSE Pefomance unde Rayleigh Flat Fading... x

11 6.. MSE Pefomance unde Multi-path Rayleigh Fequency Selective Fading Pefomance Evaluation of Poposed Joint Symbol Timing and Caie Fequency Recovey Methods Timing Metic and S-cuve of the Poposed Methods MSE Pefomance of the Poposed Methods unde AWG MSE Pefomance of the Poposed Methods unde Rayleigh Flat Fading MSE Pefomance of the Poposed Methods unde Multi-path Rayleigh Fequency Selective Fading...7 Chapte Conclusions and Futue Reseach Conclusions Futue Reseach... Refeences... xi

12 List of Tables. Simulation scenaios fo pefomance degadation due to caie fequency offset...8. Compaison of the computational complexity Simulation Paametes fo OFDM System...8 xii

13 List of Figues - Pinciple of TDMA based satellite system 6] TDMA Fame Stuctue 6] Bloc diagam of oveall caie and symbol timing ecovey system BPSK SER pefomance degadation due to caie fequency offset QPSK SER pefomance degadation due to caie fequency offset omalized mean fequency offset estimation using delay and multiply algoithm with diffeent length of M SR=dB omalized fequency offset estimation eo vaiance using delay and multiply algoithm with diffeent length M ΔfT= omalized fequency offset estimation eo vaiance using delay and multiply algoithm with diffeent peamble size L ΔfT= omalized mean fequency offset estimation using Kay algoithm omalized fequency offset estimation eo vaiance using Kay algoithm with diffeent peamble size L ΔfT= omalized mean fequency offset estimation using L&R algoithm with diffeent length of M SR=dB omalized fequency offset estimation eo vaiance using L&R algoithm with diffeent length M ΔfT=...56 xiii

14 - omalized fequency offset estimation eo vaiance using L&R algoithm with diffeent peamble size L ΔfT= omalized mean fequency offset estimation using Chen algoithm with diffeent length of M SR=dB omalized fequency offset estimation eo vaiance using Chen algoithm with diffeent length M ΔfT= omalized fequency offset estimation eo vaiance using Chen algoithm with diffeent peamble size L ΔfT= omalized mean fequency offset estimation using Fitz algoithm with diffeent length of M SR=dB omalized fequency offset estimation eo vaiance using Fitz algoithm with diffeent length M ΔfT= omalized fequency offset estimation eo vaiance using Fitz algoithm with diffeent peamble size L ΔfT= omalized mean fequency offset estimation using M&M algoithm with diffeent length of M SR=dB omalized fequency offset estimation eo vaiance using M&M algoithm with diffeent length M ΔfT= omalized fequency offset estimation eo vaiance using M&M algoithm with diffeent peamble size L ΔfT= omalized mean fequency offset estimation using poposed algoithm with diffeent length of M SR=dB...6 xiv

15 - omalized fequency offset estimation eo vaiance using poposed algoithm with diffeent length of M ΔfT= omalized fequency offset estimation eo vaiance using poposed algoithm with diffeent peamble size L ΔfT= Pefomance of nomalized mean fequency offset estimation SR=dB Pefomance compaison of nomalized fequency offset estimation eo vaiance ΔfT=.5, L=8, M= Pefomance compaison of nomalized fequency offset estimation eo vaiance ΔfT=., L=8, M= Pefomance compaison of nomalized fequency offset estimation eo vaiance ΔfT=., L=8, M= Bloc diagam of OFDM tansmitte and eceive Bloc diagam of synchonization pocess in the OFDM eceive Signal space diagam fo 8PSK Signal space diagam fo 6-QAM Stuctue of poposed OFDM peamble P Stuctue of poposed OFDM peamble P Pobability of symbol eo fo BPSK OFDM ove AWG channel =8, g = Pobability of symbol eo fo QPSK OFDM ove AWG channel =8, L=.. 6- Pobability of symbol eo fo BPSK OFDM ove Rayleigh flat fading channel with CP =8, g =.... xv

16 6- Pobability of symbol eo of QPSK OFDM ove Rayleigh flat fading channel with CP =8, g = Pobability of symbol eo fo BPSK OFDM ove fequency selective Rayleigh fading channel Tap=, = Pobability of symbol eo fo BPSK OFDM ove fequency selective Rayleigh fading channel Tap=5, = Pobability of symbol eo fo BPSK OFDM ove AWG channel with lage FFT size Pobability of symbol eo fo QPSK OFDM ove AWG channel with lage FFT size Pobability of symbol eo fo 8PSK OFDM ove AWG channel =8, g =.7 6- Pobability of symbol eo fo 8PSK OFDM ove AWG channel with lage FFT size Pobability of symbol eo fo 6-QAM OFDM ove AWG channel =8, g = Pobability of symbol eo fo 6-QAM OFDM ove AWG with lage FFT size Compaison of timing metic of existing estimatos g = Compaison of timing metic of existing estimatos g = MSE of the timing offset estimation ove AWG channel MSE of the timing offset estimation ove Rayleigh flat fading channel MSE of the timing offset estimation ove fequency selective Rayleigh fading channel....7 xvi

