Técnicas de diversidad para sistemas de DVB de última generación.

º Tesis Doctoral Ingenería de Telecomunicación Autor: Darío Pérez-Calderón Rodríguez Director: Vicente Baena Lecuyer Técnicas de diversidad para sistemas de DVB de última generación Dep. Ingeniería Electrónica Escuela Técnica Superior de Ingeniería Universidad de Sevilla 2017 i ii Tesis Doctoral Ingeniería de Telecomunicación Técnicas de diversidad para sistemas de DVB de última generación Autor: Darío Alfonso Pérez-Calderón Rodríguez Director: Vicente Baena Lecuyer Profesor titular Dep. de Ingeniería Electrónica Escuela Técnica Superior de Ingeniería Universidad de Sevilla Sevilla, 2017 PhD Dissertation Diversity Techniques for DVB cutting-edge systems Dar´ıo Alfonso P´erez-Calder´on Rodr´ıguez Departamento de Ingenier´ıa Electr´onica Universidad de Sevilla Tutor: Vicente Baena Lecuyer PhD Dissertation 24-05-2016 Abstract After the Analog black out (in Spain this happened in April 2010), the digital TV era started officially. In Europe the Digital Video Broadcasting (DVB) Project leaded the standardization process that would end with a series of worldwide accepted standards as DVB Terrestrial (DVB-T). The standardization process of the different standards allowed to develop cutting edge technologies to be applied in them making this family of standards state of the art ones in the date they were released. The technical content in this PhD. dissertation was developed while participating in the standardization process of Digital Video Broadcasting Second generation Terrestrial (DVB-T2) and Digital Video Broadcasting Next Generation Hand-held (DVB-NGH). The dissertation is focused in the diversity techniques applied in the aforementioned standards, Rotated Constellations in DVB-T2 and Multiple Input - Multiple Output (MIMO) schemes in DVB-NGH. The gain that can be obtained by applying these techniques in their corresponding standards and also the drawbacks that can be found. As main contribution, several techniques to reduce the complexity introduced in the system by the usage of Rotated Constellations and MIMO are presented. All the techniques presented are focused on the detection process of the transmitted data and look for a trade off between performance loss and hardware complexity reduction, that is a critical aspect when regarding to mobile systems as DVB-NGH. To my parents, Dar´ıo and Margarita, and sister, Margarita, who made me to get to be who I am. To my wife Lena, whose love enlightened even the darkest of my days. To my son Daniel,the first, and who I hope one day can read and probably understand this. To my daughter Aurora, the second and last by now, who also I hope can read and understand what I am writing here and if necessary explain to Daniel ... Acknowledgements There are many people who I would like to acknowledge, so just in case I forget someone in particular I would like to thank everyone that shared a moment with me these years, because from everyone I could learn a bit and step by step I could reach this moment when I am finishing my PhD. dissertation. So let’s start from the beginning, although I already dedicated this dissertation to my parents also I must thank them for my education, the values they gave to me, all that somehow made me be the stubborn being that can reach the “end of the way”. They have always been there when I needed them and also when I didn’t, I will never be able to repay all they have given to me. Thanks to my sister, yes I know I’m being repetitive but it is a must. She has been very competitive from the very beginning and I wasn’t, she made me be much more ambitious and always wanting to improve myself. Thanks to Dani, not my son, he has been my friend since I was 6, and still he is! (Some raindrops have fallen since then, also my hair!). Thanks to him I started loving sports and started playing handball, he also was there in difficult moments and of course in many more good ones. Thanks to Juanma, we also know each other since we were six, but our friendship really bloomed when we went to the university, we studied together and lived in the same room for 5 years, he was my example to follow in many aspects, I really admire him. I wouldn’t say he is just my friend he is apart from some ADN incompatibilities like my older brother (although he is younger than me). Thanks to N´u˜nez, we have lived a lot of magical moments together (we have crafted together our own armors and weapons!, not everyone has performed together as Saint Seiya characters), spent hours and hours playing countless games and also he has been there always I have needed him, and it feels also like he were my younger brother. Thanks to Fran, one of the most innocent (in the good sense of the word) people I have ever met and probably one of the most brilliant one. He could stand a year living with me, and that is a difficult task. Thanks to Vicente, that trusted in me and let me work with a wonderful team. I have learned a lot from him, and in many aspects is an example to follow if I want to get where I really want to be in my life. Thanks to Jos´e, a really kind heart person with whom I have spent who knows how many hours in these years, countless parties, fairs, and funny moments that I will barely ever forget. And also I have learned a lot from him not only about engineering. The same I could say about Patricio, but with the advantage that foreigners can understand him when speaking Spanish. Thanks to my friends Carou, Copito, Nai, Lija, Luki, and Ana they have been always there since I knew them and without them many things wouldn’t had had any sense. They are my non blood related family the ones with whom I have shared travels, dinners, games, celebrations, countless hours ... they are just great and they made me a better person. Thanks to Cinta and Jorge, that have helped me in my academic way (as teachers and after as team mates). Thanks thanks to Luismi, Antonio, Sebas, Vito, Elo, Jos´e Luis for the moments we shared in the Cave. Also to the new Cave generation, Ignazio, Inma, Lidia, Abril, ... they make much more enjoying the workday. And at last but not least, indeed not even close, thanks to my wife, Acronym list 3GPP Third Generation Partnership Project ACE Active Constellation Extension AWGN Additive White Gaussian Noise BCH Bose-Chaudhuri-Hocquengham BER Bit Error Rate BICM Bit Interleaving Coding and Modulation CBC Complement Block Coding CDMA Code Division Multiple Access CNR Carrier to Noise Ratio COFDM Coded Orthogonal Frequency Division Multiplexion CP Cyclic Prefix CPE Continual Pilots Estimation CR Coding Rate CSP Common Simulation Platform DBPSK Differential Binary Phase Shift Keying xi ACRONYM LIST DFT Discrete Fourier Transform DV B Digital Video Broadcasting DV B −CDigital Video Broadcasting Cable DV B −HDigital Video Broadcasting for Handheld DV B −NGH Digital Video Broadcasting Next Generation for Handhelds DV B −SDigital Video Broadcasting by Satellite DV B −S2Second Generation Digital Video Broadcasting Satellite DV B −SH Digital Video Broadcasting Satellite Handheld DV B −TDigital Video Broadcasting Terrestrial DV B −T2Second Generation Digital Video Broadcasting Terrestrial ELG European Launch Group eSFN enhanced Single Frequency Network eSM enhanced Spatial Multiplexing ETSI European Telecommunications Standards Institute FEC Forward Error Correction FFT Fast Fourier Transform FURIA Futura Red Integrada Audiovisual GI Guard Interval GIE Grupo de Ingenier´ıa Electr´onica HD High Definition HDL Hardware Description Language HDTV Hihgh Definition Televsion xii ACRONYM LIST HLSI Hierarchical Local Service Insertion ICI Inter Carrier Interference IFFT Inverse Fast Fourier Transform ISI Inter Symbol Interference JRC Journal Citation Report LDPC Low Density Parity Check LLR Log Likelihood Ratio LTE Long Term Evolution MC −CDMA Multi Carrier Code Division Multiple Access CBC Modified Complement Block Coding MFN Multiple Frequency Network MIMO Multiple Input - Multiple Output MISO Multiple Input Single Output ML Maximum Likelihood MoU Memorandum of Understanding MPE Multi Protocol Encapsulation OFDM Orthogonal Frequency Division Multiplexing OLSI Orthogonal Local Service Insertion PAM Pulse Amplitude Modulation PAPR Peak to Average Power Ratio PDF Probability Density Function PH Phase Hopping xiii ACRONYM LIST PhD Philosophae Doctor PLP Physical Layer Pipes PSK Phase Shift Keying QQuadrature component QAM Quadrature Amplitude Modulation QPSK Quadrature Pulse Shift Keying RF Radio Frequency RME Rayleigh Memoryless with Erasures RMSE Root Mean Square Error RQD Rotated Constellation and cyclic Q Delay rSM rotated Spatial Multiplexing SBC Simple Block Coding SC Single Carrier SCOFDM Single Carrier Orthogonal Frequency Division Multiplexion SDTV Standard Definition Television SFN Single Frequency Network SIMO Single Input - Multiple Output MISO Single Input Single Output SM Spatial Multiplexing SMS Short Message Service SNR Signal to Noise Ratio SSD Signal Space Diversity xiv ACRONYM LIST TDM Time Division Multiplexion TDT Terrestrial Digital Television TFS Time Frequency Slicing TM Technical Module TR Tone Reservation TV Television UHF Ultra High Frequency UMTS Universal Mobile Telecommunications System XPD Cross Polar Discrimination xv Figure list 2.1 Cyclic Prefix insertion . . . . . . . . . . . . . . . . . . . . . . . . 17 2.2 DVB-T2 block diagram . . . . . . . . . . . . . . . . . . . . . . . . 20 2.3 DVB-T2 block diagram . . . . . . . . . . . . . . . . . . . . . . . . 21 2.4 Alamouti MISO transmission scheme . . . . . . . . . . . . . . . . 22 2.5 Preamble structure in the DVB-T2 frame . . . . . . . . . . . . . . 23 2.6 PP7. Pilot pattern used in DVB-T . . . . . . . . . . . . . . . . . 24 2.7 PP3 .................................. 24 2.8 256-QAM constellation . . . . . . . . . . . . . . . . . . . . . . . . 25 2.9 QPSK constellation, rotated QPSK, and RQD QPSK . . . . . . . 26 2.10 Overhead reduction example . . . . . . . . . . . . . . . . . . . . . 27 2.11 ACE technique for 16-QAM . . . . . . . . . . . . . . . . . . . . . 28 2.12 DVB-NGH base profile BICM . . . . . . . . . . . . . . . . . . . . 31 2.13 Hierarchical modulation scenario example . . . . . . . . . . . . . 33 3.1 Original 16-QAM constellation(a) and rotated version (b) . . . . . 39 3.2 Virtual constellation after the RQD process . . . . . . . . . . . . 40 3.3 DVB-T2 transmission chain block scheme . . . . . . . . . . . . . . 42 3.4 Demapperscheme........................... 44 3.5 Iterative demapper scheme . . . . . . . . . . . . . . . . . . . . . . 47 3.6 Two possible candidates in a genie aided demapping process . . . 48 3.7 Rotated and not rotated constellation performance comparison in AWGN channel for QPSK constellation . . . . . . . . . . . . . . . 50 xvi FIGURES 3.8 Rotated and not rotated constellation performance comparison in AWGN channel for 256-QAM constellation . . . . . . . . . . . . . 51 3.9 Rotated and not rotated constellation performance comparison in Rayleigh channel for QPSK constellation . . . . . . . . . . . . . . 52 3.10 Rotated and not rotated constellation performance comparison in Rayleigh channel for 16-QAM constellation . . . . . . . . . . . . . 52 3.11 Rotated and not rotated constellation performance comparison in Rayleigh channel for 64-QAM constellation . . . . . . . . . . . . . 53 3.12 Rotated and not rotated constellation performance comparison in Rayleigh channel for 256-QAM constellation . . . . . . . . . . . . 53 3.13 Rotated and not rotated constellation performance comparison in RME channel for QPSK constellation . . . . . . . . . . . . . . . . 55 3.14 Rotated and not rotated constellation performance comparison in RME channel for 16-QAM constellation . . . . . . . . . . . . . . . 55 3.15 Rotated and not rotated constellation performance comparison in RME channel for 64-QAM constellation . . . . . . . . . . . . . . . 56 3.16 Rotated and not rotated constellation performance comparison in RME channel for 256-QAM constellation . . . . . . . . . . . . . . 56 3.17 1rst quadrant(a), 2nd quadrant(b) and 4th quadrant(c) histograms 59 3.18 Used subsets for the 256-QAM constellation . . . . . . . . . . . . 61 3.19 Performance of the studied subsets for 3/4 LDPC code rate . . . . 61 3.20 Performance loss for each subset and coding rate . . . . . . . . . . 62 3.21 Subquadrants division . . . . . . . . . . . . . . . . . . . . . . . . 63 3.22 1rst subquadrant(a), 2nd subquadrant(b) and 4th subquadrant(c) histograms............................... 65 3.23 Performance for 5/6 CR in the P1 channel . . . . . . . . . . . . . 67 3.24Subsetexample ............................ 67 3.25 Performance for 3/4 CR in the P1 channel . . . . . . . . . . . . . 68 3.26 Performance for 1/2 CR in the P1 channel . . . . . . . . . . . . . 69 3.27 Performance for 5/6 code rate in the 0dB Echo channel . . . . . . 69 3.28 Performance for 3/4 code rate in the 0dB Echo channel . . . . . . 70 3.29 Performance for 1/2 code rate in the 0dB Echo channel . . . . . . 70 3.30 Performance for 5/6 CR in the RME channel . . . . . . . . . . . . 71 xvii FIGURES 3.31 Performance for 3/4 CR in the RME channel . . . . . . . . . . . . 72 3.32 Performance for 1/2 CR in the RME channel . . . . . . . . . . . . 72 3.33 Performance for 5/6 CR in DVBT-P channel with real channel estimation............................... 73 3.34 Performance for 3/4 CR in DVBT-P channel with real channel estimation............................... 74 3.35 Performance for 1/2 CR in DVBT-P channel with real channel estimation............................... 75 3.36 Method proposed in [1] . . . . . . . . . . . . . . . . . . . . . . . . 77 3.37 Decision areas for Manhattan approximation and Euclidean distances with rotated constellations . . . . . . . . . . . . . . . . . . 79 3.38 Mismatch rate between Manhattan approximation and Euclidean distance ................................ 