17 6-8 Compaison of S-cuves of the conventional fequency offset estimatos MSE of the fequency offset estimation ove AWG channel MSE of the fequency offset estimation ove Rayleigh flat fading channel MSE of the fequency offset estimation ove fequency selective Rayleigh fading channel Timing metic of the poposed estimatos S-cuves of the poposed estimatos MSE of the poposed timing eo estimatos ove AWG channel MSE of the poposed fequency offset estimatos ove AWG channel MSE of the poposed timing eo estimatos ove Rayleigh flat fading channel MSE of the poposed fequency offset estimatos ove Rayleigh flat fading channel MSE of the poposed timing eo estimatos ove multi-path Rayleigh fequency selective fading channel MSE of the poposed fequency offset estimatos ove multi-path Rayleigh fequency selective fading channel...8 xvii

18 List of Abbeviations 8PSK...8-ay Phase Shift Keying 6-QAM...6-ay Quadatue Amplitude Modulation AWG...Additive White Gaussian oise BER...Bit Eo Rate BPSK...Binay Phase Shift Keying CBTR...Caie and Bit Timing Recovey CFO...Caie Fequency Offsets CFR...Caie Fequency Recovey CPO...Caie Phase Offsets CHF...Chaacteistic Function CDMA...Code Division Multiple Access CAZAC...Constant Amplitude Zeo Autocoelation CRLB...Camè-Rao Lowe Bound CP...Cyclic Pefix DA...Data-Aided DD...Decision Diected DAB...Digital Audio Boadcasting DOL...Diect On Line DVB-S...Digital Video Boadcasting Satellite Second Geneation DVB-SH...Digital Video Boadcasting Satellite Sevices to Handhelds DVB-T...Digital Video Boadcasting Teestial DSB...Diect Satellite Boadcasting DFT...Discete Fouie Tansfom EEV...Estimation Eo Vaiance FB...Feedbac xviii

19 FF...Feed-fowad FT...Fouie Tansfom FDMA...Fequency Division Multiple Access ICI...Inte-Caie Intefeence ISDB-T...Teestial Integated Sevices Digital Boadcasting ISI...Inte-Symbol Intefeence LSE...Least Squae Estimation LOS...Line Of Sight M-PSK...M-ay Phase Shift Keying MAPE...Maximum A Posteioi Estimation MLE...Maximum Lielihood Estimation MSE...Mean Squae Eo MILSATCOM...Militay Satellite Communications MMSE...Minimum Mean-Squae Estimation MCRB...Modified Camé Rao lowe bound MA...Multiple Access DD...on Decision Diected DA...on-Data Aided OFDM...Othogonal Fequency Division Multiplexing PSD...Powe Spectal Density PDF...Pobability Density Function P...Pseudo-oise QAM...Quadatue Amplitude Modulation QPSK...Quadatue Phase Shift Keying RDMA...Random Pacet Division Multiple Access RB...Refeence Bust RRC...Root Raised Cosine S-DMB...Satellite Digital Multimedia Boadcasting xix

20 SR...Signal to oise Ratio SDMA...Space Division Multiple Access SER...Symbol Eo Rate SSD...Signal Space Decomposition STE...Symbol Timing Eo STR...Symbol Timing Recovey TDMA...Time Division Multiple Access TB...Taffic Bust UW...Unique Wod VSAT...Vey Small Apetue Teminal xx

21 Chapte Intoduction Satellite communication is to use the satellite as a elay to povide data exchanges between two o moe eath stations fixed o mobile fo long distance communication. The coveage aea of satellite communication can geatly exceed that of a teestial system, and the tansmission cost of a satellite is independent of the distance fom the cente of the coveage aea. Moeove, satellite to satellite communication is vey pecise ]. Theefoe, satellite communication is inceasingly used fo global communications, as well as fo adio and television boadcastings ]. In the satellite communication system, thee ae mainly five pincipal foms of multiple access MA techniques, which allow seveal teminals connected to the same multi-point tansmission medium to tansmit ove it and to shae its capacity ]. In fequency division multiple access FDMA, all uses shae the satellite at the same time, but each use tansmits its data at a unique allocated fequency. With time division multiple access TDMA all uses ae sepaated by time, pesenting the sequential data tansmission in tansponde at the same caie fequency. In code division multiple access CDMA technique all uses tansmit othogonally coded spead spectum signals that can be sepaated at the eceive by coelation with the tansmitted code ]. Space

22 division multiple access SDMA is a scheme whee all concened eath stations can use the same fequency at the same time within a sepaate space available fo each lin ]. Random Pacet division multiple access RDMA is a scheme whee a lage numbe of satellite uses shae the same tansponde asynchonously by andomly tansmitting shot bust o pacet divisions ]. Among these five multiple access techniques, TDMA has been widely employed in seveal pactical applications of satellite communication systems, because the TDMA signals can be divided by time, thus they ae easily econfigued fo changing taffic demands, and esistant to noise and intefeence. Moeove, the entie bandwidth of the system is used fo all the uses, only one signal is pesented in the tansponde at one time, thus eliminating the caie to inte-modulation noise and educing the consequences of othe non-linea impaiments 5]. In the TDMA system, both single-caie modulation and multi-caie modulation can be employed, 6]. The application of single-caie TDMA system can be found in the satellite systems such as vey small apetue teminal VSAT satellite communication systems 7-8], digital video boadcasting satellite second geneation DVB-S 9-], diect satellite boadcasting DSB ] and high speed satellite TDMA gound system HiSTARS of KROEASAT- ]. On the othe hand, as a special case of multi-caie modulation, othogonal fequency division multiplexing OFDM modulation has been attacted much inteest in many wieless and boadcast applications such as IEEE 8.WIFI, IEEE 8.6WIMAX, GPP LTE ], as well as Euopean digital audio boadcasting DAB, digital video boadcasting teestial DVB-T, teestial integated sevices digital boadcasting ISDB-T, and digital subscibe line xdsl 5]. Simultaneously OFDM modulation is also becoming pomising scheme fo

23 satellite communication systems 6-8], and it has been adopted by digital video boadcasting satellite sevices to handhelds DVB-SH 9-]. On the ETSI side, the ETSI S-UMTS goup initiated seveal new activities to the study of OFDM-based ai intefaces fo G mobile satellite systems ]. Moeove, the application of OFDM techniques in the physical laye of a satellite system devoted to the povision of satellite digital multimedia boadcasting S-DMB sevices has also been consideed -]. Futhemoe, OFDM modulation has ecently been exploed in joint teminal engineeing office JTEO eseach effots fo identifying the optimal ai-inteface fo militay satellite communications MILSATCOM netwos though wideband global SATCOM WGS systems 6].. Motivation of Reseach In a TDMA based satellite system, it is well nown that pefect synchonization is cucial towad tansmission quality, because the synchonization eo causes sevee inte-symbol intefeence ISI o inte-caie intefeence ICI which eventually leads to a consideable pefomance degadation 5]. The majo synchonization issues encounteed in the TDMA based system ae caie synchonization, symbol timing synchonization, fame synchonization and netwo synchonization 5-6]. Caie synchonization o called caie ecovey efes to the technique to estimate and compensate caie fequency offsets CFO and caie phase offsets CPO in the eceive. CFOCPO may due to the mismatch of oscillatos between the tansmitte and eceive, o the nonlinea chaacteistic of the wieless channel, such as Dopple shift 7].

24 In TDMA-based satellite data tansmission, because thee is no discete spectal component povided to the eceive at the symbol ate, the demodulato theefoe has no diect nowledge of exactly when the eceived symbols stat and end. Even when the incoming symbol ate is nown, the eceive needs to deduce the phase of the eceived symbols elative to its own local symbol cloc efeence. Symbol timing synchonization o called symbol timing ecovey is defined to estimate symbol timing eo STE and identify the pecise moment of time when individual symbols stat and end 5]. Fame synchonization plays an impotant ole in multi-uses channels to identify the boundaies between diffeent uses and establish whee the infomation is coming fom and to whee it must be outed. etwo synchonization is also essential in twoway communication system. Bit aiving at a given multiplexe must be available at the ight time so that the assigned time slots ae coectly filled and no bits ae lost 6]. o matte what modulation schemes ae used, both STE and CFO will degade the pefomance of TDMA system. Thus, it is impotant to estimate and compensate these synchonization eos to guaantee the eliable data tansmission. ote that in the OFDM system, although the effects of ISI caused by STE can be significantly educed by inseting cyclic pefix CP, STE still causes the phase otation of the sub-caies in fequency domain even if STE is shote than the CP. Thus, the estimation accuacy of STE is also citical to the pefomance of OFDM systems as CFO estimation. Because the symbol duation fo single-caie modulation scheme is vey shot unde high ate tansmission, single-caie modulation system is moe vulneable to ISI. On the othe hand, OFDM can effectively ovecome ISI by the insetion of CP. Howeve, OFDM is much moe sensitive to CFO than that of single-caie modulation system,