80 3.39Proposedmethod........................... 81 3.40 BER against C/N for 256-QAM, CR=3/4, in 0dB Echo channel . 83 3.41 BER against C/N for 256-QAM, CR=3/4, in Rayleigh channel . . 83 3.42 BER against C/N for 256-QAM, CR=3/4, in RME(15%) channel 84 4.1 MIMO generic scenario . . . . . . . . . . . . . . . . . . . . . . . . 88 4.2 Spherepacking ............................ 92 4.3 Reception diversity example . . . . . . . . . . . . . . . . . . . . . 98 4.4 Transmission diversity example . . . . . . . . . . . . . . . . . . . 99 4.5 Spatial Multiplexing scheme . . . . . . . . . . . . . . . . . . . . . 100 5.1 MIMO 2x2 Alamouti scheme . . . . . . . . . . . . . . . . . . . . . 103 5.2 Transmitter block diagram . . . . . . . . . . . . . . . . . . . . . . 111 5.3 Pessimistic snapshot scenario simulation results . . . . . . . . . . 113 5.4 Typical snapshot scenario simulation results . . . . . . . . . . . . 114 5.5 Optimistic snapshot scenario simulation results . . . . . . . . . . 115 5.6 SFN indoor scenario results for 9 dB power imbalance . . . . . . . 119 5.7 SFN indoor scenario results for 6 dB power imbalance . . . . . . . 120 5.8 SFN indoor scenario results for 3 dB power imbalance . . . . . . . 120 5.9 SFN indoor scenario results for 0 dB power imbalance . . . . . . . 121 5.10 SFN outdoor scenario results for 9 dB power imbalance . . . . . . 122 xviii FIGURES 5.11 SFN outdoor scenario results for 6 dB power imbalance . . . . . . 122 5.12 SFN outdoor scenario results for 3 dB power imbalance . . . . . . 123 5.13 SFN outdoor scenario results for 0 dB power imbalance . . . . . . 123 5.14 Simulation results for 33.3 Hz Doppler . . . . . . . . . . . . . . . 127 5.15 Simulation results for 194.8 Hz Doppler . . . . . . . . . . . . . . . 128 5.16 Parameters λand γoptimization . . . . . . . . . . . . . . . . . . 133 5.17 Proposed method for MIMO demapper simplification . . . . . . . 134 5.18 Simulation results for 3/15 code rate . . . . . . . . . . . . . . . . 135 5.19 Simulation results for 4/15 code rate . . . . . . . . . . . . . . . . 136 5.20 Simulation results for 5/15 code rate . . . . . . . . . . . . . . . . 136 5.21 Simulation results for 6/15 code rate . . . . . . . . . . . . . . . . 137 5.22 Simulation results for 7/15 code rate . . . . . . . . . . . . . . . . 137 5.23 Simulation results for 8/15 code rate . . . . . . . . . . . . . . . . 138 5.24 Simulation results for 9/15 code rate . . . . . . . . . . . . . . . . 138 5.25 Simulation results for 10/15 code rate . . . . . . . . . . . . . . . . 139 5.26 Simulation results for 11/15 code rate . . . . . . . . . . . . . . . . 139 5.27 DVB-T2 block diagram . . . . . . . . . . . . . . . . . . . . . . . . 142 5.28 eTFM output (cross) and input (circle) constellations and angle evolution for even symbols . . . . . . . . . . . . . . . . . . . . . . 144 5.29 eTFM output (cross) and input (circle) constellations and angle evolution for odd symbols . . . . . . . . . . . . . . . . . . . . . . 144 5.30 RQD output (cross) and input (circle) constellations and angle evolution for even symbols . . . . . . . . . . . . . . . . . . . . . . 145 5.31 RQD output (cross) and input (circle) constellations and angle evolution for odd symbols . . . . . . . . . . . . . . . . . . . . . . 145 5.32 Theoretical SER for different diversity degrees (L) . . . . . . . . . 146 5.33 BER evolution for QPSK constellation . . . . . . . . . . . . . . . 149 5.34 BER evolution for 16-QAM constellation . . . . . . . . . . . . . . 150 5.35 BER evolution for 64-QAM constellation . . . . . . . . . . . . . . 150 5.36 Simulation results for 64-QAM and 5/6 code rate in RME 15% channel ................................ 152 5.37 Simulation results for 64-QAM and 5/6 code rate in a 0dB echo channel ................................ 153 xix FIGURES 5.38 Simulation results for QPSK and 1/2 code rate in a P1 channel . 154 xx 1. INTRODUCTION DVB-NGH MIMO Decoders, IEEE Transactions on Broadcasting, vol. 61, issue 1, pp. 84-90, 2015. •P´erez-Calder´on-Rodr´ıguez, Dar´ıo Alfonso; Baena, V.; Garc´ıa-Doblado, Jos´e; Oria-Oria, Ana Cinta; L´opez-Gonz´alez, Patricio; “Simplified Metrics Calculation for Soft Bit Detection in DVB-T2,” Radioengineering, vol. 23, issue 1, pp. 399-404, 2014. •Oria-Oria, Ana Cinta; L´opez-Gonz´alez, Patricio; Garc´ıa-Doblado, Jos´e; P´erez-Calder´on-Rodr´ıguez, Dar´ıo Alfonso; Baena, V.; “L1 Signaling Mobility Performance in DVB-T2 Receivers,” Microelectronics Journal, 2013. •Garc´ıa-Doblado, Jos´e; Oria-Oria, Ana Cinta; P´erez-Calder´on-Rodr´ıguez, Dar´ıo Alfonso; L´opez-Gonz´alez, Patricio; Baena-Lecuyer, Vicente; “Improved power efficiency for DVB-SH transmitters,” Microelectronics Journal, vol. 44, issue 10, pp. 897-903, 2013. •P´erez Calder´on Rodr´ıguez, Dar´ıo Alfonso; Baena Lecuyer, Vicente; Oria Oria, Ana Cinta; L´opez Gonz´alez, Patricio; Garc´ıa Doblado, Jos´e; “Simplified Rotated Constellation Demapper for Second Generation Terrestrial Digital Video Broadcasting,” IEEE Transactions on Broadcasting, vol.59, Issue 1, pp. 160-167, 2013. •Garc´ıa Doblado, Jos´e; Oria Oria, Ana Cinta; P´erez Calder´on Rodr´ıguez,Dar´ıo Alfonso; L´opez Gonz´alez, Patricio; Baena Lecuyer, Vicente “Improved power efficiency for DVB-SH transmitters,”; Microelectronics Journal, Elsevier, ISSN 0026-2692, 2013. •Oria Oria, Ana Cinta; Baena Lecuyer, Vicente; Granado Romero, Joaquin; Chavez Orzaez, Jorge; L´opez Gonz´alez, Patricio; Garc´ıa Doblado, Jos´e; P´erez Calder´on Rodr´ıguez, Dar´ıo Alfonso; “Reduced complexity ICI cancellation scheme for OFDM DVB-SH receivers,” Microprocessors and Microsystems.Elsevier, vol.36, Issue 5, pp. 393-401, 2012. •D. P´erez Calder´on Rodr´ıguez, Dar´ıo Alfonso; Baena Lecuyer, Vicente; Oria Oria, Ana Cinta; L´opez Gonz´alez, Patricio; Garc´ıa Doblado, Jos´e; “Rotated 5 1. INTRODUCTION constellation demapper for DVB-T2,” IEE Electronic Letters, vol.47, Issue 1, pp. 31-32, 2011. •Oria Oria, Ana Cinta; Baena Lecuyer, Vicente; Granado Romero, Joaquin; Garc´ıa Doblado, Jos´e; P´erez Calder´on Rodr´ıguez, Dar´ıo Alfonso; L´opez Gonz´alez, Patricio; Ortiz,David; “Performance Evaluation of Complementary Code Combining Diversity in DVB-SH,” Digital Signal Processing, Elsevier, Issue 6, pp. 718-724, 2011. •Garc´ıa Doblado, Jos´e; Baena Lecuyer, Vicente; Oria Oria, Ana Cinta; P´erez Calder´on Rodr´ıguez, Dar´ıo Alfonso; L´opez Gonz´alez, Patricio;“Coarse time synchronisation for DVB-T2,” IEE Electronic Letters, vol.46, Issue 11, pp. 797-799, 2010. Moreover the performed work has lead also to contributions to national and international congress: •P´erez-Calder´on-Rodr´ıguez, Dar´ıo Alfonso; Baena, V.; Garc´ıa-Doblado, Jos´e; L´opez-Gonz´alez, Patricio; Oria-Oria, Ana Cinta; Chavez-Orzaez, Jorge; “CDMA PLC communications system” in XXIX Conference on Design of Circuits and Integrated Systems, 2014, Madrid (Spain). •L´opez-Gonz´alez, Patricio; Baena, V.; Oria-Oria, Ana Cinta; Garc´ıa-Doblado, Jos´e; Chavez-Orzaez, Jorge; Granado-Romero, Joaquin; P´erez-Calder´on- Rodr´ıguez, Dar´ıo Alfonso; Dominguez-Amarillo, Eugenio; “A Power Line Sensor Network for monitoring MPPTs” in 28th Edition of Conference on Design of Circuits and Integrated Systems (DCIS), 2013, San Sebastai´an (Spain). •P´erez-Calder´on-Rodr´ıguez, Dar´ıo Alfonso; Garc´ıa-Doblado, Jos´e; L´opez- Gonz´alez, Patricio; Baena, V.; Chavez-Orzaez, Jorge; Oria-Oria, Ana Cinta; Dominguez-Amarillo, Eugenio; “CDMA based PLC system for SMGs” in 28th Edition of Conference on Design of Circuits and Integrated Systems (DCIS), 2013, San Sebastai´an (Spain). 6 1. INTRODUCTION •P´erez-Calder´on-Rodr´ıguez, Dar´ıo Alfonso; Garc´ıa-Doblado, Jos´e; L´opez- Gonz´alez, Patricio; Baena, V.; Oria-Oria, Ana Cinta; “T´ecnica de diversidad temporal-frecuencial para sistemas OFDM” in XXVIII SIMPOSIUM NACIONAL DE LA UNI ´ ON CIENT´ IFICA INTERNACIONAL DE RADIO, 2013, Santiago de Compostela (Spain). •L´opez Gonz´alez, Patricio; Oria Oria, Ana Cinta; Baena Lecuyer, Vicente;Garc´ıa Doblado, Jos´e; P´erez Calder´on Rodr´ıguez, Dar´ıo Alfonso;“L1 Signaling Mobility Performance in DVB-T2 Receivers,”in Proc. 2012 Int. Conf on Digital Circuit and IntegratedSystems (DCIS 2012), 2012, Avignon (France). •P´erez Calder´on Rodr´ıguez, Dar´ıo Alfonso; Garc´ıa Doblado, Jos´e; L´opez Gonz´alez, Patricio; Oria Oria, Ana Cinta; Baena Lecuyer, Vicente;“Simplified metrics for DVB-NGH MIMO,” in Proc. 2012 Int. Conf on Digital Circuit and Integrated Systems (DCIS 2012), 2012, Avignon (France). •P´erez Calder´on Rodr´ıguez, Dar´ıo Alfonso; Garc´ıa Doblado, Jos´e; L´opez Gonz´alez, Patricio; Oria Oria, Ana Cinta; Baena Lecuyer, Vicente; “Tecnolog´ıas clave en DVB-NGH,” in XXVII Simposium Nacional de Uni´on Cient´ıfica Internacional de Radio, 2012, Elche (Espa˜na). •Garc´ıa Doblado, Jos´e; P´erez Calder´on Rodr´ıguez, Dar´ıo Alfonso; L´opez Gonz´alez, Patricio; Baena Lecuyer, Vicente; “Improved power efficiency for DVB-SH transmitters,” in Proc. 2011 Int. Conf on Digital Circuit and Integrated Systems (DCIS 2011), 2011, Albufeira (Portugal). •L´opez-Gonz´alez, Patricio; Baena-Lecuyer, Vicente; Oria-Oria, Ana Cinta; P´erez-Calder´on-Rodr´ıguez, Dar´ıo Alfonso; “M-binary Turbo Decoder Acceleration using CUDA,” in Proc. 2011 Int. Conf on Digital Circuit and Integrated Systems (DCIS 2011), 2011, Albufeira (Portugal). •P´erez Calder´on Rodr´ıguez, Dar´ıo Alfonso; Garc´ıa Doblado, Jos´e; L´opez Gonz´alez, Patricio; Oria Oria, Ana Cinta; Baena Lecuyer, Vicente; “Codificaci´on MIMO para NGH,” in XXVI Simposium Nacional de Uni´on Cient´ıfica Internacional de Radio, 2011, Legan´es (Espa˜na). 7 1. INTRODUCTION •Oria Oria, Ana Cinta; L´opez Gonz´alez, Patricio; Garc´ıa Doblado, Jos´e; P´erez Calder´on Rodr´ıguez,Dar´ıo Alfonso; Baena Lecuyer, Vicente; Ortiz, David; “Reduced complexity ICI cancellation scheme for OFDM DVB-SH receivers,” in Proc. 2010 Int. Conf on Digital Circuit and Integrated Systems (DCIS 2010), 2010, Lanzarote (Espa˜na). •P´erez Calder´on Rodr´ıguez, Dar´ıo Alfonso; Garc´ıa Doblado, Jos´e; L´opez Gonz´alez, Patricio; Oria Oria, Ana Cinta; Baena Lecuyer, Vicente; “An´alisis del ´angulo de rotaci´on en DVB-T2,” in XXV Simposium Nacional de Uni´on Cient´ıfica Internacional de Radio, 2010, Bilbao (Espa˜na). •P´erez Calder´on Rodr´ıguez, Dar´ıo Alfonso; Oria Oria, Ana Cinta; Garc´ıa Doblado, Jos´e; L´opez Gonz´alez, Patricio; Baena Lecuyer, Vicente; Lacadena, Ignacio; “Rotated constellation for DVB-T2,” in Proc. 2009 Int. Conf on Digital Circuit and Integrated Systems (DCIS 2009), 2009, Zaragoza (Espa˜na). •Oria Oria, Ana Cinta; Garc´ıa Doblado, Jos´e; L´opez Gonz´alez, Patricio; P´erez Calder´on Rodr´ıguez, Dar´ıo Alfonso; Granado Romero, Joaquin;Ortiz, David; “Complementary code combining in DVB-SH,”in Proc. 2009 Int. Conf on Digital Circuit and Integrated Systems(DCIS 2009), 2009, Zaragoza (Espa˜na). Part of the work in this PhD. study has also presented in DVB-NGH standardization meetings and teleconferences: •Dar´ıo P´erez-Calder´on Rodr´ıguez, Vicente Baena Lecuyer, Carlos Pardo, ”SFNs”, DVB Document TM-NGH373, August 2010. •Dar´ıo P´erez-Calder´on Rodriguez, “AICIA SFNs Power Imbalance” DVB Document TM-NGH419, September 2010. •Dar´ıo P´erez-Calder´on Rodr´ıguez, Vicente Baena Lecuyer, “AICIA NGH 15 channel model results” DVB Document TM-NGH656, February 2011. •Dar´ıo P´erez-Calder´on Rodr´ıguez, “AICIA results high correlation” DVB Document TM-NGH731, February 2011. 8 1. INTRODUCTION 1.6 Projects and contracts During the PhD. studies the candidate has participated in the following projects and contracts: •Name: TECNOPORT 2025 Code: 0452/2284 ´ Ambito: Nacional Manager: Torralba-Silgado, Antonio Jesus Beginning date: 15/07/2014 End date: 15/09/2015 Quantity: AC6.374.310,10 •Name: CENIT SINTON´ IA/INTEGRASYS Code: PI-0583/2009 Project manager: Baena-Lecuyer, Vicente Beginning date: 23/11/2009 End date: 30/11/2011 Quantity: AC16.000,00 •Name: FURIA 3: FUTURA RED INTEGRADA Code: SN-0745/2009 Project manager: Baena-Lecuyer, Vicente Beginning date: 01/01/2009 End date: 31/12/2010 Quantity: AC113.953,00 •Name: INTEGRASYS/CENIT INTEGRA Code: PI-0692/2009 Project manager: Baena-Lecuyer, Vicente 9 1. INTRODUCTION Beginning date: 01/12/2008 End date: 31/12/2011 Quantity: AC44.400,00 •Name: TID-DVB SIDSA Code: PI-0693/2009 Project manager: Baena-Lecuyer, Vicente Beginning date: 01/01/2008 End date: 31/12/2010 Quantity: AC450.000,00 •Project: EVECOM-COMUNICACIONES PLC PARA RECARGA DE VEH´ ICULOS EL´ ECTRICOS code: PI12/1225 Project range: Auton´omica Founding program: Junta de Andaluc´ıa. Proyectos de aplicaci´on de conocimiento 2012 Project manager: Baena, V. Beginning date: 30/06/2014 End date: 30/06/2015 Quantity: AC32.640,68 •Project: HELIOS Code: PI-57086 Project range: Auton´omica Founding program: Incentivos de la Junta de Andaluc´ıa. Proyectos de Aplicaci´on de Conocimiento Founding entity: Junta de Andaluc´ıa Project manager: Baena, V. 10 1. INTRODUCTION Beginning date: 03/01/2014 End date: 30/04/2015 Quantity: AC80.446,40 •Project: EDELWEISS: DISE˜ NO DE UN SISTEMA DE COMUNICACIONES INAL´ AMBRICAS INTRA-SAT´ ELITE DE ALTA EFICIENCIA Code: TIC-7095 Project range: Auton´omica Founding program: Proyectos de Excelencia de la Junta de Andaluc´ıa Project manager: Aguirre Echanove, Miguel ´ Angel Beginning date: 01/08/2012 End date: 31/07/2015 Quantity: AC180.000,00 •Project: AYUDA SUPLEMENTARIA PLAN PROPIO. PROYECTO: ARQUITECTURAS PARA LA FUTURAGENERACI´ ON DE EST´ ANDARES DE TELEVISI´ ON DIGITAL Code: CICE Project range: Auton´omica Founding program: Plan Propio, Universidad de Sevilla Project manager: Baena-Lecuyer, Vicente Beginning date: 30/03/2011 End date: 31/12/2011 Quantity: AC2.176,90 •Project: NUEVAS ARQUITECTURAS PARA LA FUTURA GENERACI´ ON DE EST´ ANDARES DE TELEVISI´ ON DIGITAL Code: TEC2010-17029 Project range: Nacional Founding program: Otros programas del plan nacional I+D, Ministerio de Ciencia y Tecnolog´ıa 11 1. INTRODUCTION Project manager: Baena-Lecuyer, Vicente Beginning date: 01/01/2011 End date: 31/12/2013 Quantity: AC41.190,00 •Project: AYUDA A LA CONSOLIDACI´ ON DE GRUPOS DE INVESTIGACI´ ON Code: 2009/TIC-192 Project range: Autonomic Founding program: Otros programas, Junta de Andaluc´ıa Project manager: Torralba-Silgado, Antonio Jesus Beginning date: 01/01/2010 End date: 31/12/2011 Quantity: AC24.925,64 •Project: FURIA2: FUTURA RED INTEGRADA AUDIOVISUAL Code: TSI-020100-2008-630 Project range: National Founding program: Programa de Fomento de la Investigaci´on T´ecnica (PROFIT), Ministerio de Industria, Turismo y Comercio Project manager: Baena-Lecuyer, Vicente; Torralba-Silgado, Antonio Jesus Beginning date: 01/01/2008 End date: 31/12/2008 Quantity: AC54.000,53 1.7 Intelectual property The following patent has been developed during the PhD. studies, being one of the proposed simplifications in this dissertation. 