25 because the CFO can destoy the othogonality of sub-caies and lead to seve ICI. Thus, both single-caie and OFDM-based multi-channel systems have thei own advantages and disadvantages. The selection of eithe of these schemes should depend on the specific application scenaios. Since CFO can cause significant degadation to the pefomance of TDMA system, and most synchonization schemes cannot completely compensate this offset, it is also impotant to quantitatively identify the effect of CFO, which can establish citeia to evaluate diffeent synchonization schemes. Instead of tansmitting data in a continuous steam, TDMA systems always opeate in a bust mode with a set of nown modulation symbols inseted to the use data pacet at the tansmission fo synchonization. In this study, caie fequency ecovey CFR and symbol timing ecovey STR schemes have been fully investigated and exploed fo the application of bust mode TDMA based satellite system employing both single-caie and OFDM modulation schemes.. Bacgound and Pevious Wos Most of the available synchonization methods can be simply classified into two categoies: eithe data-aided DA o non-data aided DA also nown as blind synchonization algoithms. In geneal, DA algoithm exploits the infomation fom nown pilot symbols o peamble taining sequences, and it usually yields high estimation accuacy with the tadeoff of low implement efficiency 8]. On the contay, DA algoithm usually esults in low estimation accuacy with computation ovehead, but no need of exta data tansmission. Moeove, based on the way how to estimate 5

26 synchonization eos, the synchonization schemes can also be classified as joint o sepaate estimation algoithms. In the case of single-caie modulation based TDMA system, peamble taining sequence DA method is always used to povide synchonization infomation because of its supeio estimation accuacy compaed with DA method. Much effot has been devoted to find optimum estimation algoithms fo caie and symbol timing ecovey 9]. The efeence ] poposed a computationally efficient CFR algoithm based on least squae method, which can yield good pefomance at high signal to noise atio SR egion. M. Luise and R. Reggiannini ] pesented an efficient technique fo fast CFR based on ML estimation theoy, which poves to be closed to the Came-Rao lowe bound CRLB in the absence of fading effects. Jingxin Chen 7] poposed a fast caie ecovey CR method using a shot peamble, which is applicable to bust mode communication system. Michael P. Fiz innovated a pactical plana filteed caie fequency estimation algoithm by exploiting the ML estimato ]. He extended his eseach ] and poposed a new fequency estimato, which is suitable to the communication system that equies high speed, and ecusive fequency estimation. It was found that it can obtain nea optimum pefomance at the modeate signal to noise atio. A new data aided estimation algoithm was also invented by Umbeto Mengali and Aldo. D Andea ] and its theshold is lowe and acquisition ange is lage than those of Fitz ] and L&R algoithm ] fo a given estimation accuacy. Towads accuate estimation and pope compensation of the synchonization eos in the data ecovey of OFDM symbols, multiple efficient synchonization ando estimation algoithms have been poposed 5, 5]. Moose 6] poposed a DA method 6

27 to estimate the fequency offset using two epeated OFDM symbols, but the estimation ange of this ML-based algoithm is limited to the half of the sub-caie spacing. Schmidl and Cox 7] pesented a DA method using a peamble with two identical halves fo joint symbol timing and caie fequency synchonization. Since thei method does not conside the effect of cyclic pefix CP towad its maximum lielihood ML estimato, the timing metic exhibits a plateau 8] fo whole duation of the CP, which inevitably yields lage vaiance in the timing estimate. To educe the vaiance of 7] caused by the CP, Minn 8] initially poposed a method in which Schmidl and Cox s timing metic is aveaged ove a window of CP length. Although this method can alleviate the plateau poblem of 7], the estimation vaiance is still lage due to the compaable timing metics aound the optimum timing. Late Minn 8-9] also poposed anothe method by modifying the Schmidl and Cox s peamble with a new sequence, which contains fou equal pats but the last two pats with diffeent sign. This method gives elatively smalle estimato vaiance but still exhibits lage mean squae eo MSE in the multipath fading channel. To ovecome the deficiency of Minn s appoach, Pa ] pesented a novel DA timing offset estimation method using a peamble symbol which is consist of fou pats, with two symmetic pats at the beginning and thei espective conjugates at the end of the peamble. His method yields significantly smalle mean squaed eo MSE. Howeve, the fequency offset estimation cannot be obtained fom Pa s method and the pefomance is seiously degaded unde multipath fading condition. In ode to impove the pefomance unde fading channel, Seung ] poposed a novel DA method, which uses a specially designed peamble to achieve small MSE unde fading. In, ], the supe-imposed OFDM timing and fequency scheme 7