12 1. INTRODUCTION •Decodificador de constelaciones rotadas y m´etodo de decodificaci´on Type: Patente de invenci´on, Propiedad industrial Concession date: 24/11/11 Type of protection: European Authors: Baena-Lecuyer, Vicente; P´erez-Calder´on-Rodr´ıguez, Dar´ıo Alfonso; Oria-Oria, Ana Cinta; Garc´ıa-Doblado, Jos´e; L´opez-Gonz´alez,Patricio 13 Chapter 2 DVB Standards 2.1 Introduction By the end of 1991 the broadcasting institutions, equipment manufacturers, and European regulator institutions gathered to discuss about the foundation of a group which target were to supervise the introduction of the Digital Television. This group, that was known as European Launch Group (ELG), set a frame based on the agreement, and through it the most suitable technologies to be used in this deployment were proposed. A Memorandum of Understanding (MoU) was elaborated establishing the basis for all the competitors in order to have a market based on the confidence and mutual respect. This memorandum was signed in 1993 by all the components of the ELG. Since that date, the group changed its name and adopted the one that holds nowadays: DVB Project. Today the DVB Project is the organization in charge of creating and proposing the standardization processes for the digital television. It is constituted by more than 270 institutions and enterprises from all over the world. The standards proposed by the DVB Project have had a wide international acceptance in Europe and almost every continent, just in the USA and Japan the DVB Project standards coexist with other systems. All over the world more than 500 million receivers meet DVB standards [8]. In this chapter some of the DVB standards will be briefly described, focusing on the ones in which contributions this dissertation presents are based on: DVB- 14 2. DVB STANDARDS 2.3.1.1 Additional bandwidths For professional use DVB-T2 includes 10 MHz bandwidth (communication between cameras and mobile studios), this mode is not expected to be available for traditional consumers. In order to make DVB-T2 able to operate in narrower RF channels assignations (as III and L bands) a 1.712 MHz mode (suitable for mobile applications) is also included. 2.3.1.2 Extended carrier mode (for 8K, 16K, and 32K) The higher the number of carriers of the mode the more abrupt its spectrum is attenuated outside the interest band. Also increasing the number of carriers allows to increase the amount of data in a OFDM symbol. The gain that can be obtained is between 1.4% in 8k mode and 2.1% in 32k mode. This feature is optional due to the fact that it is not compatible with the critical spectrum mask defined in [3] so it can only be applied when the normal one is accepted as shown in Figure 2.3. 3.7 3.8 3.9 44.1 4.2 Normal mask 2k 32k normal 32k extended Relative to central carrier frequency (MHz) Normalized PSD (dB) 10 0 -10 -20 -30 -40 -50 Figure 2.3: DVB-T2 block diagram 21 2. DVB STANDARDS 2.3.1.3 Alamouti MISO DVB-T already supported SFNs, this kind of configurations can lead to a performance degradation due to the presence of similar power signals generating deep fadings. In order to overcome this problem DVB-T2 introduces Alamouti coding (as depicted in Figure 2.4). This MISO system is characterized by transmitting by one antenna an unmodified version of the original stream and by the other one a modified version, in this case the symbols are swapped in pairs (in the frequency domain) and conjugated, additionally one of them suffers a sign inversion. By performing this process at the receiver the signals can be combined in a way that the resulting SNR is higher. The extra complexity introduced by the use of this technique is negligible at the receiver side. S0, S1-S0*, S1* Tx1 Tx2 Rx h1h2 Figure 2.4: Alamouti MISO transmission scheme 22 2. DVB STANDARDS 2.3.1.4 Preamble The OFDM symbols of a DVB-T2 signal are grouped in frames that start with a preamble composed by several special symbols, P1 and P2. DVB-T2 Frame P1 P2 Data Symbols Figure 2.5: Preamble structure in the DVB-T2 frame The P1 symbol is used for temporal and frequency synchronization tasks. It is modulated with a BPSK constellation and guard intervals are included at the beginning and the end of the symbol. After the P1 symbol there are one or more P2 symbols (depending on the FFT size). These symbols carry the static, dynamic and configurable signalling information for the physical layer. The first group of bits have a predefined codification and modulation, the rest of the bits can be modulated using QPSK, 16-QAM or 64-QAM but have a fixed code rate of 1/2. P2 symbols can also contain data from the common PLP or the data PLPs. 2.3.1.5 Pilot patterns As in DVB-T, pilot carriers are defined in DVB-T2 for different purposes (channel estimation, synchronization, ...). Among them three different types can be distinguished: scattered and continual. Scattered pilots have defined phase and magnitude and are mainly used for channel estimation. These pilots are inserted into the DVB-T2 signal in regular time and frequency intervals, as explained in [3]. DVB-T used a fixed scattered pilot pattern, DVB-T2 offers a higher flexibility having 8 available scattered pilot patterns. The usage of the different pilot patterns depends on the FFT size and the guard interval selected, having as main target to reduce the pilot overhead maintaining a good channel estimation quality. In the example shown in Fig- 23 2. DVB STANDARDS ure 2.6 and Figure 2.7 it can be seen how the pilot overhead is reduced from 8% to a 4% by using the pattern named PP3 [3] instead of PP7 with a guard interval of 1/8. One of the pilot patterns, PP8, is conceived to be used by receivers with data based channel estimation. Symbol (time) Carrier (frequency) Figure 2.6: PP7. Pilot pattern used in DVB-T Symbol (time) Carrier (frequency) Figure 2.7: PP3 For continual pilots the overhead percentage reduction depends also on the FFT size, leading to a reduction that varies from 2.7% for the 8K mode to 0.7% for the 32K mode, without affecting the performance of the fine frequency synchronization neither of the detection algorithms based on Continual Pilots Estimation (CPE). 2.3.1.6 High order constellation In DVB-T the higher order constellation was 64-QAM, having a rate of 6 bits per symbol and carrier. In DVB-T2, the use of 256-QAM, depicted in Figure 2.8, 24 2. DVB STANDARDS increases this rate to 8 bits per symbol and carrier, representing an increase of a 33% in spectral efficiency and capacity for a given code rate. This constellation will require a higher SNR (about 5dB higher depending on the propagation scenario and the code rate used in the FEC). 00000000 00000001 00000010 00000011 00000101 00000100 00010100 00010101 00010001 00010000 00000111 00000110 00010110 00010111 00010011 00010010 00001010 00001011 00001111 00001110 00011110 00011111 00011011 00011010 00001000 00001001 00001101 00001100 00011100 00011101 00011001 00011000 00101000 00101001 00101101 00101100 00111100 00111101 00111001 00111000 00101010 00101011 00101111 00111110 00111111 00111011 00111010 00101110 00100010 00100011 00100111 00110110 00110111 00110011 00110010 00100110 00100000 00100001 00100101 00110110 00110101 00110001 00110000 00100100 Figure 2.8: 256-QAM constellation As the rest of constellations the 256-QAM follows a Gray mapping in order to minimize the errors while detecting the signal, this can be observed in Figure 2.8 for the first quadrant, the rest of the quadrants are symmetrically distributed (the two most significant bits determine the quadrant where the point lies). 2.3.1.7 Rotated Constellations Rotated constellations was one of the promising innovations introduced in DVB- T2. This technique consists in applying a rotation in the complex plane to the original QAM constellation and after to perform a cyclic delay in the quadrature 25 2. DVB STANDARDS component (Q), that’s why it is known as Rotated Constellation and cyclic Q delay (RQD). The shape of the constellation after the rotation process and the cyclic Q delay can be observed in Figure 2.9. Φ QPSK Rotated QPSK RQD QPSK Figure 2.9: QPSK constellation, rotated QPSK, and RQD QPSK Once the constellation is rotated each component (in phase and quadrature) has enough information to determine what point was transmitted. However if the carrier is attenuated or erased by a fading event the information is lost, that’s why a cyclic Q delay is performed, if one of the components is erased the remaining one, that is transported by another subcarrier, will be enough at the receiver side to perform a correct detection. This technique will be widely studied in chapter 3, being one of the main topics in this PhD. dissertation. 2.3.1.8 16K and 32K FFT sizes Increasing the FFT size leads to a smaller spacing between carriers in the frequency domain and to a longer time duration of the OFDM symbol. The first consequence of a higher FFT size is a higher sensibility to the Doppler dispersion. This makes high FFT sizes less suited for mobile reception. The higher time duration of the OFDM symbol brings an advantage when adding the guard 26 2. DVB STANDARDS interval, for a given guard interval duration the guard interval fraction is smaller when the symbol is longer in the time domain. As an example Figure 2.10 shows two different time duration symbols with the same guard interval duration. The overhead reduction in DVB-T2 goes from 2.3% to 17.6%, note that this increases the effective data rate in the same percentage. 8k symbol 32k symbol GI GI 25% overhead 6% overhead Figure 2.10: Overhead reduction example 2.3.1.9 Forward error correction, LDPC/BCH FEC codes applied in DVB-T were based in convolutional and Reed-Solomon codes. DVB-T2 applies the codes proposed by Gallager [12], the LDPC. Before the LDPC encoding, a BCH coding is also applied. The LDPC codes have a more complex physical implementation than the convolutional codes but the correcting performance is much higher. 2.3.1.10 Interleaving stages (time, bit, cell, and frequency) The main target of the interleaving stages is to scatter the information along the frame in the different domains (time, bit, cell, and frequency) so impulsive noise or selective frequency fading don’t affect to long consecutive bit sequences of the original data stream. The most important addition in DVB-T2 is the time interleaver, protecting the data stream from impulsive noise and temporal fading in the propagation channel. 2.3.1.11 PAPR reduction techniques A high Peak to Average Power Ratio (PAPR) can decrease the efficiency of the RF power amplifier. DVB-T2 introduces two techniques to reduce this undesired 27 2. DVB STANDARDS effect in the OFDM signal, ACE (Active Constellation Extension) and tone reservation. These techniques lead to a considerable reduction of the PAPR having as main drawback a slight increase of the average transmitted power (in the case of ACE) or a 1% capacity loss (in the case of tone reservation). ACE (Active Constellation Extension) reduces PAPR by extending the external points of the constellation in a determinate way, as shown in Figure 2.11. Tone reservation reduces the peaks of the signal in the time domain by using reserved carriers in frequency. Both techniques are complementary, so they can be combined. Figure 2.11: ACE technique for 16-QAM 2.4 DVB-NGH Standards as DVB-H [6] and DVB-SH [13] tried to bring the television to handheld devices in the last two decades, but the lack of a clear business model and the high deployment costs made these standards not to succeed as expected. The changes in the use habits of the consumers of technological products, tablets and smartphones mainly, have renewed the interest in multimedia broad- 28 2. DVB STANDARDS casting for mobile devices. This trend has impulsed the appearance of new technologies that can cope with the consumer’s demands. Digital Video Broadcasting Next generation for Handhelds (DVB-NGH [5]) evolves from DVB-T2 [3] to deal with the demanding mobile environment in the same way that DVB-H did with DVB-T. DVB-T2 is still a reference in terms of robustness, flexibility, and spectral efficiency (offering up to 50% more than any other system existing nowadays), this is the reason why DVB-NGH takes it as a basis. The main target of this standard is to become a reference for mobile broadcasting; in order to achieve this, DVB-NGH is the first mobile broadcasting system using a Multiple Input-Multiple Output (MIMO) scheme in one of its profiles surpassing in this way the Shannon capacity limit for traditional communication systems with only one transmitter/receiver antenna. The standardization process of DVB-NGH started in 2010 and finished in 2012, but even surpassing nowadays systems performance its commercial success is not clear, there are no plans for a commercial implementation in the moment of writing this dissertation. However, whether it commercially succeeds or not, the usage of state of the art technologies applied in DVB-NGH will make it a reference standard for future generations of broadcasting systems. DVB-NGH presents 4 profiles, base terrestrial, MIMO terrestrial, hybrid terrestrial/satellite, and hybrid MIMO terrestrial/satellite. The only mandatory profile is the base one, based on the mobile profile of DVB-T2, known as DVB-T2 Lite (and developed during the standardization process of DVB-NGH as a result of it). This profile introduces many technical novelties respecting to DVB-T2: •Lower code rates for the Forward Error Coding (FEC), down to 1/5. •Non uniform constellations for 64-QAM and 256-QAM. •4 dimension rotated constellations for QPSK. •Inter frame convolutional interleaver. •Adaptive cell quantization. •Improvement of the L1 signalling and overhead. 