28 is poposed by including a pe-defined pseudo-noise P sequence. The diffeence between these two methods is that the P sequence used in ] is in the fom of constant amplitude zeo autocoelation CAZAC sequence and the peamble is same as that of 7]. Although this method has the advantage of bette bandwidth utilization, it will equie exta powe to tansmit the supe-imposed P sequence. Instead of utilizing the coelation of the identical pats in the peamble, Bin ] developed a novel symbol timing eo estimation method based on the cyclic stuctue of the OFDM symbol and the DFT-based channel estimato. This DA method ] uses the noise subspace of channel estimates to find the stating point of the symbol. Although it was claimed that the fequency offset can be estimated fist by using the identical method of Schmidl s 7] because of the same taining sequence used, no details ae available on the way to deal with fequency offset in this method, since the fequency offset is ecoveed afte the symbol timing synchonization in Schmidl s method. As to the DA estimation algoithm, most algoithms ae based on cyclic coelation. Pa 5] poposed a blind synchonization algoithm, which fistly exploits the second-ode cyclostationaity of the eceived signals and then using the infomation in the cyclic coelation. Jan-Jaap 6] poposed a DA joint synchonization method fo OFDM system using a coelation with CP. Howeve, the estimation ange of this ML based algoithm is limited to the half of the sub-caie spacing only. In 7], anothe cyclic coelation based STR method is poposed, and it is poved that the maximum of this method points to the coect timing offset, iespectively of channel conditions. In addition, inh 8] also poposed a cyclic coelation based synchonization method. Howeve, his CP consists two pats. One pat is a copy of the head of the last symbol and the othe pat is a copy of the tail of that 8

29 symbol. Besides the cyclic coelation based algoithm, Daniel 9] poposed a ML symbol timing estimato by exploiting both the edundancy in the CP and available pilot symbol used fo channel estimation. Daniel also pesented anothe DA timing synchonization method using both the cyclic pefix and the pilot symbols., but his estimato is highly sensitivity to the caie fequency vaiation. Consideing the evaluation of multi-caie system based on OFDM impaied by CFO, numeous esults ae available on the evaluation of eo pobability of OFDM systems. As is summaized in 5], mainly two diffeent methods ae widely used to quantify the effect of ICI caused by CFO: Gaussian appoximation and chaacteistic function CHF. In 5-5], the eo pobability is deived by appoximating the ICI as a Gaussian pocess based on the cental limit theoem. Howeve, this appoach is poved to be inaccuate fo low SR 5]. On the othe hand, Sathananthan and Tellambua deived the exact Bit Eo Rate BER o Symbol Eo Rate SER fo OFDM with CFO using chaacteistic function CHF of the ICI unde additive white Gaussian noise AWG 5]. The main weaness of 5] is that the esults of the aveage eo pobability ae expessed in the fom of an infinite seies expansions athe than a closed fom eo pobability fomula. In 55], the authos evaluate the eo pobability of OFDM impaied CFO by analyzing the CHF of the eceived signal s eal and imaginay components sepaately. Howeve, the eal and imaginay pats of the eceived signal ae not independent. Recently, in 5], the authos follow the pocedue pesented in 5], and deive an exact closed fom of BERSER expession using the elationship between sum and poduct of cosine functions fo OFDM systems with CFO. The esults of 5] have been futhe impoved by 56], because the agument of Q-function is assumed to be 9

30 positive in 5]. Howeve, both of them only conside the OFDM system employing M- ay phase shift eying MPSK modulation schemes, and do not povide any analysis fo the OFDM employing quadatue amplitude modulation QAM modulation scheme. As can be seen fom the existing esults fo pefomance evaluation in the pesence of CFO, pactical evaluation of eo pobability equies appoximation of infinite seies and solution is limited to the applications of BPSK and QPSK cases with ectangula decision egion mostly. Fo highe ode modulation lie 6-QAM and the cases with nonectangula decision egion lie 8-ay phase shift eying 8PSK, thee ae only few esults 5, 57] based on infinite seies expansion. With espect to pefomance evaluation of single-caie system impaied by CFO, it is easonable to nomalize the CFO as in the OFDM system fo a fai compaison. In 58-6], the eo pobability of single-caie system in the pesence of CFO is evaluated based on the Gaussian appoximation pocess fo the ICI tems. Howeve, the Gaussian appoximation method has been poved to be less accuate when SR is high. Although many synchonization methods ae cuently available in the liteatue, most of them have been concentated on vey specific cicumstances only. Hence, it is had to genealize a synchonization method, which is obust unde diffeent cicumstances. Moeove, many synchonization methods ae highly dependent on paticula paametes, which may intoduce a tadeoff between the estimation accuacy and estimation ange. Thus, initially it is necessay to povide a complete assessment on the synchonization methods to identify thei limitations. Then futhe eseach is still needed to impove the synchonization pefomances in estimation accuacy, estimation ange and implementation complexity fo a obust application to diffeent cicumstances.