29 2. DVB STANDARDS •Local content insertion techniques (hierarchical and orthogonal). •New distributed MISO technique known as enhanced SFN (eSFN). The base profile provides a complete description of the system, the other three only remark the differences with the base one. The MIMO profile describes a 2x2 system with cross-polarization, allowing spatial multiplexing through the pairs of antennas at the transmitter and the receiver side. The main contributions of this profile are the use of a coding technique called enhanced Spatial Multiplexing (eSM) and a new bit interleaver optimized for iterative decoding. The hybrid profile allows to complement the terrestrial coverage with a satellite component in the S or L bands. A long time interleaver is added (of about 10 seconds) and Single Carrier OFDM (SC-OFDM) is used. 2.4.1 Base profile As mentioned previously, DVB-NGH is based in DVB-T2, introducing robustness and spectral efficiency improvements.The block diagram of the base profile BICM of DVB-NGH is shown in Figure 2.12, where the main differences with DVB-T2 are remarked. As it can be observed, only the short codeword length (16K) for the LDPC is available, new and more robust code rates are adopted. The interleaving memory is halved. The rotated constellations are no longer applied to 256-QAM. 4D rotated constellations are available for QPSK; non-uniform constellations are available for 64-QAM and 256-QAM. The parity interleaver is applied to all the constellations (QPSK included). The I/Q interleaver substitutes the Q-Delay applied in DVB-T2 in order to achieve a better usage of the signal/space diversity separating the I and Q components of the same data the maximum possible in time and frequency. An improvement is introduced in the time interleaver by using a convolutional interleaver for the inter frame interleaving, maintaining the block interleaver only for intra frame interleaving. Adaptive quantization is used to optimize the interleaving memory in low order constellations. As a comparison, with the same memory as in DVB-T2, DVB-NGH can obtain 4 times deeper time interleaving. 30 Chapter 3 Diversity in DVB-T2 The constantly evolving world of communication systems represents an essential part in the nowadays society, not in vain it is called the information society. The appearance of new devices as smartphones and tablets have taken this social truth to a new stage changing the concept of communication and allowing people to stay permanently connected to a huge resource of information. In this each time more demanding environment the improvement of the communication techniques is a must. The nowadays state of the art technologies for wireless communications as 802.11(a, b, g, n), 802.16 (WiMax), UMTS (Universal Mobile Telephone System) LTE (Long Term Evolution) are in a very mature state, reaching a capacity close to the Shannon’s limit theory [2]. In order to surpass the limitations described by the information theory to this kind of systems it is necessary to use new schemes that provide new propagation paths and hence a capacity increase. The first studies in this field were carried out by Winters [15], Foschini [16] and Telatar [17], and their promising results leaded to a very high interest in this new field. MIMO techniques represent one of the main investigation lines in the aim of increasing the robustness and capacity in communication systems. Before starting with this kind of schemes it is convenient to carry out an overview of the main precedents of diversity techniques applied in DVB standards, as the one used in DVB-T2 [3] known as Rotated constellations and Q Delay (RQD). 37 3. DIVERSITY IN DVB-T2 3.1 Introduction The constellation rotation technique, originally suggested in [18], was introduced in the DVB-T2 standard. This technique was studied in [19]-[20] as a technique to improve the diversity order of BICM scheme over fading channels. Subsequently, in [21] a study about the application of this technique to advanced FEC code solutions such as turbo or LDPC was presented. On the other hand, this technology has been previously applied in different systems as MC-CDMA (Multi-Carrier Code-Division Multiple Access) [22]. This method is also known as Signal Space Diversity (SSD), since its final purpose is to increase the diversity order, that is, to achieve a redundancy in information bits of the coded modulation, to improve the receiver performance, above all in severe propagation scenarios. In this chapter, the principles of this technique are presented. Also, a study in detail of how it is implemented in DVB-T2 is carried out. Finally, some proposals will be presented in order to overcome the main drawback introduced by the use of this technique. 3.2 Rotated Constellation theory Rotated constellation technique is introduced in the DVB-T2 standard to improve the performance of the system exploiting the frequency diversity that this technique offers. The main principles of this technique consist in rotating the constellation points a certain angle in the complex plane. The aspect of the traditional Quadrature and Amplitude Modulation (QAM) constellation and the one after the rotation can be seen in Figure 3.1(a) and Figure 3.1(b) respectively. As a result of the rotation, if the angle is properly chosen, each of the projections on the imaginary and real axis of the constellation points bear all the information relative to the mmapped bits of the 2m-QAM constellation. As an example, for a 16-QAM constellation, instead of 4 projections in each axis there will be 16 as it can be followed from Figure 3.1(b). After the rotation the I and Q components of a same constellation point are separated by an interleaving process. As a result the I and Q components of the same symbol travel in different carriers and time instants. This also makes two 38 3. DIVERSITY IN DVB-T2 Q I (a) Q I (b) Figure 3.1: Original 16-QAM constellation(a) and rotated version (b) different carriers to bear the same information, as if an internal repetition code was used. In the case of a 16-QAM modulation, the resulting ”virtual” constellation obtained after the whole process is depicted in Figure 3.2. As it can be easily seen the result is a higher order irregular QAM constellation (256 constellation points), maintaining the spectral efficiency of the 16-QAM. This provides additional diversity that improves the performance when correcting errors derived from the propagation along the transmission channel. When a carrier is attenuated or erased because of a fading, the information in it can still be recovered because it travels also in another carrier. The optimum rotation angle depends on the order of the chosen modulation, the propagation scenario and the mapping. It is not possible to take into account every possible propagation scenario, hence the rotation angles applied usually are a good trade-off solution; they will not be the optimum for every propagation scenario, but will improve the system performance when compared to the usage of traditional QAM constellations. Clearly, the main advantage of the rotated constellation technique is that it allows to recover the transmitted information at the receiver’s side even in the extreme case when one of the components (I or Q) is completely erased because 39 3. DIVERSITY IN DVB-T2 Q I Figure 3.2: Virtual constellation after the RQD process of the propagation along the transmission channel. 3.2.1 Rotated Constellations in DVB-T2 In a DVB-T2 system, the block that carries out the constellation rotation is included in the BICM scheme, after the bit interleaver and before the cell and time interleavers (that further improve the redundancy obtained by applying the interleaving process to the I and Q components). The position of this block is highlighted in Figure 3.3. After rotating and interleaving the I and Q components, the resulting constellation is similar to sending a higher-order constellation (Figure 3.2), but with optimization of the spectral efficiency as stated in section 3.2. This is observed when the projections of the rotated constellation points on the I and Q axis are carried out. For example, in a M-QAM constellation, if the rotation technique is used, the number of projections on each axis is M. However, if a non-rotated M-QAM constellation is used, the number of projections is on each axis √M. In addition, as DVB-T2 specifies the use of Gray symbol map- 40 3. DIVERSITY IN DVB-T2 ping, I and Q channels of a conventional constellation can be mapped separately as two independent √M-PAM. Instead, when a rotated M-QAM constellation is applied, the result is equivalent to map two related M-PAM. DVB-T2 specification proposes to implement the interleaving process to make I and Q fade independently through a simple cell delay in only one of the two components. This idea was initially suggested in [23]. More precisely, a cyclic delay of an OFDM cell within a FEC block in the quadrature component must be applied. This cyclic Q delay must be removed in the receiver when rotated constellations are used. This is done by delaying the I component another OFDM cell. Thus, both components are aligned. The LDPC decoding requires the use of soft decisions or metrics, known as Log Likelihood Ratios (LLRs). If rotated constellations are not used, a conventional one-dimensional demapper can be applied. However, in the case of rotated constellations, the classical demapper must be replaced by a two-dimensional LLR demapper. The Implementation Guideline of DVB-T2 [24] describes the main concepts of this demapper and proposes different implementation methods, as the 2D LLR demapper applying the Max-Log approximation, or the 2D LLR demapper with iterative demapping (see section 3.3). The impact on computational complexity is very noticeable due to the need of a 2D demapper, as it will be shown in subsection 3.3.4. The constellation rotation technique improves the performance of a DVB-T2 receiver in all propagation scenarios, except in Gaussian channels: this kind of channels don’t present fading or erasures events and this makes rotated constellations to show no gain nor loss of performance. 3.2.2 Rotation angle in DVB-T2 The rotation angles applied in DVB-T2 standard are chosen trying to have a good trade off in every possible scenario, obtaining a noticeable gain respect to the case when RQD is not used. The study in which the choice of the rotation angles for DVB-T2 was based is presented in [25]. As a summary, the most important factors taken into account in the selection process are the ones listed below. 41 3. DIVERSITY IN DVB-T2 BCH/ LDPC Demux bits to cells Bit interleaver Mapper (cells to constellation) Constellation rotation and cyclic Q-delay Cell interleaver Time interleaver Cell Mapper Frequency interleaver MISO Single or multiple input streams Mode adaptation Stream adaptation Input processing BICM Frame Builder OFDM generation Group 1 RF signal Pilot insertion & dummy tone reservation IFFT IFFT Guard Interval insertion P1 symbol insertion Group 2 RF signal PAPR reduction Figure 3.3: DVB-T2 transmission chain block scheme 42 3. DIVERSITY IN DVB-T2 •Distance Product-L: gives an internal vision of the asymptotic behaviour of the BER curves without coding for channels with fadings. Is a function of the measurement of the 2D distance. •Minimum 1D distance: It has an important role in the asymptotic behavior of the uncoded BER curves in channels with fadings. •Minimum bit difference between two neighbor symbols of the constellation with a minimum 1D distance: when the mean bit difference is increased, if an error occurs at the demapper the number of erroneous bits at the output increases leading to a worse performance of the LDPC decoder. •Mean bit difference between two neighbor symbols of the constellation with a minimum 1D distance after projecting over the axis: when the mean bit difference is increased, if an error occurs at the demapper the number of erroneous bits at the output increases leading to a worse performance of the LDPC decoder. As a result of the study in [25] the rotation angles used in DVB-T2 are the ones shown in Table 3.1. Modulation QPSK 16-QAM 64-QAM 256-QAM θ(o) 29 16.8 8.6 atan(1/16) Table 3.1: Rotation angles in DVB-T2 In Table 3.1, atan(1/16) represents the arctangent of 1/16 expressed in degrees. 3.3 Demapping rotated constellations In section 3.2 the constellation rotation technique has been introduced, and its main principles explained from the transmitter point of view. In this subsection it will be studied how the detection process is carried out when RQD is used in DVB-T2. 43 3. DIVERSITY IN DVB-T2 3.3.1 Removing the cyclic delay The first step to be carried out when demapping rotated constellations is to remove the one cell cyclic delay applied to the Q component at the transmitter side. This can be easily done at the receiver by delaying the I component by a cell. By doing this, the I and Q components of the symbols generated at the transmitter’s mapper are allocated again in the same cell. 3.3.2 2D LLR demapping In the case of traditional QAM constellations with Gray mapping, approximations based on I/Q separation were available to simplify the detection process and its hardware implementation [26]. On the contrary, the use of rotated constellations requires at the receiver side to compute LLRs that are a 2 dimensions function. The scheme of the demapper is as the one presented in Figure 3.4. 