31 By the esults, the eliable and efficient synchonization methods fo the TDMA based satellite system can be detemined finally.. Reseach Objective The majo objective of this eseach will focus on developing and evaluating the feasible synchonization schemes to mitigate the pefomance degadation of TDMA based satellite communication system unde vaious citical synchonization impaiments and diffeent channel intefeences. To accomplish this majo objective, extensive eview has been initially pefomed to analyze the conventional synchonization methods used to estimate CFO fo singlecaie modulation based TDMA system. And then the pefomance in tems of estimation ange and accuacy of conventional synchonization methods is discussed and evaluated unde AWG channel by both mathematical deivation and compute simulation. The cucial factos fo each conventional synchonization methods have been detemined, and the tadeoff between estimation ange and accuacy o implementation complexity is identified. Finally, this study develops an impoved CFR algoithm to mitigate the pefomance degadation due to CFO. Then, the pefomance of poposed CFR algoithm has been evaluated and compaed with the conventional CFR algoithms with espect to estimation ange, accuacy and implementation complexity. Since single-caie modulation based TDMA systems suffe geat pefomance degadation unde fading effect, and the mobile satellite channel is dominant by multipath fading intoduced intefeence, this eseach also conside the application of multi caie OFDM-TDMA system, because of its geat immunity to fading and ISI. In ode

32 to investigate the suitable synchonization methods in the OFDM system, this eseach initially analyze the eo pefomance of OFDM system impaied by CFO unde vaious channel envionments. Due to the lac of existing pefomance evaluation fo highe ode modulation schemes and modulation schemes with non-ectangula decision egion, this study also gives a fully investigation on eo pefomance of 8PSK OFDM and 6- QAM OFDM unde AWG. Secondly, the pefomance of conventional synchonization methods fo OFDM systems is discussed and evaluated with espect to estimation ange and accuacy, espectively. Finally, this study pesents two novel obust joint CFR and STR methods fo multi-caie OFDM systems towad the impoved the estimation accuacy. Then, the pefomance of poposed algoithm is compaed with othe exiting joint CFR and STR methods though compute simulations and theoetical analyses.. Dissetation Oganization The est of this dissetation is oganized as follows. Chapte two biefly descibes TDMA based satellite system and its channel model. In ode to evaluate diffeent synchonization methods, the estimation citeia and classification of synchonization methods ae also povided. In Chapte thee, the pefomance degadation of single-caie modulation based TDMA system impaied by CFO is mathematically evaluated fistly. Then, the conventional CFR methods ae numeically investigated with espect to estimation ange and accuacy. Based on the CFO estimation fomula, the cucial paametes influencing the pefomance of each CFR method ae identified. Finally, this chapte explains the poposed CFR though mathematical deivations.

33 Chapte fou pesents on the degadation of eo pobability due to CFO in the single-caie modulation based TDMA system. The exiting synchonization algoithms ae evaluated and compaed though compute simulations. In addition, the pos and cons of diffeent synchonization methods ae pesented based on the simulation esults. Finally, the pefomance evaluation of poposed method is conducted though compute simulation. And its impoved estimation accuacy at the high signal to noise atio SR egion with educed computational ovehead will be illustated. In Chapte five, fistly, the symbol eo pefomance of OFDM TDMA system is analyzed in the pesence of CFO, and the symbol eo pobabilities of 6-QAM OFDM and 8PSK OFDM systems impaied by CFO unde AWG ae mathematically deived. Then, multiple conventional DADA CFR and STR methods ae intoduced. In the end, two new joint CFR and STR methods ae poposed, and numeically evaluated in detail. Chapte six povides the simulation esults on the pefomance degadation of OFDM system in the pesence of CFO and the mathematical esults deived in Chapte five ae veified. Then, the poposed joint CFR and STR methods ae evaluated and compaed with the conventional synchonization methods. Based on the simulation esults, the poposed methods ae poved to be obust unde diffeent intefeences. Finally, some concluding emas and suggestions fo futue eseach ae made in Chapte seven.

34 Chapte Synchonization in TDMA based Satellite Communication Systems As discussed in Chapte one, the synchonization is vey impotant fo the TDMA based satellite communication systems. In ode to analyze the synchonization methods in both single-caie TDMA and OFDM-based multi-caie TDMA systems, it is necessay to establish the model of the TDMA based satellite system and its channel models fistly. Since the fixed satellite communication channel is pimay a line of sight LOS lin fo the uplin and downlin, the channel noise in LOS lin is geneally modeled as the AWG. Howeve, the channel envionment fo mobile satellite communication is much moe complex. Tansmission between a satellite and a mobile teminal on the gound is geneally no longe just simple LOS lin. The adio wave may encounte a multiplicity of obstacles in the path, subjecting the tansmitted wave to eflections, diffaction, and scatteing, esulting in a multiplicity of ays eaching the eceive antenna 6]. Also, because the tansmitte o eceive teminal is moving, the powe eceived is also vaying, esulting in signal fading 6]. As in statistics, thee ae vaious estimation methods, and each of them has its own advantage depending on the implementation scenaios. Thus, it is necessay to

35 classify the estimation methods into diffeent categoies at the beginning. Moeove, some estimation citeia, such as bias and eo vaiance, must be taen into consideation fo evaluating the pefomance of an estimato. Howeve, it is geneally expected that the estimato can be unbiased, and achieve a small eo vaiance as much as it can. Theefoe, CRLB usually establishes a fundamental lowe limit to any unbiased estimato. Since no estimato can povide lowe eo vaiance than this bound, it can be seved as an evaluation standad. Hence, any estimation method is aiming to impove its accuacy towads this bound. Unfotunately, this CRLB cannot be easily computed in many pactical situations. Instead the modified CRLB MCRB is mostly pefeed to educe computational complexity 6].. Intoduction of TDMA based Satellite Communication System The TDMA based satellite communication system typically consists of a set of taffic stations also efeed to as teminals and a maste efeence station as shown in Figue - 6]. Figue -: Pinciple oftdma based satellite system 6]. 5

36 In this technique, a fame is divided into multiple time slots, which ae shaed among the teminals. A maste efeence station with visibility of the whole netwo maintains access discipline and scheduling duing the fame. It tansfes a efeence bust to povide fame synchonization and signaling infomation and identifies the tansmitting stations. Refeence bust RB will be followed by the taffic busts, which ae tansmitted by the coesponding taffic stations. These bust of TDMA fame contain a peamble sequence, which is used to povide the synchonization infomation. Figue - 6] gives the composition of efeence and taffic busts stuctue. RB: Refeence Bust, TB: Taffic Bust, GT: Guad Time, CR: Caie Recovey, BTR: Bit Timing Recovey, UW: Unique Wod, SC: Signaling Channel, TD: Taffic Data Figue -: TDMA Fame Stuctue 6]. Refeence bust RB is tansmitted by a designated efeence station to achieve the stat of fame fo all eath stations involved. The fame so-defined is divided into time slots, and each eath station is assigned a time slot in which to tansmit its infomation. Then, each eath station sends its own taffic bust TB, among which also includes peamble sequences fo caie ecovey and symbol timing duing one station. Since thee ae significant time delays between uses, each use eceives the efeence bust with a diffeent phase, and its taffic bust is tansmitted with a coespondingly diffeent phase within the time slot. Thee is theefoe a need fo guad times to tae account of this uncetainty. Each time slot TS is theefoe longe than the peiod needed fo the 6

37 actual taffic bust, theeby avoiding the ovelap of taffic bust even in the pesence of these popagation delays. Then, a unique wod UW sequence is followed to pefom the fame and bust synchonization. The analysis pesented in this study concentates on the caie and symbol timing ecovey methods. It is assumed that fame synchonization has been accomplished by suitable unique wod detection schemes.. Model of TDMA based Mobile Satellite Channel In the mobile satellite channel, the tansmitted signal is not only coupted by the classical AWG, but also suffes fom fading. In othe wods, the signal tansmitted to the eceive contains not only a diect LOS adio wave, but also a lage numbe of distacted, scatteed, and eflected adio waves. Even wose in uban centes o mountain aeas, the LOS wave is often bloced by obstacles, and a collection of diffeently delayed waves is all what is eceived by a mobile eath teminal. The fading in adio popagation can be categoized as lage scale fading and small scale fading 6]. Lage scale fading manifests itself as the aveage signal powe attenuation o path loss due to motion ove lage aeas; wheeas small scale fading efes to the damatic changes in the signal amplitude and phase that occu due to small changes in the spatial sepaation between the tansmitte and eceive 6]. Small scale fading can be examined in eithe time o fequency domain. In the case of time domain, coheent time T c is defined as the peiod of time ove which the fading pocess is coelated. Thus, if coheent time T c is smalle than symbol duation T s, the channel is fast fading. Othewise, the channel is small fading fo T c >T s. With 7

38 espect to the fequency domain, coheent bandwidth B c measues the fequency ange ove which the fading pocess is coelated, and it is defined as the fequency bandwidth ove which the coelation function of two samples of the channel esponse taen at the same time but diffeent fequencies falls below a suitable value. Hence, the channel is said to exhibit fequency selective fading if B c <B w, wheeas the channel is said to expeience flat fading if B c >B w, whee B w denotes the signal bandwidth 65]. Based on the mobility effect, the mobile satellite signals can be affected by multipath, shadowing o blocage, among which multipath fading is by fa the most citical poblems. Since multipath popagation leads to apid fluctuations of the phase and amplitude of the signal, it has a small-scale fading effect. Moeove, since fading is geneated by path length diffeences between the multiple ays, it can be classified as naowband o wideband depending on the chaacteistics and location of scattes with espect to the mobile. aowband implies that the channel unde consideation is sufficiently naow that its fequency esponse can be consideed flat 6], while the message bandwidth significantly exceeds the channel's coheence bandwidth in the wideband system. Thus, the fequency esponse of wideband system expeiences fequency selective fading. Fo example, the channel fo voice and low data ate Hz ange mobile satellite systems can be consideed as puely naowband, while boadband multimedia mobile adio and satellite based mobile systems infomation bandwidth is in the MHz to GHz ange have to be consideed as wideband fading lins 6]. In this eseach, only the multipath fading is focused fo the mobile satellite lin. As is nown, AWG is always the dominant channel impaiment in the mobile satellite channel. Let st denotes the tansmitted signal, which can be epesented as 8