2m-QAM rotated constellation demapper I Q LLR(b0) LLR(b1) LLR(bm-2) LLR(bm-1) Figure 3.4: Demapper scheme 3.3.2.1 Exact LLR calculation In order to obtain the LLRs at the output of the demapper (as shown in Figure 3.4) form the received points (I,Q) Equation 3.1 is applied, where birepresents the i-th bit of the 2m-QAM symbol and Pr(bi= 0|I, Q) the probability of the i-th bit being equal to 0 being I and Q received. LLR(bi) = ln Pr(bi= 1|I, Q) Pr(bi= 0|I, Q)(3.1) 44 3. DIVERSITY IN DVB-T2 A positive LLR indicates that the bit biis more likely to be a “1” whereas a negative LLR indicates that the bit is more likely to be a “0”. The problem is how to measure the likelihood level and then how to calculate the LLRs. A certain point of the constellation, x, is transmitted with values Ixand Qxin its real and imaginary components respectively. Ideally the same values should be found at the receiver, but in a real communications scenario this is far from being true. The carriers that bear Iand Qare affected by amplitude fading with factors ρIand ρQrespectively and noise is added, decreasing considerably the quality of the signal. Taking this into account the conditional probability density function (pdf), having transmitted x, can be expressed as shown in Equation 3.2, where σ2represents the noise variance. p(I, Q|x) = 1 2πσ2e−(I−ρIIx)2+ (Q−ρQQx)2 2σ2(3.2) Considering the reception of a particular bit, bi, if it was transmitted as “1” implies that any of the 2m−1possible symbols with bi= 1 could had been transmitted. The set of points of the constellation whose i-th bit takes the value j (“0” or “1”) will be denoted Cj i. The conditional pdf for the received values I,Qsupposing bitransmitted as “1” is given in Equation 3.3, where it is supposed that the 2mpossible constellation points are transmitted with the same probability. p(I, Q|bi= 1) = 1 2mπσ2X x∈C1 i e−(I−ρIIx)2+ (Q−ρQQx)2 2σ2(3.3) The conditional pdf for the received I,Qsupposing bitransmitted as ”0” has the same aspect as Equation 3.3 but in this case x∈C0 i. Making use of the Bayes theory and supposing the same probability for bito take the value “0” or “1”, the LLR can be expressed as in Equation 3.4. 45 3. DIVERSITY IN DVB-T2 LLR(bi) = ln Pr(bi= 1|I, Q) Pr(bi= 0|I, Q)= ln Pr(I, Q|bi= 1) Pr(I, Q|bi= 0)= = ln          1 2mπσ2X x∈C1 i e−(I−ρIIx)2+ (Q−ρQQx)2 2σ2 1 2mπσ2X x∈C0 i e−(I−ρIIx)2+ (Q−ρQQx)2 2σ2          (3.4) 3.3.2.2 Approximate LLR calculation The calculus of the LLR can be simplified by using the Max-log approximation, that consists in approximating the natural logarithm of the addition of exponentials by the natural logarithm of the maximum exponential in the sum, this is mathematically expressed in Equation 3.5. ln(ea0+ea1+... +ean)≈max i=0...k(ai) (3.5) Applying Equation 3.5 to Equation 3.4 the simplified expression shown in Equation 3.6 can be obtained. LLR(bi)≈minx∈C0 i((I−ρIIx)2+ (Q−ρQQx)2))− minx∈C1 i((I−ρIIx)2+ (Q−ρQQx)2)) (3.6) This method is already proposed in [24]. 3.3.3 Iterative demapping When iterative demapping is used, the scheme shown in Figure 3.4 must be modified to use the extrinsic bit information coming from the LDPC decoder, as shown in Figure 3.5. As it can be seen, the information obtained from the LDPC decoding is used to recalculate the LLRs in a new iteration. Genie-aided demapping represents iterative demapping when the number of 46 3. DIVERSITY IN DVB-T2 11 12 13 14 15 16 17 18 19 20 21 C/N(dB) 10-9 10-8 10-7 10-6 10-5 10-4 10-3 10-2 10-1 BER 64-QAMR 1/2 64-QAM 1/2 64-QAMR 3/4 64-QAM 3/4 64-QAMR 5/6 64-QAM 5/6 Figure 3.11: Rotated and not rotated constellation performance comparison in Rayleigh channel for 64-QAM constellation 14 16 18 20 22 24 26 C/N(dB) 10-8 10-7 10-6 10-5 10-4 10-3 10-2 10-1 BER 256-QAMR 1/2 256-QAM 1/2 256-QAMR 3/4 256-QAM 3/4 256-QAMR 5/6 256-QAM 5/6 Figure 3.12: Rotated and not rotated constellation performance comparison in Rayleigh channel for 256-QAM constellation 53 3. DIVERSITY IN DVB-T2 Table 3.3 summarizes the performance improvement that can be achieved in P1 channel measured at a BER of 10−5. A maximum gain of 1.9dB can be obtained. Constellation QPSK 16-QAM Coding rate 1/2 3/4 5/6 1/2 3/4 5/6 CNQAM+RQD 2.0 6.1 7.8 7.5 12.4 14.6 CNQAM 2.7 8.0 10.7 8.0 13.3 16.1 Gain(dB) 0.7 1.9 2.9 0.5 0.9 1.5 Constellation 64-QAM 256-QAM Coding rate 1/2 3/4 5/6 1/2 3/4 5/6 CNQAM+RQD 11.9 17.7 20.3 15.5 22.7 25.5 CNQAM 12.0 18.1 20.9 15.6 22.8 25.6 Gain(dB) 0.1 0.4 0.6 0.1 0.1 0.1 Table 3.3: Performance gain in Rayleigh channel 3.4.1.3 Simulation results for RME channel Due to the random carrier loss that emulates this channel model, the gain obtained by using rotated constellations compared to traditional QAM constellations is very appreciable, and more when the carrier lost ratio increases. In the simulations results shown below, the carrier loss ratio has been set to 0.2 (a 20% carrier loss). In these simulation results it can be observed that some configurations cannot reach the BER target of 10−5. This always applies to configurations that don’t use RQD, so when comparing them to the ones using RQD the gain will be represented as infinity. Table 3.4 summarizes the performance improvement that can be achieved in RME channel measured at a BER of 10−5. A maximum gain of 9.8dB can be obtained when the reception is still possible not using RQD. In some cases the reception cannot be performed without RQD in those cases the gain has been represented using ∞. 54 3. DIVERSITY IN DVB-T2 2 4 6 8 10 12 14 16 18 20 22 C/N (dB) 10-8 10-7 10-6 10-5 10-4 10-3 10-2 10-1 BER QPSKR 1/2 QPSK 1/2 QPSKR 3/4 QPSK 3/4 QPSKR 5/6 QPSK 5/6 Figure 3.13: Rotated and not rotated constellation performance comparison in RME channel for QPSK constellation 8 10 12 14 16 18 20 22 24 26 C/N(dB) 10-8 10-7 10-6 10-5 10-4 10-3 10-2 10-1 100 BER 16-QAMR 1/2 16-QAM 1/2 16-QAMR 3/4 16-QAM 3/4 16-QAMR 5/6 16-QAM 5/6 Figure 3.14: Rotated and not rotated constellation performance comparison in RME channel for 16-QAM constellation 55 3. DIVERSITY IN DVB-T2 12 14 16 18 20 22 24 26 28 30 32 C/N(dB) 10-7 10-6 10-5 10-4 10-3 10-2 10-1 100 BER 64-QAMR 1/2 64-QAM 1/2 64-QAMR 3/4 64-QAM 3/4 64-QAMR 5/6 64-QAM 5/6 Figure 3.15: Rotated and not rotated constellation performance comparison in RME channel for 64-QAM constellation 18 20 22 24 26 28 30 32 34 36 38 C/N(dB) 10-8 10-6 10-4 10-2 100 BER 256-QAMR 1/2 256-QAM 1/2 256-QAMR 3/4 256-QAM 3/4 256-QAMR 5/6 256-QAM 5/6 Figure 3.16: Rotated and not rotated constellation performance comparison in RME channel for 256-QAM constellation 56 3. DIVERSITY IN DVB-T2 Constellation QPSK 16-QAM Coding rate 1/2 3/4 5/6 1/2 3/4 5/6 CNQAM+RQD 2.9 7.9 10.6 8.7 15.0 18.5 CNQAM 4.4 17.7 ∞10.0 24.2 ∞ Gain(dB) 1.5 9.8 ∞1.3 9.2 ∞ Constellation 64-QAM 256-QAM Coding rate 1/2 3/4 5/6 1/2 3/4 5/6 CNQAM+RQD 13.8 21.7 26.7 18.4 28.3 35.4 CNQAM 14.6 30.2 ∞18.8 34.7 ∞ Gain(dB) 0.8 8.5 ∞0.4 6.4 ∞ Table 3.4: Performance gain in RME channel 3.4.1.4 Results summary As it has been shown in section 3.4, the use of rotated constellations only brings advantages in performance terms. In order to achieve a same BER level, if rotated constellations are used, the required C/N is lower (up to 9.72dB improvement can be obtained, when the gain can be numerically expressed). This gain can be translated into a wider coverage area or in the possibility of selecting a less robust LDPC coding rate and then achieving a higher capacity (more data is transmitted instead of redundancy). From the simulation results it can be concluded that the lower corrector capacities the selected coding rate has the higher gain can be obtained by using rotated constellations. The other factor that influences the gain achieved is the constellation order, the higher the constellation density is, the lower gain can be obtained. 3.5 RQD demapping simplification As shown in this chapter, the RQD technique adopted by the DVB-T2 standard [3] increases the robustness of the receiver in severe multipath propagation scenarios. The main drawback of the RQD technique is the high complexity of the 57 3. DIVERSITY IN DVB-T2 required LLR demapper. As shown in section 3.3, for each received point, the demapper must compute the Euclidean distances to all the M-QAM constellation points. In this section simplification methods based on different approaches will be exposed. In the first approach the number of distance to compute will be reduced based on statistical properties, whereas the second will focus on approximating the Euclidean distance calculation for a more simple function in terms of hardware implementation. 3.5.1 Constellation candidates reduction This first approach is based on reducing the amount of constellation points to which an Euclidean distance must be calculated in order to compute de LLRs needed by the LDPC decoder. 3.5.1.1 Quadrant division Depending on the quadrant in which the received point lies in the complex plane, the 2D distances are computed to a reduced subset of the entire constellation, decreasing the number of required operations and so the hardware complexity of the demapper. It will be shown that, if the subsets are carefully selected, this method will have a negligible impact on the system performance while still reducing the hardware complexity. In the case of a Max-log LLR demapper, the subsets can be easily determined analysing the histogram of the minimum 2D distances (see Figure 3.17). The histogram is constructed as follows: for all the points received in a particular quadrant (the number of received points is set to ten millions), the constellation symbols that are more likely to be at the minimum distance are determined, also the probability of occurrence of each of them. The subset for this particular quadrant will be determined by the Nsymbols with greatest occurrence probability, where Nis the number of constellation points in the subset. The resulting histogram for a 256-QAM constellation and a point belonging to the first quadrant is shown in Figure 3.17 (the same analysis can be carried out for all the quadrants). For the sake of clarity, the first, second and fourth quadrants 58 3. DIVERSITY IN DVB-T2 (a) (b) (c) Figure 3.17: 1rst quadrant(a), 2nd quadrant(b) and 4th quadrant(c) histograms 59 3. DIVERSITY IN DVB-T2 (Q1, Q2 and Q4 respectively) have been depicted separately. The histogram of the third quadrant (Q3) has been omitted because, in this case, the occurrence probability is close to 0. Consequently, in the case a point in Q1 is received, it is clear that the constellation points of Q3 may be ignored in the computation of the LLRs, which means a 25 % reduction in the number of operations. Moreover, we will see in the following section that some points of Q2 and Q4 can be also removed from the subset, having a negligible impact on the system performance. 3.5.1.1.1 Performance results The impact of the proposed demapper on the system performance has been analyzed via BER curves at the output of the LDPC decoder of a DVB-T2 system. The results have been obtained by using the DVB-T2 CSP [7]. All the simulations presented here have been obtained for a 256 QAM, a LDPC block size of 64.800 bits, ideal channel estimation, different code rates (CR), and the worst case channels defined in [24], a Rayleigh memoryless channel, setting the erasure events to a 15% of the carriers. Figure 3.18 shows the five reduced subsets that have been evaluated for the proposed demapper. These subsets correspond to different reductions in the number of required operations (see Table 3.5). The losses have been obtained comparing the BER curves to the ideal Max-log demapper performance and measuring the difference in terms of C/N at a BER of 10-4. Subset 1 2 3 4 5 Reduction in the number of operations 25% 31% 37.5% 44% 50% Table 3.5: Subset reductions(%) Figure 3.19 shows the BER curves for the five subsets and the ideal Maxlog LLR demapper in the case of a CR of 3/4. It can be seen that the subset number 2, which corresponds to a 31% reduction, has a negligible impact on the system performance. In the case of the subset number 4, a 44% reduction can be obtained, but the losses are increased to 0.5dB. 60 3. DIVERSITY IN DVB-T2 Figure 3.18: Used subsets for the 256-QAM constellation 28 28.5 29 29.5 30 30.5 31 10−4 10−3 10−2 10−1 C/N, dB BER MaxLog Subset 1 Subset 2 Subset 3 Subset 4 Subset 5 Figure 3.19: Performance of the studied subsets for 3/4 LDPC code rate 61 3. DIVERSITY IN DVB-T2 Figure 3.20 shows the losses compared to the ideal Max-log demapper for different code rates and for the five proposed subsets. For the most robust code rates, CR=1/2 and CR=2/3, the losses are always below 0.2dB for all the subsets. However, in the case of the less robust code rates, CR=4/5 and CR=5/6, only the first and second subsets have losses below 0.2dB. That is, for optimum performance, a 31% reduction in the number of operations can be obtained by using the subset number 2. For more realistic scenarios (like a Rayleigh) simulations have shown that the reduction can be even higher with no loss in system performance. 1 2 3 4 5 0 0.5 1 1.5 2 2.5 3 3.5 Subset Loss, dB CR=1/2 CR=2/3 CR=3/4 CR=4/5 CR=5/6 Figure 3.20: Performance loss for each subset and coding rate The results shown in this subsection were published in [27]. 3.5.1.2 Subquadrant division In subsubsection 3.5.1.1, depending on the quadrant in which the received point lies in the complex plane, the 2D distances are computed to a reduced subset of the entire constellation, decreasing the number of required operations and so the hardware complexity of the demapper. Making a good choice of the subsets will lead to a complexity reduction with almost no performance loss, as simulations 62 3. DIVERSITY IN DVB-T2 16.5 16.55 16.6 16.65 16.7 16.75 16.8 16.85 16.9 C/N (dB) 10-6 10-5 10-4 10-3 10-2 BER MaxLog N=36 N=34 Figure 3.26: Performance for 1/2 CR in the P1 channel 26.2 26.3 26.4 26.5 26.6 26.7 26.8 26.9 C/N (dB) 10-7 10-6 10-5 10-4 10-3 10-2 BER MaxLog N=64 N=56 N=48 N=40 N=36 N=35 N=34 Figure 3.27: Performance for 5/6 code rate in the 0dB Echo channel 69 3. DIVERSITY IN DVB-T2 23.1 23.15 23.2 23.25 23.3 23.35 23.4 23.45 C/N (dB) 10-6 10-5 10-4 10-3 10-2 BER MaxLog N=56 N=36 Figure 3.28: Performance for 3/4 code rate in the 0dB Echo channel 16.4 16.45 16.5 16.55 16.6 16.65 16.7 16.75 16.8 C/N (dB) 10-6 10-5 10-4 10-3 10-2 10-1 BER MaxLog N=56 N=36 Figure 3.29: Performance for 1/2 code rate in the 0dB Echo channel 70 3. DIVERSITY IN DVB-T2 Results for RME channel Figure 3.30 shows the BER curves of the system for different values of N and for an RME channel with 15% of erasure events. The results for the ideal maxlog demapper are also shown for comparison purposes. From this figure it can be seen that for a performance degradation under 0.1dB this channel require subsets with a minimum of 152 points. 34 34.2 34.4 34.6 34.8 35 35.2 C/N (dB) 10-6 10-5 10-4 10-3 10-2 BER MaxLog N=152 N=148 N=144 N=136 Figure 3.30: Performance for 5/6 CR in the RME channel Due to the harder propagation conditions of the RME channel, the subsets need more points in order to show a negligible performance loss. Anyway, in terms of complexity reduction, the required number of 2D Euclidean distances is still greatly reduced; a 41% reduction is obtained. Figure 3.31 and Figure 3.32 show the BER curves for the code rates 3/4 and 1/2 respectively, both for the RME channel. Each figure shows the curves for N=152 and N=144. As in the previous subsection, it can be seen that these more robust code rates show lower losses so the optimization of the subsets can be done focusing on the worst case code rate, that is, CR 5/6. As stated earlier, the RME channel is clearly the worst case propagation 71 3. DIVERSITY IN DVB-T2 28 28.1 28.2 28.3 28.4 28.5 28.6 28.7 28.8 28.9 C/N (dB) 10-6 10-5 10-4 10-3 10-2 BER MaxLog N=152 N=144 Figure 3.31: Performance for 3/4 CR in the RME channel 18.5 18.6 18.7 18.8 18.9 19 19.1 19.2 C/N (dB) 10-9 10-8 10-7 10-6 10-5 10-4 10-3 10-2 10-1 BER MaxLog N=152 N=144 Figure 3.32: Performance for 1/2 CR in the RME channel 72 3. DIVERSITY IN DVB-T2 scenario defined in [24]. In fact, it does not correspond to a real channel but it allows testing the DVB-T2 systems in very hard propagation conditions. That is, this channel is very unlikely to occur, so a low cost receiver could use the subsets found for the P1 or the 0dB echo channels. Real channel estimation effect The aim of this subsection is to show the impact of the real channel estimation on the proposed method. A 2D pilot aided channel estimator based on Wiener filters [28] has been added to the CSP. Simulations have been run for the P1 channel, different code rates (5/6, 3/4 and 1/2) and the PP1 pilot pattern [3]. Figure 3.33 shows two BER curves for CR 5/6, one for the ideal max-log demapper and one for our proposed scheme with N=36. In both curves the receiver uses the 2D channel estimator. When compared with the results of Fig. 5 (ideal channel estimation) a loss of 0.65dB can be denoted. According to [24] a performance loss of 0.41dB can be expected due to the pilot boosting, so the losses of the real channel estimator can be estimated to be 0.24dB. 26.15 26.2 26.25 26.3 26.35 26.4 26.45 26.5 26.55 26.6 26.65 C/N (dB) 10-7 10-6 10-5 10-4 10-3 10-2 BER MaxLog N=36 Figure 3.33: Performance for 5/6 CR in DVBT-P channel with real channel estimation 73 3. DIVERSITY IN DVB-T2 From Figure 3.33 it can be seen that the performance loss between the maxlog demapper and the proposed scheme is almost the same as the one with ideal channel estimation shown in Figure 3.23. The same conclusions can be drawn from Figure 3.34 and Figure 3.35 where the same curves are depicted for CR 3/4 and 1/2 respectively. 23.4 23.45 23.5 23.55 23.6 23.65 23.7 23.75 23.8 C/N (dB) 10-7 10-6 10-5 10-4 10-3 10-2 10-1 BER MaxLog N=36 Figure 3.34: Performance for 3/4 CR in DVBT-P channel with real channel estimation The results presented in this subsection were published in [29] 74 3. DIVERSITY IN DVB-T2 16.95 17 17.05 17.1 17.15 17.2 17.25 17.3 17.35 17.4 C/N (dB) 10-7 10-6 10-5 10-4 10-3 10-2 10-1 BER MaxLog N=36 Figure 3.35: Performance for 1/2 CR in DVBT-P channel with real channel estimation 3.5.2 Metrics simplification In order to reduce the complexity introduced by the use of the RQD technique, other approaches have been proposed in the literature. In [30], the authors propose to decorrelate the received signal in order to be able to detect it as two independent PAM modulations, obtaining a great complexity reduction. However, this method already shows a performance loss of 0.3 dB in some configurations for a BER level of only 10−3. The second approach is the one suggested in [27],[29] (these two explained in the pervious subsections),[1],[31],[32]. In them the authors show that, using different subsets of the original constellation, the number of distances to be calculated can be considerably reduced. Depending on the accuracy of the chosen subset the performance will be more or less affected by the reduction of the number of distances to be computed. Mathematically, the expression for the LLR metrics for the i-th bit (bi) of the received symbol (I+jQ) when the max-log approximation is used is shown in Equation 3.6. 75 3. DIVERSITY IN DVB-T2 From Equation 3.6, 2mterms like the one shown in Equation 3.8 must be calculated in the case of a 2m-QAM constellation. Note that Drepresents the Euclidean distance between the received complex value and the ideal constellation point propagated through the channel (ρIIx+jρQQx). D2= (I−ρIIx)2+ (Q−ρQQx)2) (3.8) In the DVB-T2 standard, mcan take values up to 8 in the case of a 256-QAM constellation. This means that for each received constellation point, 256 terms as the one in Equation 3.8 must be calculated, which makes the detection process very complex in terms of hardware implementation. The third approach, which is compatible with the previous one, consists in simplifying Equation 3.8 by using an approximation. This approach has been used in [1], where the authors propose to use Equation 3.9 as a simplification when calculating Equation 3.8. This approximation leads to the scheme shown in Figure 3.36, which avoids the use of multipliers for the squared terms in Equation 3.8, and leads to a negligible performance loss. D≈         max(a, b) if min(a, b)≤max(a, b) 4 max(a, b)− min(a, b)−max(a, b) 4 2if Otherwise (3.9) In Equation 3.9 a= abs(I−ρIIx) and b= abs(Q−ρQQx). In this subsection a technique that follows this approach and further reduces the hardware complexity will be shown. As it will be shown by means of simulation results, the performance loss will remain negligible. 3.5.2.1 Proposed method: Abs approximation When max-log approximation is used as shown in Equation 3.6 it is needed to find the point of the constellation that minimizes the distance between the received signal and the subset of constellation points in which i-th bit value is “0” or 76 3. DIVERSITY IN DVB-T2 abs(·) abs(·) output selector 2 bits right shift + comparator min max D a b xI II   +- 1 bit right shift m u x xQQQ   + Figure 3.36: Method proposed in [1] “1”. It can be found many distance metrics in the literature different from the traditional Euclidean distance; however all of them need to accomplish some conditions. Being D the distance between the points i and j these restrictions are: D(i, j)>= 0 (3.10) D(i, i) = 0 (3.11) D(i, j) = D(j, i) (3.12) D(i, j)<=D(i, h) + D(h, j) (3.13) Equation 3.13 expresses what is known as triangle inequality. The main target of this work is to obtain a method that can reduce the hardware implementation complexity of the soft bit decision detection as the used in standards as DVB-T2. So the proposed used distance metric needs to be as simple as possible keeping a good performance. Manhattan distance for an n-dimensional space can be formulated as indicated in Equation 3.14. D(i, j)≈ |xi1−xj1|+|xi2−xj2|+... +|xin −xjn|(3.14) 77 3. DIVERSITY IN DVB-T2 This approximation has a much simpler implementation than the traditional Euclidean one, and one of the most simple among the distance metrics that meet Equation 3.10-Equation 3.13. To reduce the complexity of the RQD demapper, the simplicity of Equation 3.8 leads us to propose using the 2 dimensions expression of the Manhattan distance, where aand brepresent the absolute value of the difference between the I and Q components of the ideal constellation point propagated through the channel and the received one respectively: D≈a+b(3.15) This approximation will be used instead of Equation 3.9 when looking for the minima of the distances needed in Equation 3.6. Focusing in the case of QAM constellations used in DVB-T2 standard it can be found that with traditional non rotated constellations the decision area for every constellation takes the shape of a square mesh when the Euclidean distance is used. When non rotated constellations are used the aforementioned grid represents also the decision regions for the Manhattan distance approximation. This means that either using the Euclidean distance or Manhattan approximation leads to the same result when looking for the constellation point that minimizes the distance with a received signal, so the performance loss is null. Note that the value of this minimum is not the same for both but the constellation point that produces it is. However in DVB-T2, the use of the rotated constellation technique changes the decision regions as shown in Figure 3.37 where the decision regions for the Euclidean distance are bounded with dashed lines and the decision regions for the Manhattan approximation are limited by continuous lines (for the sake of simplicity the same channel gain has been assumed for I and Q components, if not the shape of the regions would be not square but rectangular depending on the power gain ratio between ρIand ρQ). As it can be easily noticed, the regions for the two distance metrics are different, but they are coincident in the majority of the area. The grey marked zones represent the non coincident areas, thus the points where the Euclidean distance 78 3. DIVERSITY IN DVB-T2 Code rate 0dB echo Rayleigh RME(15%) 1/2 <0.05 <0.1<0.1 3/5 <0.05 <0.1<0.1 2/3 <0.05 <0.1<0.1 3/4 <0.05 <0.1<0.1 4/5 <0.05 <0.1<0.05 5/6 <0.05 <0.1<0.1 Table 3.9: 16-QAM performance loss(dB) 3.6 Conclusions In this chapter one of the most promising diversity techniques included in DVB- T2, RQD, has been presented. This technique has been widely explained, showing its theoretical principles and how it is implemented in DVB-T2 systems. The performance of the RQD technique has been evaluated by means of computer simulations. These simulations results show a very noticeable performance improvement as shown in section 3.4, making the communication feasible in very adverse propagation environments in which without using RQD the communication would not be possible. The main drawback of using RQD has been studied, that is, the hardware complexity increase at the receiver side. The rotation and cyclic delay implies that the information of the in phase and quadrature components are transmitted in different subcarriers, that makes both to suffer different fadings, that’s why simplifications based on decomposing real and imaginary components to have 1D distances instead of 2D ones are not suitable when rotated constellations are applied, 2D Euclidean distances must be computed. In order to reduce the complexity problem two different strategies have been followed. The first one consists in selecting a subset of points of the original constellation as candidates to compute the distances needed at the receiver side. With this approach a 78% reduction can be achieved, in terms of distances to be computed per received symbol. The second approach focuses in using an approximation when calculating the 2D distances needed at the receiver. With 85 3. DIVERSITY IN DVB-T2 the proposed method the use of multipliers is almost avoided, instead just adders are needed reducing the consumed silicon area. The two shown approaches are compatible, so they can be combined to obtain a higher complexity reduction. 86 Chapter 4 MIMO 4.1 Introduction To increase the data rate in power, bandwidth and complexity constrained systems is a very challenging target. To this aim DVB-NGH adopts one of the most promising techniques nowadays, MIMO which consists in using several transmitter and receiver antennas. The use of MIMO allows to greatly increase the available capacity for the system. The first works on this unexplored field were carried out by Winters [15], Foschini [16], and Telatar [17] and they brought a very high interest on what by the time was known as Smart Antenna Technologies. This interest was due to the high spectral efficiency that can be obtained when the propagation channel has a high dispersion level and its variations can be estimated accurately. All this promising results were the starting point for hundreds of investigations about MIMO in order to characterize theoretical and practical issues related to the wireless MIMO channels. The aforementioned high spectral efficiency is based on the fact that a channel with enough dispersion offers independent propagation paths for the information transmitted by every antenna in the system. For MIMO systems a capacity of min(M, N) channels can be achieved, where Mrepresents the number of transmitter antennas and Nthe number of receiver ones. This makes the capacity to be scaled by min(M, N) when compared with the SISO equivalent system. This capacity gain requires a scattered propagation environment so that the 87 4. MIMO channel matrix has full range and it can be estimated perfectly at the receiver side. The capacity of a MIMO channel strongly depends on the statistical and correlation properties of the propagation paths between antennas. Exhaustive works have been carried out to characterize the MIMO channel properties and develop analytical models as the ones used in the DVB-NGH standardization process [33]. The correlation of the paths between antennas varies drastically with the distance between transmitter and receiver and the Doppler dispersion. 