39 s t smg t mt m. whee s m is a sequence of tansmitted data symbol, gt is the impulse esponse of tansmitte shaping filte. Then, the equivalent baseband eceived signal yt unde only AWG impaiment can be expessed as 66] y t al exp j l t] s t l ] n t. whee l t f cos t l c l l l. β epesents the Dopple shift, θ l is the lth path wave aiving angle, nt is the AWG noise with zeo mean and double sided powe spectal density PSD of, f c is the caie fequency, a l denotes the amplitude attenuation facto fo path l, whose popagation delay is given by τ l. Thus, the channel impulse esponse of the complex multipath channel is given as h, t al exp j t] l ] l. hτ,t denotes the channel esponse at time t due to an impulse applied at time t-τ, and δ is the diac delta function. If the diffeence between diffeent path delays is smalle than the symbol duation, the τ l in equation. ae all appoximately equal to τ. Then, the channel impulse esponse can be expessed as ' ' h, t al exp j t] h t.5 l 9

40 Because the Fouie tansfom of equation.5 is htexp-jπfτ, all fequency components in the eceived signal ae subject to the same complex gain ht. In this case, the eceived signal is said to exhibit flat fading. When thee is no dominant LOS lin, the amplitude of the eceived signal is Rayleigh distibuted and the phase of the eceived signal is nomal distibuted by the cental limit theoem. Thus, the pobability density functions of the amplitude and phase of ht at any time t ae given as 66] p h t x x x exp.6 p ag h t ] x, x.7 When components of a signal tavelling though diffeent paths aive at the eceive with delays significantly lage as compaed to the bit o symbol duation, the signal will undego significant amount of distotion acoss the infomation bandwidth. Then it esults in fequency selective fading o wideband fading. The impulse esponse of a wideband channel model can be witten as Q h, t l h l t ] l.8 whee h l t and τ l t denote the time vaying complex gain and excess delay of the path l. Q is the total numbe of path. Thee ae vaious methods to geneate the time vaying complex gain h l t, such as filteing two zeo mean independent white Gaussian noise with same vaiance. In addition, Cla poposed a simple and efficient method to geneate h l t, which can be expessed as 65] h v j t cos j i i j i j t cos i l t ie e ie e i i.9

41 whee λ and f epesents the wavelength and fequency of the signal, θ i denotes the angle between the diection of mobile movement and the ith scatte, α i and φ i ae the complex amplitude and phase of the ith scatte espectively, is the appoximate numbe of scatte, and β epesents the Dopple shift. The amplitude is assumed to be constant fo all scattes in path l, then i P l. whee P l is the aveage powe fo the eceived signal lth path. Thus, the time vaying complex gain and excess delay of the path l can be expessed as 65] h t l Pl i e ji e j t cos i.. Identification of Synchonization Methods Thee ae mainly fou estimation methods used in the unnown paamete estimation: least squae estimation LSE, maximum lielihood estimation MLE, minimum mean-squae estimation MMSE, and maximum a posteioi estimation MAPE. Let ψ epesents a set of signal paametes to be estimated, the tansmitted and eceived signal is epesented by a set of L symbols, then the eceived signal in equation. can be ewitten as s n. whee,,, ] and s s, s,, s ]. LSE is tying to find the L L estimate paamete ˆ, which can minimize the squae eo of 8] ˆ agmin ] ] LSE L i i s i.

42 MLE is to choose an estimate of the unnown paamete ψ, which gives the eceived signal a high pobability of occuence. Then, the maximum lielihood ML estimate is of the value which maximizes the conditional pobability density function p. Fo example, the condition pobability is defined fom equation. p i si ] exp L L i. The ML estimate is given by 6] ˆ ag max p ] MLE.5 It is obviously to see that the ML estimate also needs to minimize the tem L i i s i ]. Theefoe, ML estimate and LS estimate ae equivalent. MMSE descibes the appoach which minimizes the mean squae eo MSE, which is given as 6] MSE E ˆ ] Thus, the MMSE is the estimate value that minimize the aveage function of ˆ agmin ˆ MMSE E p ].6.7 whee p is a posteioi pobability density function of andom paamete ψ with the given eceived signal. Fom the Bayesian theoem, the conditional pobability density function p can be denoted as 6] p p p p.8