4.2 MIMO scenario definition A generic MIMO scenario presents Mtransmitter antennas and Nreceiver antennas as depicted in Figure 4.1. h11 h21 hN1 hNM Tx1 Tx2 TxM Rx1 Rx2 RxN Figure 4.1: MIMO generic scenario In this case the channel can be expressed mathematically as a NxMmatrix. The propagation process can be then modeled in the frequency domain as shown in Equation 4.1. 88 4. MIMO y=Hx+n(4.1) Where yis a Nx1 vector representing the received signal, xaMx1 vector representing the transmitted data, naNx1 vector representing the noise at the receiver’s antenna and HaNxMmatrix representing the propagation channel. Hcan be described as follows in Equation 4.2. H=    h11 ··· h1M . . ..... . . h1N··· hNM    (4.2) Where hij models the channel response between the j-th transmitter antenna and the i-th receiver antenna. This value can be taken as an instantaneous sample of the channel in a determinate frequency and time instant. When there is multipath in the communication with a high spread because of the delays, Hchanges with the frequency. If the transmitter or the receiver are mobile elements Doppler spread will appear and the elements of Hvary with the time. If the transmitter antennas are sufficiently separated and the same occurs with the receiver ones, hij can be modeled as independent complex Gaussian variables with zero mean, unity variance (Rayleigh fading). This model is known as identically distributed Gaussian MIMO channel. In general, if the antennas are separated by more than the half of the wavelength, the channel fadings can be modeled as Gaussian random variables. From this generic scenario different systems can be derived: SISO (M= 1, N= 1), SIMO (M= 1, and any entire number different from 1 for N), and MISO (N= 1, and any entire number different from 1 for M). 4.3 MIMO channel capacity The science that studies the limits of the achievable capacity in communication systems is known as Information Theory. The basic measurement of the per- 89 4. MIMO formance of a communication process is known as channel capacity; this can be defined as the maximum data rate which allows to obtain an arbitrarily low error rate. Along this section this channel capacity will be studied for different scenarios, beginning with the simplest one, a SISO AWGN channel, that will be used as a starting point for the analysis of other fading channels which don’t have a unique definition of the channel capacity. However, several different notions of it are developed and can be used to carry out a systemic analysis of the capacity limits. These different notions can be used to distinguish the different resources available for the system planning. 4.3.1 AWGN channel capacity The first principles of the Information Theory were settled by Claude Shannon in 1948 [2] trying to characterize the the limits of a reliable communication process. Before Shannon developed his theory it was believed that the only way to reach a reliable communication in a noisy channel was to reduce the data rate, for example using a repetition code. Shannon showed that this was not right, and that using a more efficient coding a reliable communication could be obtained. Although, as stated before, it exists a maximum data rate known as channel capacity, above which the error rate cannot be reduced to zero. In order to carry out the aforementioned study the expression in Equation 4.3 will be taken as the communication system model. y[m] = x[m] + n[m] (4.3) Where x[m] and y[m] represent the system input and output respectively, n[m] is an additive white Gaussian noise with zero mean and unity standard deviation, and mthe discrete time index. 4.3.1.1 Repetition codes Using Bipolar Phase Shift Keying (BPSK) symbols, where x[m] = ±√P, being Pthe power of the symbol, the error rate can be written as Q(pP/σ2), being Qthe tail probability function of the normal standard distribution, defined as 90 4. MIMO Q(x) = R∞ x 1 √2πe −u2 2du. In order to reduce the error rate the same symbol can be transmitted Ntimes. That makes a code block with length N. The codewords would be then xA=√P[1, ..., 1]tand xB=−√P[1, ..., 1]t. These codewords have a power of NP. If vector xAis sent the received vector ywill be as described in Equation 4.4, where n= [n[1], ..., n[N]]. y=xA+n(4.4) An error will occur when yis closer to xBthan to xA. In this case, the error rate will follow the expression in Equation 4.5, where k·krepresents the module operator. Q rkxA−xBk 2σ!=Q rNP σ2!(4.5) This expression exponentially decreases with the value of N. Therefore, it can be concluded that the error rate can be reduced arbitrary just by increasing the block length. As a drawback, the data rate is reduced by an Nfactor. The data rate can be increased by using a multilevel Pulse Amplitude Modulation (PAM). Repeating a M-PAM symbol, with the symbols equispaced ±√P, the bit rate becomes log2(M)/N bits per symbol time. In this case, the error rate for the internal points of the constellation is the one shown in Equation 4.6 Q √NP (M−1)σ!(4.6) As long as the order of the M-PAM modulation increases less than √N, a reliable communication can be assured. However, the data rate is limited by log2(M)/N. Even in this way the data rate tends to zero when the block length increases. Geometrically, repetition codes align all the codewords in the same dimen- 91 4. MIMO sion. But the signal space has bigger space, in this case N. For a more efficient communication the codewords should be spread in the Ndimensions. An estimation of the maximum number of codewords that can cope with the power constraint Pcan be obtained applying the classic sphere packing problem. Because of the large numbers law, the N-dimensional received vector would be inside of a sphere of radius pN(P+σ2), so without loss of generality, only what happens inside the sphere must be studied, see Figure 4.2. If N→ ∞, it can be obtained that: 𝑁 𝑃+𝜎2 𝑁𝜎2 𝑁𝑃 Figure 4.2: Sphere packing 1 N N X n=1 n2[m]→σ2(4.7) For a high value of N, the received vector ywill lie, with high probability, inside a sphere of radius √Nσ with center x, that is, the transmitted codeword. A reliable communication can be obtained if the spheres don’t overlap. The maximum number of codewords that fit inside the sphere without overlapping is 92 4. MIMO the quotient between the volume of the sphere and the noise spheres, as shown in Equation 4.8 pN(P+σ2)N √Nσ2N(4.8) This implies that the maximum number of bits per symbol that can be reliably transmitted is the one shown in Equation 4.9 1 Nlog2 pN(P+σ2)N √Nσ2N!=1 2log21 + P σ2(4.9) The expression shown in Equation 4.9 is what is known as AWGN channel capacity. This reasoning is based only in the number of codewords that can be packed assuring a reliable communication. Taking into account that P σ2=SNR Equation 4.9 can be rewritten as in Equation 4.10. Cawgn =log2(1 + SNR)bits/s/Hz (4.10) From the engineering point of view, the difficult task is to find codes that make coding and decoding process as simple as possible with a performance close to the channel capacity. The AWGN channel capacity is probably one of the most known results of the Information Theory, but it is not more than the application of the Shannon theory to a concrete scenario. 4.3.1.2 SIMO channel A transmission scenario with a transmitter antenna and Nreceiver antennas can be described as in Equation 4.17, where ynrepresents the received signal at the n-th receiver antenna, hnthe complex gain of the path between the transmitter 93 4. MIMO antenna and the n-th receiver antenna, and nnthe aditive white Gaussian noise. yn[m] = hnx[m] + nn[m] (4.11) In order to perform the detection of the received signal it will be enough to apply Equation 4.12, where h= [h1, ..., hN]t,n= [n1, ..., nN]t, and y= [y1, ..., yN]t. ˇyn[m] := h∗y[m] = khk2x[m] + h∗n[m] (4.12) This represents an AWGN channel with a SNR at the receiver side of Pkhk2/σ2, where Prepresents the average energy per transmitted symbol, and σthe noise power. Thus, the channel capacity for a SIMO scenario can be expressed as: CSIMO =log21 + Pkhk2 σ2bits/s/Hz (4.13) The usage of multiple antennas increases the effective SNR, leading to a performance improvement in terms of signal power level. 4.3.1.3 MIMO channel capacity The aim of this section is to show how MIMO can improve the available system capacity, but to obtain a closed expression for a MIMO scenario is very complicated. Due to this, several hypothesis will be taken into account to achieve a closed expression for the MIMO channel capacity in order to compare the different performances that can be obtained with the different schemes proposed in this section. For the sake of simplicity a 2x2 MIMO channel will be used (the simplest MIMO scenario). The channel can be modelled in this case as a complex 2x2 matrix, Equation 4.14, where hij represents the attenuation between transmitter 94 4. MIMO 4.4.1.4 Conclusions In this subsection the principal advantages of the MIMO systems have been summarized, using illustrative examples with 2 transmitter and/or receiver antennas (the spatial diversity simplest case). It is important to remark that is not possible to obtain benefit of all the properties explained at the same time. This is because the spatial degrees of liberty are limited and it is needed to find a trade off between them. The optimal space time/frequency solution depends on the channel properties and on the network requirements. 101 Chapter 5 MIMO performance in DVB-NGH 5.1 Introduction In chapter 4 the basic theory of MIMO systems was briefly exposed. In this chapter the main candidate technologies proposed for DVB-NGH [5] will be presented and the performance of all of them will be studied using MATLAB based simulations. The simulation results shown in this chapter were used during the DVB-NGH standardization process in order to select the most adequate MIMO codification, eSM. The figure of merit used to compare the different MIMO codifications, SISO and SIMO will be its achievable capacity. 5.2 MIMO schemes proposed in DVB-NGH As stated in subsection 4.4.1, MIMO systems have clear advantages over the traditional SISO systems. However to obtain the maximum profit of them it is essential to apply an efficient coding to the data transmitted. In this subsection the main target is to show the different techniques that were taken as possible candidates for MIMO data coding in DVB-NGH. Some of them well known, as Golden, Silver and Alamouti codes and other newly proposed for the standard as eSM. 102 5. MIMO PERFORMANCE IN DVB-NGH 5.2.1 Alamouti MIMO This MIMO scheme was originally proposed by Siavash M. Alamouti in [14] and its MISO version was already included in DVB-T2. The aim of this technique is to obtain a higher robustness in 2x2 MIMO systems, it follows a scheme as the one shown in Figure 5.1. h0 h3 h1 Tx1 Tx2 Rx1 Rx2 S -S0 1 * S S1 0 * h2 Channel estimation combiner Channel estimation n n0 1 n n2 3 h2 h0h1 h3 S0 ^ S1 ^ Figure 5.1: MIMO 2x2 Alamouti scheme Mathematically it can be described as shown in Equation 5.1, where Xrepresents the output of the MIMO coder and s1, s2the two input M-QAM symbols. X(s1, s2) = "s1−s∗ 2 s2s∗ 1#(5.1) This code uses diversity to enhance the system robustness, but reaches half the data rate available for a MIMO 2x2 configuration. 103 5. MIMO PERFORMANCE IN DVB-NGH 5.2.2 Golden Code Golden Code [36] is a space-time code for MIMO systems with 2 transmitter and receiver antennas. The generation of the code follows the expression in Equation 5.2, where s1,s2,s3and s4represent four consecutive M-QAM constellation symbols to be transmitted, θ= (1 + √5)/2 (the golden number that gives name to the codification), and σ= 1 −θ. X(s1, s2, s3, s4) = 1 √5"(1 + σi)s1+ (θ−i)s2(1 + σi)s3+ (θ−i)s4 (i−θ)s3+ (1 + σi)s4(1 + θi)s1+ (σ−i)s2#(5.2) The main properties of this code are: •Full rank: The determinant of the difference of two codewords is always different from zero. •Full rate: The four degrees of freedom in the system are used. •No null determinant: the minimum determinant is 1/√5. •Reaches the frontier of diversity in multiplexing gain. •The spectral efficiency is 2 log2(M)bits/s/Hz. •The number of distances to be computed in ML detection is M4 5.2.3 Silver Code Silver Code [37] is a time-space code for MIMO 2x2 systems that allows a simpler detection than Golden Code. The codification of the codewords follows the expression indicated in Equation 5.3, where Xais defined in Equation 5.4, Xbin Equation 5.5 and Equation 5.6 and Tin Equation 5.7. X=Xa(s1, s2) + TXb(z1, z2) (5.3) Xa(s1, s2) = "s1−s∗ 2 s2s∗ 1#(5.4) 104 5. MIMO PERFORMANCE IN DVB-NGH Xb(z1, z2) = "z1−z∗ 2 z2z∗ 1#(5.5) "z1 z2#=U"s3 s4#=1 √7"1 + i−1+2i 1+2j1−i#"s3 s4#(5.6) T="1 0 0−1#(5.7) where sirepresents a M-QAM symbol to be transmitted. This codification technique has a very similar characteristics to the ones of the Golden Code, and the main ones are summarized below. •Full rank: The determinant of the difference of two codewords is always different from zero. •Full rate: The four degrees of freedom in the system are used. •No null determinant: the minimum determinant is 1/√7. •Reaches the frontier of diversity in multiplexing gain. •The spectral efficiency is 2 log2(M)bits/s/Hz. •Fast decoding: In the worst case the number of distances to be calculated while detecting will be 2M3for a ML detection. 5.2.4 eSM This technique is based in the well known SM (given as an example in subsubsection 4.4.1.3) and was originally proposed in [38]. This technique pretends to obtain a high multiplexing gain in uncorrelated channels and robustness in channels with a high correlation. 105 5. MIMO PERFORMANCE IN DVB-NGH The eSM scheme proposed for DVB-NGH is described in Equation 5.8. "x1 x2#=1 √1 + a2"1a a−1#"s1 s2#(5.8) In Equation 5.8, s1and s2represent normalized QAM symbols, and x1and x2 the codfied MIMO symbols. The aparameter is defined depending on the constellation used, as shown in Equation 5.9. a=   (√2 + 3 + √5)/(√2+3−√5) if QPSK+QPSK (√2 + 4)/(√2 + 2) if 16-QAM+16-QAM (5.9) Equation 5.8 can be also seen as a rotation between both of the transmitter antennas, and then to be rewritten as shown in Equation 5.10, where θrepresents the aforementioned rotation angle. "x1 x2#="cos(θ)−sin(θ) sin(θ)cos(θ)#"s1 s2#(5.10) 5.3 Effective bit per cell calculation In order to be able to represent the capacity against the noise power it is necessary to calculate the effective bits per cell for each of the possible MIMO configurations. This calculation quantifies the effective number of data in every cell taking into account the following parameters: •Bits per cell: this parameter depends on the chosen modulation and corresponds with the number of bits represented by the chosen M-QAM constellation point transmitted into the corresponding cell. The number of bits per cell is indicated in Table 5.1. •Coding rate: this parameter refers to the proportion between the total number of bits of the codeword output of the forward error correction block 106 5. MIMO PERFORMANCE IN DVB-NGH Modulation QPSK 16-QAM 64-QAM 256-QAM Bits per cell (Bitsc) 2 4 6 8 Table 5.1: Bits per cell and the data bits in it. In this case LDPC codes are used. CR =Bitsdata Bitsdata +Bitsredundancy (5.11) •BCH header: this corresponds to the redundancy bits inserted by the BCH code. The number of redundancy bits depends on the LDPC code rate, as shown in Table 5.2. Code rate Data bits Data+header bits 3/15 3072 3240 5/15 5232 5400 6/15 6312 6480 4/9 7032 7200 Table 5.2: BCH overhead •Scattered pilots: they are added for channel estimation purpose. This pilots make the amount of effective bits per cell to decrease. The pilot pattern selected for MIMO study in DVB-NGH standardization process was PP1 [3]. Depending on what kind of configuration is used (MIMO, SIMO, SISO) the correction to apply will be 1/6 or 1/12. This is because in PP1 one cell in every 12 is a pilot and then this capacity is lost. For 2x2 MIMO twice the number of pilots are needed. •Half/full rate MIMO coding (M): In the case of MIMO full rate coding the whole capacity that the channel offers is used to transmit different data, in half rate codes half of this available capacity is used to increase the robustness of the system with additional redundancy, so the capacity is 107 5. MIMO PERFORMANCE IN DVB-NGH halved in comparison. In this document the only half rate MIMO code is Alamouti. Taking into account all these parameters, the calculus of the number of effective cells is as shown in Equation 5.12. EBPC =M·Bitsc·CR·Bitsdata BitsBCHoverhead ·(1 −PP1overhead) (5.12) For all the configurations contemplated during the standardization process the calculus of the EBPC is shown in Table 5.3, Table 5.4, and Table 5.5. Modulation BPC Code rate Data bits Data+header bits EBPC QPSK 2 3/15 3072 3240 0.35 5/15 5232 5400 0.59 6/15 6312 6480 0.71 4/9 7032 7200 0.80 16-QAM 4 3/15 3072 3240 0.7 5/15 5232 5400 1.18 6/15 6312 6480 1.43 4/9 7032 7200 1.59 64-QAM 6 3/15 3072 3240 1.04 5/15 5232 5400 1.78 6/15 6312 6480 2.14 4/9 7032 7200 2.39 256-QAM 8 3/15 3072 3240 1.39 5/15 5232 5400 2.37 6/15 6312 6480 2.86 4/9 7032 7200 2.18 Table 5.3: SISO and SIMO EBPC 108 5. MIMO PERFORMANCE IN DVB-NGH Modulation BPC Code rate Data bits Data+header bits EBPC QPSK 2 3/15 3072 3240 0.32 5/15 5232 5400 0.54 6/15 6312 6480 0.65 4/9 7032 7200 0.72 16-QAM 4 3/15 3072 3240 0.63 5/15 5232 5400 1.08 6/15 6312 6480 1.30 4/9 7032 7200 1.45 64-QAM 6 3/15 3072 3240 0.95 5/15 5232 5400 1.61 6/15 6312 6480 1.95 4/9 7032 7200 2.17 256-QAM 8 3/15 3072 3240 1.26 5/15 5232 5400 2.15 6/15 6312 6480 2.60 4/9 7032 7200 2.89 Table 5.4: MIMO half rate EBPC Modulation BPC Code rate Data bits Data+header bits EBPC QPSK 2 3/15 3072 3240 0.63 5/15 5232 5400 1.08 6/15 6312 6480 1.3 4/9 7032 7200 1.45 16-QAM 4 3/15 3072 3240 1.26 5/15 5232 5400 2.15 6/15 6312 6480 2.60 4/9 7032 7200 2.89 Table 5.5: MIMO full rate EBPC 109 5. MIMO PERFORMANCE IN DVB-NGH 5.4 Performance analysis In this section the capacity curves obtained by means of computer simulations will be presented. In each subsection it will be shown first the system configuration and after the capacity curves that characterize the system performance for that given configuration. The CSP [7] was taken as a basis for the simulation platform, being modified to add the MIMO simulation capability. As a result a system as the one shown in Figure 5.2 was obtained. The simulation chain consists of 4 groups of blocks regarding the functionality that is carried out in them: data input porcessing, BICM, frame builder and OFDM signal generation. To compare SISO and MIMO it should be taken into account that in the MIMO channel generation the resulting propagation scenario is normalized to have an unity gain. Due to this fact, when one of the paths of the MIMO channel is taken to generate the SISO propagation scenario it will have a gain different from the unity (lower in this case). This makes the noise power inserted at the receiver in the SISO scenario lower than in the case of MIMO resulting in a non fair comparison. As a solution for this problem a certain noise power is inserted independently to the signal level, this is the noise power is inserted referred to a normalized signal level. In the graphics presented in this subsection it will be represented the noise power (dB) in the abscissa axis and in ordinate the effective bits per cell (calculated as shown in section 5.3). A BER target is fixed at the receiver side after the LDPC decoder (10−5generally), the noise power will be the one that provides the fixed performance level. The curves with the same token and line represent the same MIMO technique or SISO and SIMO configurations. In the following subsections the simulation results obtained for the different set of propagation scenarios will be presented, after that these results will be analysed. 110 5. MIMO PERFORMANCE IN DVB-NGH eSM vs Alamouti 2x2 EBPC Pessimistic Typical Optimistic 0.63 0.9 0.8 1.0 1.08 1.1 0.9 1.1 1.30 0.9 1.1 1.4 1.45 0.9 0.8 1.2 Table 5.8: eSM gain (dB) referred to Alamouti 2x2 eSM vs SISO EBPC Pessimistic Typical Optimistic 0.63 3.2 2.6 2.7 1.08 3.0 3.1 2.9 1.30 3.2 3.4 3.2 1.45 2.8 3.2 2.9 Table 5.9: eSM gain (dB) referred to SISO eSM vs SIMO EBPC Pessimistic Typical Optimistic 0.63 0.8 0.7 0.8 1.08 0.7 0.6 0.8 1.30 0.9 0.7 1.0 1.45 0.6 0.3 0.7 Table 5.10: eSM gain (dB) referred to SIMO 117 5. MIMO PERFORMANCE IN DVB-NGH From Table 5.6, Table 5.7,Table 5.8, Table 5.9 and Table 5.10 it can be followed that eSM offers a better performance, in terms of capacity, than the rest of the studied schemes, presenting a maximum gain over them that can be found in Table 5.11. The minimum gain observed can be found in Table 5.12. Golden Silver Alamouti SIMO SISO Max. gain (dB) 0.2 0.6 1.4 1.0 3.4 Table 5.11: eSM maximum gain (dB) Golden Silver Alamouti SIMO SISO Min. gain (dB) 0.0 0.3 0.8 0.3 2.7 Table 5.12: eSM minimum gain (dB) 5.4.2 SFNs results An SFN is the one in which several transmitters transmit the same information using the same frequency. The main target of the SFNs is the improvement of the spectral efficiency, allowing a higher performance when compared to the traditional MFNs. An SFN can increase the coverage area, this is because the total received power in the points between transmitters increases. The SFN transmissions can be considered as multipath propagation cases in very adverse conditions. The receiver receives different echoes of the same signal and the delay between them can cause severe fadings in the communication channel. In DVB-NGH the use of this kind of channels is contemplated and because of this during the standardization process the channel models described in [33] were developed. The simulation conditions in SFN scenarios were set to: 118 5. MIMO PERFORMANCE IN DVB-NGH •BER lower than 10−5after the LDPC decoder with at least 200 erroneous FEC blocks and a m´aximum of 20000 LDPC blocks for each C/N. •16k length LDPC with code rates of 1/4, 1/3, 2/5, and 1/2. •8 MHz bandwidth •Guard interval of 1/4. •4k FFT size. •Relative power imbalance between paths of 9, 6, 3, and 0 dB. In this subsection the same procedure as in the previous will be followed. First the simulation results will be presented and after they will be analysed. 5.4.2.1 SFN indoor simulation results This kind of scenario models a SFN reception inside a typical building. 012345678910 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1/σ2 Effective bits per cell SISO SIMO Alamouti 2x2 eSM Figure 5.6: SFN indoor scenario results for 9 dB power imbalance 119 5. MIMO PERFORMANCE IN DVB-NGH 012345678910 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1/σ2 Effective bits per cell SISO SIMO Alamouti 2x2 eSM Figure 5.7: SFN indoor scenario results for 6 dB power imbalance 012345678910 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1/σ2 Effective bits per cell SISO SIMO Alamouti 2x2 eSM Figure 5.8: SFN indoor scenario results for 3 dB power imbalance 120 5. MIMO PERFORMANCE IN DVB-NGH 12345678910 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1/σ2 Effective bits per cell SISO SIMO Alamouti 2x2 eSM Figure 5.9: SFN indoor scenario results for 0 dB power imbalance 5.4.2.2 SFN outdoor simulation results This kind of scenario models a SFN reception in an open area environment. 121 5. MIMO PERFORMANCE IN DVB-NGH 012345678 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1/σ2 Effective bits per cell SISO SIMO Alamouti 2x2 eSM Figure 5.10: SFN outdoor scenario results for 9 dB power imbalance 012345678 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1/σ2 Effective bits per cell SISO SIMO Alamouti 2x2 eSM Figure 5.11: SFN outdoor scenario results for 6 dB power imbalance 122 5. MIMO PERFORMANCE IN DVB-NGH 123456789 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1/σ2 Effective bits per cell SISO SIMO Alamouti 2x2 eSM Figure 5.12: SFN outdoor scenario results for 3 dB power imbalance 123456789 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1/σ2 Effective bits per cell SISO SIMO Alamouti 2x2 eSM Figure 5.13: SFN outdoor scenario results for 0 dB power imbalance 123 5. MIMO PERFORMANCE IN DVB-NGH 5.4.2.3 Results analysis All the results presented in subsubsection 5.4.2.1 and subsubsection 5.4.2.2 show a very similar behavior, eSM presents a better performance in every case. Also it is remarkable that SIMO and Alamouti 2x2 show a very similar performance in the case of not having a very big power imbalance but when the power imbalance increases SIMO offers better performance. To have a quantitative idea of the performance gain obtained in terms of 1/σ2(dB) the following tables show a comparative between the different codes for the studied cases. eSM Vs SISO Indoor PPPPPPPP P PI EBPC 0.63 1.08 1.30 1.45 9 2.30 2.53 2.72 2.51 6 2.31 2.61 2.68 2.25 3 2.33 2.59 2.77 2.34 0 2.35 2.59 2.82 2.43 Table 5.13: eSM gain (dB) referred to SISO in SFN indoor scenario eSM Vs SISO Outdoor PPPPPPPP P PI EBPC 0.63 1.08 1.30 1.45 9 0.82 1.13 1.47 1.41 6 0.89 1.21 1.57 1.41 3 0.77 1.19 1.57 1.33 0 0.84 1.37 1.67 1.43 Table 5.14: eSM gain (dB) referred to SISO in SFN outdoor scenario These tables show the difference in performance existing between eSM and the rest of configurations and codes contemplated in the DVB-NGH standardization process. The capacity is measured with a BER level of 10−5and the measurements 124 5. MIMO PERFORMANCE IN DVB-NGH eSM Vs SIMO Indoor PPPPPPPP P PI EBPC 0.63 1.08 1.30 1.45 9 0.53 0.50 0.52 0.21 6 0.53 0.40 0.42 0.01 3 0.55 0.30 0.37 -0.09 0 0.43 0.28 0.37 -0.08 Table 5.15: eSM gain (dB) referred to SIMO in SFN indoor scenario eSM Vs SIMO Outdoor PPPPPPPP P PI EBPC 0.63 1.08 1.30 1.45 9 0.23 0.63 0.78 0.53 6 0.40 0.55 0.77 0.53 3 0.42 0.55 0.82 0.51 0 0.40 0.63 0.82 0.53 Table 5.16: eSM gain (dB) referred to SIMO in SFN outdoor scenario eSM Vs Alamouti indoor PPPPPPPP P PI EBPC 0.63 1.08 1.30 1.45 9 0.80 1.00 0.90 0.70 6 0.70 0.80 0.90 0.60 3 0.80 0.70 0.90 0.50 0 0.60 0.60 0.90 0.40 Table 5.17: eSM gain (dB) referred to Alamouti 2x2 in SFN indoor scenario 125 5. MIMO PERFORMANCE IN DVB-NGH eSM Vs Alamouti outdoor PPPPPPPP P PI EBPC 0.63 1.08 1.30 1.45 9 0.80 1.00 1.30 1.20 6 0.80 1.00 1.30 1.10 3 0.70 1.00 1.30 1.00 0 0.70 1.00 1.20 0.90 Table 5.18: eSM gain (dB) referred to Alamouti 2x2 in SFN outdoor scenario are taken with different EBPC and different power imbalance between paths. The maximum gain that eSM presents is shown in Table 5.19. The minimum gain found is presented in Table 5.20. SISO SIMO Alamouti 2x2 Maximum Gain (dB) 2.82 0.82 1.3 Table 5.19: Maximum eSM gain (dB) in SFN scenarios SISO SIMO Alamouti 2x2 Minimum Gain (dB) 0.82 0.23 0.4 Table 5.20: Maximum eSM gain (dB) in SFN scenarios 5.4.3 Mobile channel model results The mobile channel used is a 6 taps model and is defined in [33]. The model takes into account effects as the polarization rotation and power asymmetry between the two used polarizations. Doppler frequencies of 33.3 and 194.8 Hz (corresponding with speeds of 60 and 350 km/h at the work frequency of 600 MHz) are contemplated, defined as representative of vehicular movements of low and high speed. The models mean to be representative in the UHF band. 126